US11409751B2

US11409751B2 - Providing answers to questions using hypothesis pruning

Info

Publication number: US11409751B2
Application number: US16/267,688
Authority: US
Inventors: Jennifer Chu-Carroll; David A. Ferrucci; David C. Gondek; Adam P. Lally; James W. Murdock, IV
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2010-09-28
Filing date: 2019-02-05
Publication date: 2022-08-09
Also published as: WO2012047532A1; EP2622428A4; CN103229120A; US20120078889A1; EP2622428A1; US20160232165A1; US9323831B2; US20190171646A1; US10216804B2; US9317586B2; US20130018876A1

Abstract

A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving a query, conducting a search through one or more data sources to identify candidate answers to the query, and providing each of the candidate answers with a preliminary score. The method further comprises filtering out any of the candidate answers with a preliminary score that does not satisfy a defined condition. The candidate answers having preliminary scores that satisfy this condition form a subset of the candidate answers. Each of the candidate answers in this subset is processed to produce further scores. A ranking function is applied to these further scores to determine a ranking for each of the candidate answers in the subset; and after this ranking function is applied, one or more of the candidate answers are selected as one or more final answers to the query.

Description

CROSS REFERENCE TO RELATED APPLICATION

This application is a continuation of copending U.S. patent application Ser. No. 15/131,650, filed Apr. 18, 2016, which is a continuation of U.S. patent application Ser. No. 13/240,140, filed Sep. 22, 2011. The entire contents and disclosures of application Ser. Nos. 15/131,650 and 13/240,140 are hereby incorporated herein by reference in their entireties.

This application claims the benefit of the filing date of U.S. provisional patent application No. 61/387,157, filed Sep. 28, 2010, the disclosure of which is hereby incorporated by reference herein in its entirety.

BACKGROUND Field of the Invention

This invention generally relates to information retrieval, and more specifically, to question answering. Even more specifically, embodiments of the invention relate to query/answer systems and methods implementing parallel analysis for providing answers to questions by generating and evaluating multiple candidate answers.

Description of the Related Art

Generally, question answering (QA) is a type of information retrieval. Given a collection of documents (such as the World Wide Web or a local collection), a QA system should be able to retrieve answers to questions posed in natural language. QA is regarded as requiring more complex natural language processing (NLP) techniques than other types of information retrieval such as document retrieval, and QA is sometimes regarded as the next step beyond search engines.

QA research attempts to deal with a wide range of question types including: fact, list, definition, how, why, hypothetical, semantically-constrained, and cross-lingual questions. Search collections vary from small local document collections, to internal organization documents, to compiled newswire reports, to the world wide web.

Closed-domain question answering deals with questions under a specific domain (for example, medicine or automotive maintenance), and can be seen as an easier task because NLP systems can exploit domain-specific knowledge frequently formalized in ontologies. Alternatively, closed-domain might refer to a situation where only a limited type of questions are accepted, such as questions asking for descriptive rather than procedural information. Open-domain question answering deals with questions about nearly everything, and can only rely on general ontologies and world knowledge. Open-domainQ/A systems, though, usually have much more data available from which to extract the answer.

Access to information is currently dominated by two paradigms: a database query that answers questions about what is in a collection of structured records; and a search that delivers a collection of document links in response to a query against a collection of unstructured data (text, html etc.).

One major challenge in such information query paradigms is to provide a computer program capable of answering factual questions based on information included in a large collection of documents (of all kinds, structured and unstructured). Such questions can range from broad such as “what are the risk of vitamin K deficiency” to narrow such as “when and where was Hillary Clinton's father born”.

User interaction with such a computer program could be either a single user-computer exchange or a multiple turn dialog between the user and the computer system. Such dialog can involve one or multiple modalities (text, voice, tactile, gesture etc.). Examples of such interaction include a situation where a cell phone user is asking a question using voice and is receiving an answer in a combination of voice, text and image (e.g. a map with a textual overlay and spoken (computer generated) explanation. Another example would be a user interacting with a video game and dismissing or accepting an answer using machine recognizable gestures or the computer generating tactile output to direct the user.

The challenge in building such a computer system is to understand the query, to find appropriate documents that might contain the answer, and to extract the correct answer to be delivered to the user. Currently, understanding the query is an open problem because computers do not have human ability to understand natural language nor do they have common sense to choose from many possible interpretations that current (very elementary) natural language understanding systems can produce.

Being able to answer factual query in one or multiple dialog turns is of potential great value as it enables real time access to accurate information. For instance, advancing the state of the art in question answering has substantial business value, since it provides a real time view of the business, its competitors, economic conditions, etc. Even if QA is in a most elementary form, it can improve productivity of information workers by orders of magnitude.

U.S. patent application Ser. No. 12/152,411, the disclosure of which is hereby incorporated herein by reference in its entirety, describes a QA system that involves generating a pool of candidate answers to a query and then processing (i.e., retrieving supporting evidence, scoring, and ranking) all of the candidate answers. This processing of all of the candidate answers may not always be necessary. In at least some cases, some of the candidate answers are unlikely to be a correct answer, and for these low quality candidate answers, it may not be worth the computational cost of searching for supporting evidence.

SUMMARY OF THE INVENTION

Embodiments of the invention provide a method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving an input query, conducting a search through one or more data sources to identify a plurality of candidate answers to the input query, and providing each of the candidate answers with a preliminary score based on one or more defined criterion. The method further comprises filtering out any of the candidate answers with a preliminary score that does not satisfy a defined condition. As a result of this filtering, the candidate answers having preliminary scores that satisfy said defined condition form a subset of the candidate answers. Each of the candidate answers in this subset is processed to produce a plurality of further scores for each of these candidate answers. A candidate ranking function is applied to these further scores to determine a ranking for each of the candidate answers in the subset of candidate answers; and after the applying this candidate ranking function, one or more of the candidate answers are selected as one or more final answers to the input query.

In an embodiment, each of the candidate answers in the subset of candidate answers is processed by using a supporting passage retrieval process to identify passages in which said each of the candidate answers occur. The subset of candidate answers is sent from the supporting passage retrieval process to a candidate ranking function module, and this module is used to apply the candidate ranking function to the candidate answers in the subset of candidate answers.

In one embodiment, any filtered out candidate answers may also be sent to the candidate ranking function, and this module may also be used to apply the candidate ranking function to any filtered out candidate answers. In this embodiment, the candidate answers selected as the one or more final answers may be selected from among the candidate answers in the subset of candidate answers as well as any filtered out candidate answers.

In an embodiment, a filtering module is used to filter out any of the candidate answers with a preliminary score that does not satisfy the defined criterion includes passing the candidate answers through a filtering module to perform said filtering. Any such filtered out candidate answers are sent to the candidate ranking function module in a manner by-passing the supporting passage retrieval process.

In one embodiment, each of the candidate answers in the subset of candidate answers is processed using a context independent candidate answer process to obtain one or more post-filtering scores for said each of the candidate answers. In an embodiment, each of the candidate answers of the subset of candidate answers are ranked based on the post-filtering scores for said each of the candidate answers.

In an embodiment, each of the candidate answers is provided with the preliminary score by using a logistics regression model to score said each of the candidate answers.

In one embodiment, any of the candidate answers with a preliminary score below a defined threshold are filtered out. In an embodiment, this defined threshold is established to obtain a desired tradeoff between computational cost and a quality metric for the final answer. In an embodiment, the defined threshold is determined by running a test on a given set of data to obtain this desired tradeoff.

In an embodiment of the invention, only a subset of the candidate answers are used for finding supporting evidence. This subset may not include all of the candidate answers.

The subset of candidate answers to use for supporting passage retrieval is identified by the following procedure:

- 1) Run context independent scorers on the candidate answers (since those scorers do not require supporting passages).
- 2) Use a scoring function (e.g., a logistic regression model) to score each candidate answer.
- 3) Candidate answers with very low scores are omitted from supporting passage retrieval; i.e., they are considered to be not good enough to be worth the computational cost of searching for supporting evidence. The threshold used to identify low-scoring answers may be obtained by optimizing over a held-out data set, e.g., performing a parameter sweep to select for an optimal tradeoff between computational cost and final answer quality metrics.
- 4) The remaining candidate answers are considered to be good enough to merit further investigation.

BRIEF DESCRIPTION OF THE DRAWINGS

The objects, features and advantages of the invention are understood within the context of the Description of Embodiments of the Invention, as set forth below. The Description of Embodiments of the Invention is understood within the context of the accompanying drawings, which form a material part of this disclosure, wherein:

FIG. 1 shows a system diagram depicting a high level logical architecture and question/answering method for an embodiment of the present invention,

FIG. 2 shows a variant of the architecture of FIG. 1, where the Evidence Gathering module includes two submodules: Supporting Passage Retrieval and Candidate Answer Scoring.

FIG. 3 shows a more detailed diagram of the Query Analysis and the Candidate Answer Generation modules of FIG. 1.

FIG. 4 shows a more detailed diagram of the Candidate Answer Scoring and the Answer Ranking Modules of FIGS. 1 and 2.

FIG. 5 is an example flow diagram depicting method steps for processing questions and providing answers according to an embodiment of the invention.

FIG. 6 depicts an aspect of a UIMA framework implementation for providing one type of analysis engine for processing CAS data structures.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

As used herein, the words “question” and “query,” and their extensions, are used interchangeably and refer to the same concept, namely a request for information. Such requests are typically expressed in an interrogative sentence, but they can also be expressed in other forms, for example as a declarative sentence providing a description of an entity of interest (where the request for the identification of the entity can be inferred from the context). “Structured information” (from “structured information sources”) is defined herein as information whose intended meaning is unambiguous and explicitly represented in the structure or format of the data (e.g., a database table). “Unstructured information” (from “unstructured information sources”) is defined herein as information whose intended meaning is only implied by its content (e.g., a natural language document). “Semi structured information” refers to data having some of its meaning explicitly represented in the format of the data, for example a portion of the document can be tagged as a “title”.

FIG. 1 shows a system diagram depicting a high-level logical architecture 10 and methodology of an embodiment of the present invention. As shown in FIG. 1, the architecture 10 includes a Query Analysis module 20 implementing functions for receiving and analyzing a user query or question. According to an embodiment of the invention, a “user” refers to a person or persons interacting with the system, and the term “user query” refers to a query (and its context) 19 posed by the user. However, it is understood other embodiments can be constructed, where the term “user” refers to a computer system 22 generating a query by mechanical means, and where the term “user query” refers to such a mechanically generated query and its context 19′. A candidate answer generation module 30 is provided to implement a search for candidate answers by traversing structured, semi structured and unstructured sources contained in a Primary Sources module 11 and in an Answer Source Knowledge Base module 21 containing collections of relations and lists extracted from primary sources. All the sources of information can be locally stored or distributed over a network, including the Internet. The Candidate Answer generation module 30 generates a plurality of output data structures containing candidate answers based upon the analysis of retrieved data. In FIG. 1, one embodiment is depicted that includes an Evidence Gathering module 50 interfacing with the primary sources 11 and knowledge base 21 for concurrently analyzing the evidence based on passages having candidate answers, and scoring each of the candidate answers as parallel processing operations.

In one embodiment, the architecture may be employed utilizing the Common Analysis System (CAS) candidate answer structures, and implementing Supporting Passage Retrieval as will be described in greater detail herein below. This processing is depicted in FIG. 2 where the Evidence Gathering module 50 comprises Supporting Passage Retrieval 40A and the Candidate Answer Scoring 40B as separate processing modules for concurrently analyzing the passages and scoring some or each of the candidate answers as parallel processing operations. The Answer Source Knowledge Base 21 may comprise one or more databases of structured or semi-structured sources (pre-computed or otherwise) comprising collections of relations (e.g., Typed Lists). In an example implementation, the Answer Source knowledge base may comprise a database stored in a memory storage system, e.g., a hard drive. An Answer Ranking module 60 provides functionality for ranking candidate answers and determining a response 99 that is returned to a user via a user's computer display interface (not shown) or a computer system 22. The response may be an answer, or an elaboration of a prior answer or a request for clarification in response to a question—when a high quality answer to the question is not found.

It is understood that skilled artisans may implement a further extension to the system shown in FIG. 1 to employ one or more modules for enabling I/O communication between a user or computer system and the system 10 according to, but not limited to, the modalities of text, audio, video, gesture, tactile input and output etc. Thus, in one embodiment, both an input query and a generated query response may be provided in accordance with one or more of multiple modalities including text, audio, image, video, tactile or gesture.

The processing depicted in FIGS. 1 and 2, may be local, on a server, or server cluster, within an enterprise, or alternately, may be distributed with or integral with or otherwise operate in conjunction with a public or privately available search engine in order to enhance the question answer functionality in the manner as described. Thus, embodiments of the invention may be provided as a computer program product comprising instructions executable by a processing device, or as a service deploying the computer program product. The architecture employs a search engine (a document retrieval system) as a part of Candidate Answer Generation module 30 which may be dedicated to the Internet, a publicly available database, a web-site (e.g., IMDB.com) or a privately available database. Databases can be stored in any storage system, e.g., a hard drive or flash memory, and can be distributed over a network or not.

As mentioned, embodiments of the invention make use of the Common Analysis System (CAS), a subsystem of the Unstructured Information Management Architecture (UIMA), that handles data exchanges between the various UIMA components, such as analysis engines and unstructured information management applications. CAS supports data modeling via a type system independent of programming language, provides data access through a powerful indexing mechanism, and provides support for creating annotations on text data, such as described in (http://www.research.ibm.com/journal/sj/433/gotz.html) incorporated by reference as if set forth herein. CAS also allows for multiple definitions of the linkage between a document and its annotations, as is useful for the analysis of images, video, or other non-textual modalities.

In one embodiment, the UIMA may be provided as middleware for the effective management and interchange of unstructured information over a wide array of information sources. The architecture generally includes a search engine, data storage, analysis engines containing pipelined document annotators and various adapters. The UIMA system, method and computer program may be used to generate answers to input queries. The method includes inputting a document and operating at least one text analysis engine that comprises a plurality of coupled annotators for tokenizing document data and for identifying and annotating a particular type of semantic content. Thus it can be used to analyze a question and to extract entities as possible answers to a question from a collection of documents.

In one non-limiting embodiment, the Common Analysis System (CAS) data structure form is implemented as is described in U.S. Pat. No. 7,139,752, the whole contents and disclosure of which is incorporated by reference as if fully set forth herein.

As shown in greater detail in the more detailed logical architecture diagram of FIG. 3, the “Query Analysis” module 20 receives an input that comprises the query 19 entered, for example, by a user via their web-based browser device. An input query 19 may comprise a string such as “Who was the tallest American president?”. Alternately, a question may comprise of a string and an implicit context, e.g., “Who was the shortest?”. In this example, context may range from a simple another string e.g. “American presidents” or “Who was the tallest American president?” to any data structure, e.g. all intermediate results of processing of the previous strings—a situation arising e.g., in a multiple turn dialog. The input query is received by the Query Analysis module 20 which includes, but is not limited to, one or more the following sub-processes: Parse and Predicate Argument Structure block 202; a Focus Segment, Focus and Modifiers block 204; Lexical Answer Type block 206; Question Decomposition block 208; a Lexical and Semantic Relations module 210; a Question Classifier block 212; and a Question Difficulty module 214.

The Parse and Predicate Argument Structure block 202 implements functions and programming interfaces for decomposing an input query into its grammatical and semantic components, e.g., noun phrases, verb phrases and predicate/argument structure. An (English Slot Grammar) ESG-type parser may be used to implement block 202. The Focus Segment, Focus & Modifiers block 204 is provided to compute the focus and focus modifiers of the question, and is further described below. The Lexical Answer Type (LAT) block 206 implements functions and programming interfaces to provide additional constraints on the answer type (Lexical) as will be described in greater detail herein below. The Question decomposition block 208 implements functions and programming interfaces for analyzing the input question to determine the sets of constraints specified by the question about the target answer. There are several ways that these constraints may relate to one another: 1) Nested Constraints; 2) Redundant Constraints; and 3) Triangulation. With nested constraints, an answer to an “inner” question instantiates an “outer” question. For example, “Which Florida city was named for the general who led the fight to take Florida from the Spanish?”. With redundant constraints, one constraint uniquely identifies the answer. For instance, “This tallest mammal can run at 30 miles per hour. Which is it?”. With triangulation, each constraint generates a set of answers and the correct answer is the one answer in common in the two (or more) sets. For example, in a “puzzle”-style question “What is a group of things of the same kind, or scenery constructed for a theatrical performance”.

The Lexical and Semantic Relations module 210 is provided to detect lexical and semantic relations in the query (e.g., predicate-argument relations) as is the Question Classification block 212 that may employ topic classifiers providing information addressing, e.g., what is the question about? The Question Difficulty module 214 executes methods providing a way to ascertain a question's difficulty, e.g., by applying readability matrix to the question. It is understood that one or more of the query/question analysis processing blocks shown in FIG. 3 may be selected for a particular implementation.

The Parse and Predicate Argument Structure block 202 implements functions and programming interfaces for decomposing an input query into its grammatical components by performing a Lexical processing and a syntactic and predicate argument structure analysis as known in the art. For an example query:

“In the 1960s this largest Kansas city became the world's largest producer of general aviation aircraft”.

The Parse and Predicate Arguments block 202 will produce an example parse search results tree structure below with eX providing an index into the tree, e.g., the “become” word is e8 (the 8^thstructure of the results tree, and e7 indexes the 7^thword of the results tree structure) where 7 represents the word (“city”) that is the first argument of “become” and e13 (indexes the 13^thword of the results tree structure) is the “producer” which is the second argument of “become” in the semantic structure depicted:

in(e1,e3,e8)

the(e2,e3)

1960s(e3,u)

this(e4,e7)

large(e5,e7)

Kansas(e6,e7)

city(e7,u)

become(e8,e7,e13)

the(e9,e10)

world(e10,u,e13)

aposts(e11,e10)

large(e12,e13)

producer(e13,of:e17)

general(e15,e17)

aviation(e16,u,e17)

aircraft(e17)

The Focus Segment, Focus and Modifiers block 204 detects a Focus Segment which is the text span in the question that the correct answer replaces. For example, in the following query, the italicized words represent the focus segment in the query:

“In the 1960s this largest Kansas city became the world's largest producer of general aviation aircraft.”

To detect a focus segment, a set of rules that operate on Predicate-Argument structures and the ESG parse are implemented that match Patterns in Predicate-Argument Structure (PAS). Example patterns include, e.g., a Noun Phrase; “what/which/this/these X”, where X is another object(s); “who/what/when/where/why/this/these”; a Pronoun without a referent. An example of a pronoun pattern with the pronoun words italicized is as follows:

As a boy he built a model windmill; his calculus foe Gottfried Leibniz designed them as an adult.

With reference to the Lexical Answer Type (LAT) block 206, LAT is the question terms that identify the semantic type of the correct answer. The italicized words in the following passage represent the LAT in the following query:

“What Kansas city is the world's largest producer of general aviation aircraft”.

LATs may include modifiers if they change the meaning. For example, the italicized words represent the LAT in the following query:

Joliet and Co found that the Mississippi emptied into what body of water?

Referring to FIG. 3, an output 29 of the Question/Query analysis block 20 comprises a query analysis result data structure (CAS structure). In this embodiment, the output data structure Question/Query analysis block 20 and the candidate answer generation block 30 may be implemented to pass the data among the modules, in accordance with the UIMA Open Source platform.

The “Candidate Answer Generation” module 30 receives the CAS-type query results data structure 29 output from the Question/Query analysis block 20, and generates a collection of candidate answers based on documents stored in Primary Sources 11 and in Answer Source KB 21. The “Candidate Answer Generation” module 30 includes, but is not limited to, one or more of the following functional sub-processing modules: A Term Weighting & Query Expansion module 302; a Document Titles (Document Retrieval in Title Sources) module 304; an Entities From Passage Retrieval module 308; and an Entities from Structural Sources K.B. module 310.

The Term Weighting and Query Expansion module 302 implements functions for creating a query against modules 11 and 21 (part of query generation) with an embodiment implementing query expansion (see, e.g., http://en.wikipedia.org/wiki/Query_expansion). The Document Titles (Document Retrieval in Title Sources) module 304 implements functions for detecting a candidate answer (from sources 11 and 21). The Entities From Passage Retrieval module 308 implements functions for detecting a candidate answer in textual passages, e.g. based on grammatical and semantic structures of the passages and the query. The Entities from Structured Sources module KB 310 implements functions for retrieving a candidate answer based on matches between the relations between the entities in the query and the entities in Answer Source KB 21 (implemented e.g. as an SQL query).

As a result of implementing the functional modules of the Candidate Answer Generation block 30, a query is created and run against all of the structured and unstructured primary data sources 11 in the (local or distributed) sources database or like memory storage device(s). This query may be run against the structured (KB), semi-structured (e.g., Wikipedia, IMDB databases, a collection of SEC filings in XBRL, etc.), or unstructured data (text repositories) to generate a candidate answer list 39 (also as a CAS, or an extension of prior CAS). It should be understood that, in one embodiment, the query is run against a local copy of the listed primary source databases, or may access the publically available public database sources. Moreover, it should be understood that, in one embodiment, not all the terms from the query need to be used for searching the answer—hence the need for creating the query based on results of the query analysis. For example, to answer the question “five letter previous capital of Poland,” the terms “five letter” should not be part of the query.

As further shown in FIG. 3, the Answer Source Knowledge Base 21 is shown interfacing with the Entities from Structured Sources module 310 that includes: Typed Lists (e.g., list of all countries in world), Precise Unary (e.g., a country), Binary (e.g., country+head of state of country), Ternary (e.g., country+head of state of country+wife of head of state), n-ary Relation Extracted, etc.

With reference to FIGS. 2 and 4, the “Candidate Answer Scoring” module 40B receives a CAS-type data structure 49 (i.e., CAS or CASes) output from the Supporting Passage Retrieval (SPR) block 40A of Evidence Gathering block 50, for example. The “Candidate Answer Scoring” module 40B includes, but is not limited to, one or more the following functional sub-processing modules: a Lexical & Semantic Relations in Passage module 402; a Text Alignment module 405; a Query Term Matching in Passage module 407; a Grammatical Relations block 410; an Answer Look-up in KBs module 413; and a Candidate Answer Type Analysis module 415.

The Lexical and Semantic Relations in Passage module 402 implements functions computing how well semantic (predicate/argument) relations in the candidate answer passages are satisfied (part of answer scoring). The Text Alignment module 405 implements functions for aligning the query (or portions thereof) and the answer passage and computing the score describing the degree of alignment, e.g., when aligning answers in a quotation. The Query Term Matching in Passage module 407 implements functions for relating how well a passage in the query matches to terms in the candidate answer passages (part of answer scoring). The Grammatical Relations block 410 implements functions for detecting a grammatical relations among candidate answers which can be subsumed under the Lexical & Semantic Relations in Passage module 402. The Answer Look-up in KBs module 413 implements functions for detecting the candidate answer based on the score ranking. The Candidate Answer Type Analysis module 415 produces a probability measure that a Candidate Answer is of the correct type based, e.g., on a grammatical and semantic analysis of the document with which the Candidate Answer appears. The output of the “Candidate Answer Scoring” module 40B is a CAS structure having a list of answers with their scores given by the modules.

As described herein, multiple parallel operating modules may be implemented to compute the scores of the candidate answers with the scores provided in CAS-type data structures 59 based on the above criteria. For instance, does the answer satisfy similar lexical and semantic relations (e.g. for a query about an actress starring in a movie, is the answer a female, and does the candidate satisfy actor-in-movie relation?), how well do the answer and the query align, how well do the terms match and do the terms exist in similar order. Thus, it is understood that multiple modules are used to process different candidate answers and, thus, potentially provide many scores in accordance with the number of potential scoring modules.

With reference to FIGS. 2 and 4, the “Answer Ranking” module 60 receives a plurality of CAS-type data structures 59 output from the Evidence Gathering block 50 (which includes implementing SPR 40A and Candidate Answer Scoring 40B), and generates a score for each received candidate answer. FIG. 4 shows a machine learning implementation where the “Answer Ranking” module 60 includes a trained model component 71 produced using a machine learning techniques from prior data. The prior data may encode information on features of candidate answers, the features of passages the candidate answers come in, the scores given to the candidate answers by Candidate Answer Scoring modules 40B, and whether the candidate answer was correct or not. The machine learning algorithms can be applied to the entire content of the CASes together with the information about correctness of the candidate answer. Such prior data is readily available for instance in technical services support functions, or in more general settings on the Internet, where many websites list questions with correct answers. The model encodes a prediction function which is its input to the “Learned Feature Combination” module 73.

Thus, in the embodiment illustrated in FIG. 4, there is input to the answer ranking module 60 a list of candidate answers, as a CAS, in addition to a trained model that is stored in the trained model sub-module 71 and whose parameters depend on the type of the query. The answer ranking module 60 includes a learned feature combination sub-block 73 which implements functionality that generates a ranked list of answers 75. An output of the answer ranking module 60 includes an answer to the query (one or a list) and, optionally, a clarification question (if the system is engaging in a dialog or if none of the produced answers has a high rank). The learned feature combination sub-block 73 applies the prediction function produced by the Trained Model 71, and for example it implements methods that weight the scores of candidate answers based on the trained model. An example implementation of the training block 71 and of Learned Feature Combination 73 may be found in the reference to Ittycheriah, A. et al, entitled “{IBM}'s Statistical Question Answering System—{TREC}—“Text {REtrieval} Conference” in 2001 at ttp://citeseer.ist.psu.edu/cache/papers/cs2/7/http:zSzzSztrec.nist.govzSzpubszSztrec 10zSz.zSzpaperszSztrec2001.pdf/ittycheriah01ibms.pdf).

The application of a machine learning Trained Model 71 and the Learned Feature Combination 73 are described below in more detail. In one embodiment, a two-part task is implemented to: (1) identify a best answer among candidates, and (2) determine a confidence in that best answer. In accordance with this processing, each question-candidate pair comprises an instance, and scores are obtained from a wide range of features, e.g., co-occurrence of answer and query terms, whether a candidate matches answer type, and search engine rank. Thus, for an example question, “What liquid remains after sugar crystals are removed from concentrated cane juice.” example scores such as shown in the Table 1 below are generated based on but not limited to: Type Analysis, Alignment, Search Engine Rank, etc. TypeAgreement is the score for whether the lexical form of the candidate answer in the passage corresponds to the lexical type of the entity of interest in the question. Textual Alignment scores the alignment between question and answer passage.

TABLE 1

Candidate	Type	Align	Rank	Score

Milk	1	0.2	3	0.46
Muscovado	0	0.6	1	0.48
Molasses	1	0.5	2	0.8

Thus, in this embodiment, candidate answers are represented as instances according to their answer scores. As explained above, a classification model 71 is trained over instances (based on prior data) with each candidate answer being classified as true/false for the question (using logistic regression or linear regression function or other types of prediction functions as known in the art). This model is now applied, and candidate answers are ranked according to classification score with the classification score used as a measure of answer confidence, that is, possible candidate answers are compared and evaluated by applying the prediction function to the complete feature set or subset thereof. If the classification score is higher than a threshold, this answer is deemed as an acceptable answer. Using the numbers for Type, Align and Rank of Table 1, and the prediction function (Score) given by an example linear expression:

=0.5*Type+0.8*Align+(1−Rank)*0.1, values of 0.46, 0.48 and 0.8 are obtained for Milk, Muscovado, and Molasses, respectively (the higher value being better). These values are represented in the Score column of TABLE 1. This example of the scoring function is given for illustration only, and in the actual application more complex scoring functions may be used. For instance, the mathematical expression may be based on the logistic regression function (a composition of linear expressions with the exponential function), and may be applied to a much larger number of features.

FIG. 5 is a block diagram 500 depicting system operation. At step 510, a query is received by the system programmed to perform the steps of the invention. The set of instructions are executed in a computing environment comprising one or more processors or computing devices. The query is analyzed and parsed into elements suitable for searching by the search engine 511 (performing the information retrieval function of module 30 in FIGS. 1 and 2).

FIG. 5 also represents a Data Model defining the data structures supplied as input to, or produced as output by, the system components shown in FIGS. 1 and 2. The Data Model provides documentation about what a given component does, and it enables independently developed components to be integrated with a fairly high degree of confidence that they will correctly inter-operate. The Data Model is formally defined as a UIMA Type System and has a declarative specification in a UIMA Type System descriptor. As known, the Unstructured Information Management Architecture (UIMA) framework is an open, industrial-strength, scalable and extensible platform for building analytic applications or search solutions that process text or other unstructured information to find the latent meaning, relationships and relevant facts buried within (http://incubator.apache.org/uima/).

The Data Model is instantiated with a UIMA CAS (the container for instances of types in the type system.

The type system has a few basic design points.

- 1. A CAS represents a single question however, it is not so limited, i.e., includes the question of some prior focus (category, prior question or answer, or question meta-data some element of the context is also provided);
- 2. The question is the subject of analysis in the initial CAS view;
- 3. Processing is divided into several phases, where each phase may generate multiple, new CASes with new subjects of analysis and corresponding views, but the original question view is carried in every CAS. It is understood that variations are possible.

All processing results may be added to the original CAS (with intermediate results carrying the way through to the end of processing) and the final answer generated by the system is posted as an annotation in the CAS.

In one example embodiment, the data model includes a base annotation type that many of the types extend the uima.tcas.Annotation (see http://incubator.apache.org/UIMA). Each class (e.g., an annotator) can also provide an estimate of the correctness of the annotations it creates.

In one exemplary embodiment, as part of the question analysis performed during the questions processing phase 510, the question 19 is received and the CAS is initialized with the question (and the question context). Two initial annotations are created: a Question annotation with meta-data about the question, and a QaResult annotation that consolidates all of the question analysis results by either containing the results directly, or pointing to other annotations that represent question analysis results.

The base Question annotation type is defined to optionally include any associated meta-data such as the source of the question (TREC, technical support, TV show, etc.), prior dialog, or other contextual information (for example, about information contained in the input expressed in other modalities).

The question type can be further specialized into example subtypes modeling questions that Question which class defines a Type (i.e., question type, for example, one of FACTOID, LIST, DEFINITION, OTHER, OPINION or UNCLASSIFIED question types).

An initial step in analyzing the question is to run the NLP (Natural Language Processing) stack on the question. Natural language processing typically includes syntactic processing (e.g. using the ESG parser) and derivation of predicate-argument structure. This processing is performed in accordance with the standard UIMA fashion, where the NLP stack is run as an aggregate analysis engine on the CAS. In an embodiment of the invention, all of the NLP stack results are added to the CAS as annotations extending Hutt and ESG type systems.

After NLP stack processing, the question analysis components are run, which include question classification, answer type detection, and focus identification, for example, as shown in the query analysis block of FIG. 3. The question may be classified based on question types (metadata), each of which may require special downstream processing. The result of this classification may be stored in a QClass annotation:

Example downstream processing may include processing a puzzle question (where getting the answer requires synthesis information from multiple sources, inference, etc.); an audio_visual question that requires audio/visual processing; a simple_factoid question with quotes, or named entities, etc.; a FACTOID about a fact that can be “looked up”; and a DEFINITION that contains a definition of the answer and where the words defined by the question are expected as an answer.

The typical question analysis processes shown in FIG. 3 produces several annotations, including the focus, answer type, semantic role labels, and constraints, and marks any portion of the question that represents a definition.

For the above annotation types, the Question Analysis component 510 of FIG. 5 will create an instance of the annotation, set the span over the question text (if appropriate), and set any other features in the annotation. Note that there may be multiple instances of these annotations.

The question and the results of question analysis are used to generate an abstract representation of the query, which for purposes of description, is referred to as the AbstractQuery. The abstract query represents all searchable keywords and phrases in the question, along with the semantic answer type (if it was detected).

The abstract query is represented using the following types: a synonym (all query concepts underneath are synonyms of each other); a phrase (all query concepts in order are a phrase); a tie (an “or”, i.e., a disjunction of the argument nodes); a weight (the concepts underneath are weighted per the float stored in the operator); required (the concepts underneath are all required, if possible); and relation (the concepts underneath are below a relation, which is stored within the operator).

Referring to FIG. 5, in question processing block 510, after question analysis, search processing begins, and this may include searching primary structured and unstructured sources, e.g. Google, a local copy of Wikipedia, or database look-up.

Each search engine has a query generator that generates an engine-specific query from the abstract query and formats it in the query syntax for the search engine. The search engine then processes the query and adds a search result hit-list to the CAS. A Search object contains the search engine query, an identifier for the search engine, and the search results.

In one embodiment, a search result is represented by a SearchResult object, which contains an identifier for the result (a URI), a score for the result, and the actual content of the result, i.e., the passage text, knowledge base tuple, etc. The SearchResult may be specialized for different kinds of search engines and corresponding search results.

The Document object may be created to represent the result delivered by the search engine. This object may include a title of the document and a unique identifier for this document, and other data and meta-data. The passage object may be used with a search engine that returns passages. It may add to the document object the offset (e.g., a character offset of the start of this passage within the document that contains this passage, and a character offset of the end of this passage within the document that contains this passage) and passage length metadata for the passage hit.

As represented in FIG. 5, the data in the example CAS structure are output of the search results block of the question analysis processing step 510 and are about to be processed in parallel. A Search Hit CAS splitter mechanism 515 is used to initiate a parallel search for candidate answers. For parallel operations, the search list (search result passages) are distributed by the CAS splitter element 515 so that concurrent search results processing techniques are applied (work divided) to process each of the found search results and to perform candidate answer generation (in parallel) using the techniques described herein in the Candidate Answer Generation block 30 (FIG. 3).

During candidate answer generation, candidate answers are identified in the search result. In one example implementation, a candidate answer is represented at two different levels: a Candidate Answer Variant; and A Candidate Answers Canon. A CandidateAnswerVariant is a unique candidate answer string (possibly the result of some very simple normalization). A CandidateAnswerCanon is a canonicalized candidate answer that groups together semantically equivalent variants. Both of these types extend an abstract base class CandidateAnswer which class defines the candidate answer string and features associated with this candidate answer.

A class (e.g., CandidateAnswer) for candidate answers provides the candidate answer string and features associated with this candidate answer. In operation, one or more of its subtypes may be instantiated. One sub-type includes a variant of a candidate answer class (CandidateAnswerVariant) defined that may have multiple occurrences, all of which are collected in a variant object and defines the occurrences of this variant. A CandidateAnswerOccurrence class is provided that annotates a span of text identified as a candidate answer and defines: (1) the manner in which covered text refers to some entity, e.g. NAME, PRONOUN, CITY; (2) the source of the candidate answer; (3) the character offset of the start of this candidate answer within the text of the source; and (4) the character offset of the end of this candidate answer within the text of the source.

In one example implementation, candidate answers are derived from document titles, and another method may derive a candidate answer from one or more elements in the candidate passage. Candidate answers can be normalized whereby several spelling variants can be identified in one canonical form.

The data in the example CAS structure 39 output of the search results processing and candidate answer generation block 30 is again processed in parallel by a Candidate Answer CAS splitter mechanism 520 that is used to parallelize the candidate answers for further processing.

As mentioned above, in some cases, some of the candidate answers are unlikely to be a correct answer. For these low quality candidate answers, it may not be worth the computational cost of searching for further supporting evidence at 40A and 40B. In an embodiment of this invention, a subset of the candidate answers to use for supporting passage retrieval at 40A and 40B is identified by the following procedure. Content independent scorers are run on the candidate answers (since these scorers do not require supporting passages). A scoring function (e.g., a logistic regression model) is used to score each candidate answer, and candidate answers with low scores (for example, below a specified threshold) are omitted from supporting passage retrieval. The remaining candidate answers are considered good enough to merit further investigation.

To achieve this filtering, the architecture of FIG. 5 is provided with a Hypothesis Filtering module 522, a Pre-Filter Context Independent Candidate Answer Processing module 524, and a Post-Filter Context Independent Candidate Answer Processing module 526.

Modules

524 and 526 may operate, in an embodiment of the invention, similar to context Independent Candidate Answer Processing module 43 of the above-identified U.S. patent application Ser. No. 12/152,411. Both

modules

524 and 526 provide a score for a candidate answer independently of the passage in which the candidate answer is found—that is, independent of the content of the passage in which the candidate answer is found. For instance, if the candidate answer is obtained from the document title, the score will not be dependent on the content of the passage and is context independent.

A pathway is provided in which the candidate CAS is sent from “candidate CAS Splitter” 520 to Pre-Filter Context Independent Candidate Answer Processing module 524. A pathway 534 is also provided in which the candidate CAS is sent from the Pre-Filter Context Independent Candidate Answer Processing module 524 to the Hypothesis Filtering module 522.

In Candidate Answer Processing module 524, a scoring function (e.g., a logistic regression model) is used to score each candidate answer. Hypothesis Filtering module 522 is used to filter the candidate answers based on the scores provided in Processing module 524. Candidate answers with low scores are omitted from supporting passage retrieval at 40A; i.e., these candidate answers are considered to be not good enough to be worth the computational cost of searching for supporting evidence. The threshold used to identify low-scoring answers may be obtained by optimizing over a given data set, e.g., performing a parameter sweep to select for an optimal tradeoff between computational cost and final answer quality metrics. The remaining candidate answers are considered to be good enough to merit further investigation.

A pathway 536 is provided in which a candidate answer is sent from the Hypothesis Filtering module 522 to the Supporting Passage Retrieval Process 40A. This pathway 536 is used only for candidate answers which have received a sufficiently high score from Hypothesis Filtering module 522, a pathway 540 is also provided in which a candidate answer is sent from the Hypothesis Filtering module 522 to the Post-Filter Context Independent Candidate Answer Processing module 526. This pathway 540 is also used only for candidate answers which have received a sufficiently high score from the Hypothesis Filtering module.

A pathway 542 is also provided in which candidate answers are sent from the Hypothesis Filtering Module 522 to the Final Merging/Ranking process 570. In an embodiment of the invention, this pathway 542 is used only for candidate answers which have received a low score form the Hypotheses Filtering module 522. Candidates along this path are still included in the final answer ranking, but do not benefit from the supporting evidence and analysis that occurs in the post-filter answer scoring at 40A and 40B.

The Evidence Gathering module 50 (shown in FIGS. 1 and 3) that implements (parallel processing) and supports passage retrieval and answer scoring according to embodiments of the invention is now described in greater detail with respect to FIG. 5.

Traditionally, passage retrieval is used in candidate answer generation wherein using keywords from the question, passages are found from unstructured corpora. Then candidate answers are extracted from those passages.

In accordance with an embodiment of the present invention, Supporting Passage Retrieval (SPR) 40A operates after candidate answer generation. For each resulting candidate passage, the passage content are traversed to find/look for those passages having candidate answers in addition to question terms. It is understood that better passages can be found if it is known what candidate answer is being looked for. For each candidate answer passed to SPR 40A, the sources are traversed to find those passages having candidate answers in addition to question terms (i.e., another search is conducted against the original primary sources (databases) or the Candidate KB). In another embodiment, the search can be conducted against cached search results (past passages). It is understood that the best results are obtained if the search is repeated with candidate answers included together with the question terms.

Supporting Passages are then scored by multiple Scorers by conducting one or more of the following: Simple Term Match Score; textual Alignment; and a deeper analysis. Simple Term Match Score implements executable instructions for counting the number of terms that match, and Textual Alignment implements executable instructions for determining if words appear in the same or similar order with a similar distance between them so they are not separated much (aligned). This is advantageous to find a quotation. To find quotes, for example, an alignment would be performed to get the best result. Deeper analysis implements executable instructions for determining the meaning of the passages/question (i.e., lexical and/or semantic relations). Each of these analyses produces a score.

An example Evidence Gathering is now described for non-limiting, illustrative purposes. In this example, a query is given as follows:

‘In 2002, who became the first Republican sitting senator ever to host

aturday Night Live?’

In the Query Generation implementing stopword removal (that is removal from the query of the most frequent words such as “a”, “an”, “the”, “is/was/be . . . ”, “become/became . . . ” . . . ), the query becomes: ‘Republican first sitting senator ever host

aturday Night Live?2002’.

After query generation, in one embodiment, the query is sent to an Internet search engine, e.g., such as provided by MSN, and the top 20 result documents are read. The following depicts example passage extraction results (candidate answers) for the example question search results for ‘Republican first sitting senator ever host

aturday Night Live?2002’. In each document, passages are identified that include each candidate answer (i.e., John McCain or Al Gore), along with as many question keywords as possible shown italicized. Both example passages include a passage score that, in one exemplary embodiment, is calculated as:
Passage Score=# of query terms in passage/total # of query terms
Candidate: John McCain
Document: http://doney.net/aroundaz/celebrity/mccain_john.htm Passage: Representative from Arizona 1st District (1983-1987), POW (1967-1972), Navy pilot, first sitting Senator to host Saturday Night Live (2002) Born in the Panama Canal Zone, John McCain shares the headstrong, blunt, maverick traits of his father and grandfather, who were the first father and son four star Admirals in the U.S.
Passage Score: 8/11=0.73
Candidate: Al Gore
Document: http://www.imdb.com/title/tt0072562/news Passage: 17 Dec. 2002
(StudioBriefing) No longer a candidate for the presidency in 2004, Al Gore may have a whole new career cut out for him as the host of a late-night comedy show, judging by the ratings for the December 14 edition of NBC's Saturday Night Live.
Passage Score: 5/11=0.45

Additionally calculated, in one embodiment, is an SPR Answer Score that is calculated, in one embodiment, as a decaying sum of scores of passages containing that answer as shown in equation 1)
AnswerScore=P ₀ +nP ₁ +n ² P ₂ +n ³ P ₃+ . . . 1)
where, P_iis the i^thhighest passage score, and “n” is a constant<1 (e.g., 0.1).

Thus, for the example query ‘Republican first sitting senator ever host “Saturday Night Live” 2002,” the SPR “AnswerScore” for Candidate John McCain is calculated as:

1) Taking the first candidate answer passage for John McCain with question keywords shown italicized, to wit:

Representative from Arizona 1st District (1983-1987), POW (1967-1972), Navy pilot, first sitting Senator to host Saturday Night Live (2002). Born in the Panama Canal Zone, John McCain shares the headstrong, blunt, maverick traits of his father and grandfather, who were the first father and son four star Admirals in the U.S. there is calculated a passage score of [Score: 0.74].

2) Taking an example second candidate answer passage for John McCain, to wit:

John McCain, Meghan regaled reporters with tales of her days as an intern at Saturday Night Live in 2004. & quot; Slave work, & quot; she says, & quot; but I had an amazing time. There is calculated a passage score of [Score: 0.27].

3) Taking an example third candidate answer passage for John McCain, to wit:

The most prominent Republican Arizona Senator John McCain was portrayed as a loser because of his support for staying the course in Iraq. There is calculated a passage score of [Score: 0.18].

Thus, a Combined AnswerScore for candidate answer John McCain, in accordance with equation 1), with n=0.1, becomes: 0.74+(0.1)(0.27)+(0.01)(0.18)=0.7688.

Similarly, for the example query “2002 Republican first sitting senator ever host Saturday Night Live,” the SPR “AnswerScore” for Candidate Al Gore is calculated as:

1) Taking the first candidate answer passage for Al Gore with question keywords shown italicized, to wit:

17 Dec. 2002 (StudioBriefing) No longer a candidate for the presidency in 2004, Al Gore may have a whole new career cut out for him as the host of a late-night comedy show, judging by the ratings for the December 14 edition of NBC's Saturday Night Live28, there is calculated a passage score of [Score: 0.45].

2) Taking the second candidate answer passage for Al Gore, to wit:

Also in attendance were former Presidents George Bush, Gerald Ford, Jimmy Carter, former Vice-president Al Gore former Senator Bob Dole and all their wives. Was portrayed on “Saturday Night Live” (1975) by Phil Hartman, Chris Farley (once), David Spade (once), Chris Elliot (once), Michael McKean, and Darrell Hammond. There is calculated a passage score of [Score: 0.36].

3) Taking the third candidate answer passage for Al Gore, to wit:

Also in attendance were former Presidents George Bush, Gerald Ford, Jimmy Carter, former Vice President Al Gore, former Senator Bob Dole and all their wives. [September 2001]. Was portrayed on " Saturday Night Live" (1975) by Phil Hartman, Chris Farley (once), David Spade (once), Chris Elliott (once), Michael McKean, and Darrell Hammond. There is calculated a passage score of [Score: 0.36].

4) Taking the fourth candidate answer passage for Al Gore, to wit:

Remember Al Gore's “Saturday Night Live” skit where he pretended to be President and the world was a glorious place? There is calculated a passage score of [Score: 0.27].

Thus, a Combined AnswerScore for candidate answer Al Gore, as calculated by the SPR module in accordance with equation 1), becomes: 0.45+(0.1)(0.36)+(0.01)(0.36)=0.4896.

It is noted that an Answer scoring in accordance with a simple sum scheme would have been 1.44 for Al Gore, which would have beat a score of 1.19 in a simple sum calculation for John McCain.

The answer scores for each candidate answer would be included in the CAS.

Referring back to FIG. 5, in module 546, supporting passages are retrieved. Functionality is initiated after the CAS split and receiving filtered candidate answers from Hypothesis Filter 522. Supporting passage records created by Supporting Passage Retrieval are split by Supporting Passage Splitter 548; and since there may be many of them, the splitter routes the new CASes (with all information that was computed previously: context, query, candidate answer, supporting passage) to Answer Scoring 40B.

The results contain many CASes containing (among other elements) the three important items: a candidate answer, the question terms, and a supporting passage. Since thousands of such CASes can be generated per one question, these candidates are scored in parallel. In an embodiment, the candidate scoring performed by candidate scoring module 40B can be subdivided into two classes: context independent scoring 43 (where the answer can be scored independently of the passage), and context dependent scoring 47 (where the answer score depends on the passage content). For example, as mentioned above, if the candidate answer is obtained from the document title, the score will not be dependent on the content of the passage, and is context independent. In contrast, other types of candidate answer scoring based on text alignment (module 405, FIG. 4), grammatical relations (module 410, FIG. 4), or lexical and semantic relations (module 402, FIG. 4) require a comparison between the query and the passage, and are context dependent. Since most of these methods depend on the computation of grammatical and semantic relations in the passage, search results must be processed (in the Search Result Processing module in SPR block 40A) prior to Context Dependent Candidate Answer processing in 47.

The results of an answer scorer are saved in the CAS. During the final phase of Answer Ranking processing 60, all of the candidate answer features are aggregated and merged, and the final candidate answer scoring function is applied (as described above with respect to the example scores provided in Table 1. Since a given candidate answer may appear in multiple passages, the Final Merge/Rank annotator must collect results across CASes, normalize and merge candidate answers, merge feature scores produced by the same answer scorer across multiple instances of the candidate answer, and aggregate the results. The normalized, merged, and aggregated results are input to the scoring function to produce a final score for the candidate answer. The final scoring results are saved as an answer and//or delivered to a user. In embodiments of the invention, Final merging and ranking is incremental, i.e., the machine provides the best answer so far as the computation on different nodes completes. Once all nodes complete, the final (top) answer(s) is delivered. Thus, in one embodiment, the final AnswerList and Answers are added to the original Question view, and the question answering process is complete.

It should also be noted that if the system is unable to find an answer or to find an answer with a high score (based, e.g., upon comparison to a preset threshold), the system might ask the user a clarifying question, or deliver a collection of answers, or admit a failure and ask the user for further direction. A person skilled in the art would be able to implement such a dialog based e.g. on U.S. Pat. Nos. 6,829,603 and 6,983,252, both of which are incorporated by reference as if fully set forth herein, and a reference entitled “Natural language dialogue for personalized interaction” authored by Wlodek Zadrozny, et al. and found in Communications of the ACM archive, Volume 43, Issue 8, (August 2000), Pages: 116-120, (http://portal.acm.org/citation.cfm?id=345164).

A person skilled in the art would be able to implement a further extension to the system of the invention to employ modes of multimodal communication (using U.S. Pat. No. 7,136,909) involving multiple modalities of text, audio, video, gesture, tactile input and output etc. As mentioned above, examples of such interaction include a cell phone user who is asking a question using voice and is receiving an answer in a combination of other modalities (voice, text and image), or an interaction with a video game.

The data model and processing models described herein are designed to enable parallel processing, and to admit a “streaming” model of computation, where results become available incrementally, before all processing is complete. This streaming model may be advantageous if the analytics are able to identify and process the most likely candidates first, and continue to improve scoring estimates with more processing time.

As mentioned, in one embodiment, the above-described modules of FIGS. 1-5 can be represented as functional components in UIMA and may be embodied as a combination of hardware and software for developing applications that integrate search and analytics over a combination of structured and unstructured information. The software program that employs UIMA components to implement end-user capability is generally referred to as the application, the application program, or the software application.

The UIMA high-level architecture, one embodiment of which is illustrated in FIGS. 1-7, defines the roles, interfaces and communications of large-grained components that cooperate to implement UIM applications. These include components capable of analyzing unstructured source artifacts, such as documents containing textual data and/or image data, integrating and accessing structured sources and storing, indexing and searching for artifacts based on discovered semantic content.

Although not shown, a non-limiting embodiment of the UIMA high-level architecture includes a Semantic Search Engine, a Document Store, at least one Text Analysis Engine (TAE), at least one Structured Knowledge Source Adapter, a Collection Processing Manager, at least one Collection Analysis Engine, all interfacing with application logic. In one example embodiment, the UIMA operates to access both structured information and unstructured information to generate candidate answers and an answer in the manner as discussed herein. The unstructured information may be considered to be a collection of documents, and can be in the form of text, graphics, static and dynamic images, audio and various combinations thereof.

Aspects of the UIMA are further shown in FIG. 8, where there is illustrated an Analysis Engine (AE) 600 that can be a component part of the Text Analysis Engine. Included in the AE 600 is a Common Analysis System (CAS) 610, an annotator 620 and a controller 630. A second embodiment of a TAE (not shown) includes an aggregate Analysis Engine composed of two or more component analysis engines as well as the CAS, and implements the same external interface as the AE 600.

Common Analysis System

610

The Common Analysis System (CAS) 610 is provided as the common facility that all Annotators 620 use for accessing and modifying analysis structures. Thus, the CAS 610 enables coordination between annotators 620 and facilitates annotator 620 reuse within different applications and different types of architectures (e.g. loosely vs. tightly coupled). The CAS 610 can be considered to constrain operation of the various annotators.

The CAS 610 principally provides for data modeling, data creation and data retrieval functions. Data modeling preferably defines a tree hierarchy of types, as shown in the example Table 2 provided below. The types have attributes or properties referred to as features. In embodiments of the invention, there are a small number of built-in (predefined) types, such as integer (int), floating point (float) and string; UIMA also includes the predefined data type “Annotation”. The data model is defined in the annotator descriptor, and shared with other annotators. In the Table 2, some

ypes?that are considered extended from prior art unstructured information management applications to accommodate question answering in embodiments of the invention include:

TABLE 2

TYPE (or feature)	TYPE's PARENT (or feature type)

Query Record	Top
Query	Query Record
Query Context	Query Record
Candidate Answer Record	Annotation
Candidate Answer	Candidate Answer Record
Feature: CandidateAnswerScore	Float
Supporting Passage Record	Candidate Answer Record
Feature: SupportingPassageScore	Float

In Table 2, for example, all of the question answering types (list in the left column) are new types and extend either another new type or an existing type (shown in the right column). For example, both Query and Query Context are kinds of Query Record, a new type; while Candidate Answer Record extends the UIMA type Annotation, but adds a new feature CandidateAnswerScore which is a Float.

CAS

610 data structures may be referred to as “feature structures.” To create a feature structure, the type must be specified (see TABLE 2). Annotations (and—feature structures) are stored in indexes.

The CAS 610 may be considered to be a collection of methods (implemented as a class, for example, in Java or C++) that implements an expressive object-based data structure as an abstract data type. Preferably, the CAS 610 design is largely based on a TAE 130 Feature-Property Structure that provides user-defined objects, properties and values for flexibility, a static type hierarchy for efficiency, and methods to access the stored data through the use of one or more iterators.

The abstract data model implemented through the CAS 610 provides the UIMA 100 with, among other features: platform independence (i.e., the type system is defined declaratively, independently of a programming language); performance advantages (e.g., when coupling annotators 620 written in different programming languages through a common data model); flow composition by input/output specifications for annotators 620 (that includes declarative specifications that allow type checking and error detection, as well as support for annotators (TAE) as services models); and support for third generation searching procedures through semantic indexing, search and retrieval (i.e. semantic types are declarative, not key-word based).

The CAS 610 provides the annotator 620 with a facility for efficiently building and searching an analysis structure. The analysis structure is a data structure that is mainly composed of meta-data descriptive of sub-sequences of the text of the original document. An exemplary type of meta-data in an analysis structure is the annotation. An annotation is an object, with its own properties, that is used to annotate a sequence of text. There are an arbitrary number of types of annotations. For example, annotations may label sequences of text in terms of their role in the document's structure (e.g., word, sentence, paragraph etc), or to describe them in terms of their grammatical role (e.g., noun, noun phrase, verb, adjective etc.). There is essentially no limit on the number of, or application of, annotations. Other examples include annotating segments of text to identify them as proper names, locations, military targets, times, events, equipment, conditions, temporal conditions, relations, biological relations, family relations or other items of significance or interest.

Typically an Annotator's 620 function is to analyze text, as well as an existing analysis structure, to discover new instances of the set of annotations that it is designed to recognize, and then to add these annotations to the analysis structure for input to further processing by other annotators 220.

In addition to the annotations, the CAS 610 of FIG. 6 may store the original document text, as well as related documents that may be produced by the annotators 620 (e.g., translations and/or summaries of the original document). Preferably, the CAS 610 includes extensions that facilitate the export of different aspects of the analysis structure (for example, a set of annotations) in an established format, such as XML.

More particularly, the CAS 610 is that portion of the TAE that defines and stores annotations of text. The CAS API is used both by the application and the annotators 620 to create and access annotations. The CAS API includes, for example, three distinct interfaces. A Type system controls creation of new types and provides information about the relationship between types (inheritance) and types and features. One non-limiting example of type definitions is provided in TABLE 1. A Structure Access Interface handles the creation of new structures and the accessing and setting of values. A Structure Query Interface deals with the retrieval of existing structures.

The Type system provides a classification of entities known to the system, similar to a class hierarchy in object-oriented programming. Types correspond to classes, and features correspond to member variables. Preferably, the Type system interface provides the following functionality: add a new type by providing a name for the new type and specifying the place in the hierarchy where it should be attached; add a new feature by providing a name for the new feature and giving the type that the feature should be attached to, as well as the value type; and query existing types and features, and the relations among them, such as “which type(s) inherit from this type”.

In an embodiment, the Type system provides a small number of built-in types. As was mentioned above, the basic types are int, float and string. In a Java implementation, these correspond to the Java int, float and string types, respectively. Arrays of annotations and basic data types are also supported. The built-in types have special API support in the Structure Access Interface.

The Structure Access Interface permits the creation of new structures, as well as accessing and setting the values of existing structures. In an embodiment, this provides for creating a new structure of a given type, getting and setting the value of a feature on a given structure, and accessing methods for built-in types. Feature definitions are provided for domains, each feature having a range.

In an alternative environment, modules of FIGS. 1-5 can be represented as functional components in GATE (General Architecture for Text Engineering) (see: http://gate.ac.uk/releases/gate-2.0alpha2-build484/doc/userguide.html). GATE employs components which are reusable software chunks with well-defined interfaces that are conceptually separate from GATE itself. All component sets are user-extensible and together are called CREOLE—a Collection of REusable Objects for Language Engineering. The GATE framework is a backplane into which plug CREOLE components. The user gives the system a list of URLs to search when it starts up, and components at those locations are loaded by the system. In one embodiment, only their configuration data is loaded to begin with; the actual classes are loaded when the user requests the instantiation of a resource). GATE components are one of three types of specialized Java Beans: 1) Resource; 2) Processing Resource; 3) Language Resource; and 4) Visual Resource. Resource is a top-level interface, which describes all components. What all components share in common is that they can be loaded at runtime, and that the set of components is extendable by clients. They have Features, which are represented externally to the system as “meta-data” in a format such as RDF, plain XML, or Java properties. Resources may all be Java beans in one embodiment. ProcessingResource is a resource that is runnable, may be invoked remotely (via RMI), and lives in class files. In order to load a PR (Processing Resource), the system knows where to find the class or jar files (which will also include the metadata). Language Resource is a resource that consists of data, accessed via a Java abstraction layer. They live in relational databases. VisualResource is a visual Java bean, component of GUIs, including of the main GATE gui. Like PRs these components live in .class or .jar files.

In describing the GATE processing model, any resource whose primary characteristics are algorithmic, such as parsers, generators and so on, is modelled as a Processing Resource. A PR is a Resource that implements the Java Runnable interface. The GATE Visualisation Model implements resources whose task is to display and edit other resources are modelled as Visual Resources. The Corpus Model in GATE is a Java Set whose members are documents. Both Corpora and Documents are types of Language Resources (LR) with all LRs having a Feature Map (a Java Map) associated with them that stores attribute/value information about the resource. FeatureMaps are also used to associate arbitrary information with ranges of documents (e.g. pieces of text) via an annotation model. Documents have a DocumentContent which is a text at present (future versions may add support for audiovisual content) and one or more AnnotationSets which are Java Sets.

As UIMA, GATE can be used as a basis for implementing natural language dialog systems and multimodal dialog systems having the disclosed question answering system as one of the main submodules. The references, incorporated herein by reference above (U.S. Pat. Nos. 6,829,603 and 6,983,252, and 7,136,909) enable one skilled in the art to build such an implementation.

Embodiments of the invention can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment containing both hardware and software elements. In embodiments, the invention is implemented in software, which includes but is not limited, to firmware, resident software, microcode, etc.

The invention can take the form of a computer program product accessible from a computer-usable or computer-readable medium providing program code for use by or in connection with a computer or any instruction execution system. For the purposes of this description, a computer usable or computer readable medium can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus or device.

The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) or a propagation medium. Examples of a computer-readable medium include a semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read only memory (ROM), a rigid magnetic disk and an optical disk. Current examples of optical disks include compact disk read only memory (CD-ROM), compact disk read/write (CD-R/W), and DVD.

The system and method of the present disclosure may be implemented and run on a general-purpose computer or computer system. The computer system may be any type of known or will be known systems and may typically include a processor, memory device, a storage device, input/output devices, internal buses, and/or a communications interface for communicating with other computer systems in conjunction with communication hardware and software, etc.

The terms “computer system” and “computer network” as may be used in the present application may include a variety of combinations of fixed and/or portable computer hardware, software, peripherals, and storage devices. The computer system may include a plurality of individual components that are networked or otherwise linked to perform collaboratively, or may include one or more stand-alone components. The hardware and software components of the computer system of the present application may include and may be included within fixed and portable devices such as desktop, laptop, and server. A module may be a component of a device, software, program, or system that implements some “functionality”, which can be embodied as software, hardware, firmware, electronic circuitry, or etc.

In the preferred embodiment the term “user” refers to a person or persons interacting with the system, and the term “user query” refers to a query posed by the user. However other embodiments can be constructed, where the term “user” refers to the computer system generating a query by mechanical means, and where the term “user query” refers to such a mechanically generated query. In this context the “user query” can be a natural language expression, a formal language expression, or a combination of natural language and formal language expressions. The need for automated answering of a computer generated questions arises, for example, in the context of diagnosing failures of mechanical and electronic equipment, where the failing equipment can generate a query on the best way to fix a problem, and such a query could be answered by the system described in this invention based on a relevant corpus of textual data collected from the Internet. Methods of generating automatically natural language expressions from a formal representation have been previously disclosed, for example, in the U.S. Pat. Nos. 5,237,502 and 6,947,885, the contents and disclosures of each of which are incorporated by reference as if fully set forth herein and, can be used by the skilled in the art to create systems for automatically issuing a “user query”. Similarly, in such a diagnostic scenario the system can ask an elaboration question, e.g. to query for some additional parameters.

The embodiments described above are illustrative examples and it should not be construed that the present invention is limited to these particular embodiments. Thus, various changes and modifications may be effected by one skilled in the art without departing from the spirit or scope of the invention as defined in the appended claims.

Claims

What is claimed is:

1. A system for generating answers to questions, comprising:

a computer device comprising at least one distinct software module, each distinct software module being embodied on a tangible computer-readable medium; a memory; and at least one processor coupled to the memory and operative for:

receiving an input query;

obtaining a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;

passing the candidate answers in sequence through a pre-filter module and a hypothesis filtering module, including

the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers independent of the data content in which said each candidate answer is found, and

the hypothesis filtering module applying a defined score threshold to the pre-filter scores of the candidate answers for sending a first group of the candidate answers having pre-filter scores below the defined threshold to a post-filter module and sending, in a parallel processing path, a second group of the candidate answers having pre-filter scores above the defined threshold through a supporting passage retrieval process run by said at least one processor;

the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers independent of the data content in which said each candidate answer of the first group of candidate answers is found;

the supporting passage retrieval process retrieving supporting passages for the second group of the candidate answers;

providing content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and

returning one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers, wherein the applied defined score threshold is established by said at least one processor to obtain an optimal tradeoff between a computational cost and a quality metric of a final answer returned to the input query.

2. The system according to claim 1, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

3. The system according to claim 2, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

4. The system according to claim 3, wherein the first group of the candidate answers have a pre-filter score above a specified high score.

5. The system according to claim 4, wherein the third group of the candidate answers have a pre-filter score below a specified low score.

6. The system according to claim 1, wherein to establish said defined score threshold, said at least one processor performing a test on a given set of data to select the optimal tradeoff between a computational cost and quality metrics of a final answer to the input query.

7. An article of manufacture, comprising:

at least one non-transitory computer readable hardware medium having computer readable program code logic to generate answers to questions, the computer readable program code logic, when executing, performing the following:

receiving an input query at a computer processor system;

obtaining, at the computer processor system, a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;

passing, at the computer processor system, the candidate answers in sequence through a pre-filter module and a hypothesis filtering module, including

the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module to direct a first group of the candidate answers to a post-filter module and to direct, in a parallel processing path, a second group of the candidate answers through a supporting passage retrieval process;

providing, at the computer processor system, content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and

returning, at the computer processor system, one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

8. The article of manufacture according to claim 7, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

9. The article of manufacture according to claim 8, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

10. The article of manufacture according to claim 9, wherein the first group of the candidate answers have a pre-filter score above a specified high score.

11. The article of manufacture according to claim 10, wherein the third group of the candidate answers have a pre-filter score below a specified low score.

12. The article of manufacture according to claim 7, wherein to establish said defined score threshold, the computer readable program code logic, when executing, causes the computer processor system to perform a test on a given set of data to select the optimal tradeoff between a computational cost and quality metrics of a final answer to the input query.

13. A computer-implemented method of generating answers to questions, comprising:

receiving an input query at least one processor of a computer processor system;

obtaining, at the at least one processor of the computer processor system, a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;

passing, at the at least one processor of the computer processor system, the candidate answers in sequence through a pre-filter module and a hypothesis filtering module, including

the hypothesis filtering module applying a defined score threshold to the pre-filter scores of the candidate answers and sending a first group of the candidate answers having pre-filter scores below the defined threshold to a post-filter module and sending, in a parallel processing path, a second group of the candidate answers having pre-filter scores above the defined threshold through a supporting passage retrieval process run by the at least one processor of said computer system;

providing, at the at least one processor of the computer processor system, content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and

returning, at the computer processor system, one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers, wherein the applied defined score threshold is established by said at least one processor to obtain an optimal tradeoff between a computational cost and a quality metric of a final answer returned to the input query.

14. The method according to claim 13, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

15. The method according to claim 14, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

16. The method according to claim 15, wherein:

the first group of the candidate answers have a pre-filter score above a specified high score; and

the third group of the candidate answers have a pre-filter score below a specified low score.

17. The method according to claim 13, wherein to establish said defined score threshold, the method further comprises: performing, at said computer processor system, a test on a given set of data to select the optimal tradeoff between a computational cost and quality metrics of a final answer to the input query.