International Business Machines Corporation

Hyundai Motor Company

Kia Corporation

William G Dubyak

Charles Evan Beller

Palani Sakthi

Kristen Maria Summers

Edward Graham Katz

Brian L Keith

Nicole O'Connor

Joshua G Hong

Alexander Charles Tonetti

Sean Thomas Thatcher

Julie T Yu

Vijai Gandikota

Sean L Bethard

Robert A Sheets

Aaron K Baughman

Mohamed Nooman Ahmed

Linda M Boyer

Christopher F Ackermann

Christopher Frank Codella

Richard L Darden

Andeep S Toor

Arnold Greenland

A candidate document is received, for example, by a document filter. A determination is made based on the content of the candidate document, whether the candidate document is relevant to a document corpus. A determination is made based on the content of the candidate document, whether the candidate document is novel with respect to the document corpus. In response to determining that the candidate document is relevant to the document corpus and novel with respect to the document corpus, the candidate document is added to the document corpus to make at least a portion of the content of the candidate document available for a response to a search query.

Automated determination of document utility for a document corpus

Natural language processing is enhanced by linguistically extracting intelligence about a user. A history of user queries is analyzed by a natural language classifier to determine various user intents, and these intents are combined to form a user intent profile. The profile includes elements of sentiment, emotion and tone. The profile can be used in various ways including restricting access to documents in a collection, or refining a cognitive analysis of a query. For access restriction, a determination is made that the user intent is inconsistent with a document, and the user is denied access to the document. This determination involves a user intent score which is compared to a score of the document. For cognitive analysis, searching of reference documents is filtered by excluding documents based on the user intent. The searching includes a comparison of meta-data tags of the documents to the user intent.

Leveraging natural language processing to refine access control within collections

Embodiments are directed to interaction with an open-domain question and answer system by recognizing questions that are highly broad or abstract, and generating and processing a batch of questions expressing alternate, concrete instances of the more general, abstract question. Responses to all of the questions in the batch are considered as candidates, and the strongest general answers are returned. A weighted, ranked answer set, based on weighting individual concrete questions and scaling the answers proportional to the weight of the questions, is provided to the user. The approach, according to embodiments herein, addresses the need for responses to broad questions in which a response to any of a set of more concrete question instances may serve to answer the question.

Dispersed batch interaction with a question answering system

Batch interaction with a computerized question answering system can produce an answer that more closely relates to a user's information need. A batch of questions can be generated interactively, and provides a context for a first question received from a user. The batch of questions includes or more additional questions which have terms with a nonsynonymous semantic relation to a first term in the first question. A question answering system can process the batch of questions to determine candidate answers. An answer to the first question can be determined based, at least in part, on a combined ranking of the candidate answers.

Precision batch interaction with a question answering system

Dynamic semantic processing of text in a word processing application with engagement of question-answering system. A user provides a text stream to a computer system via an input source. The input text stream includes a first natural language statement. The system determines that the first natural language statement includes a fact-based component expressed in natural language form. The system identifies an initial span and an alternative span of the first natural language component, based on the determining. The system engages a question-answering (QA) system by providing the initial span and the alternative span of the first natural language component to the QA system, and by receiving, in response to the providing, a set of natural language results from the QA system. The system evaluates the initial span and the alternative span of the first natural language component based on the received natural language results.

Document preparation with argumentation support from a deep question answering system

Embodiments can provide a computer implemented method for harvesting training data for a training set for use by a system capable of answering questions, the system comprising a processor and a memory comprising instructions executed by the processor, the method comprising receiving, from a user, an input question; processing the input question and returning, to the user, a result set comprising one or more ranked hypotheses and one or more ranked evidence passages corresponding to the one or more ranked hypotheses; receiving, from the user, an indication that one of the one or more ranked hypotheses is to be designated a watched hypothesis; adding the input question and the watched hypothesis to a to-be-vetted question/answer (QA) pair set comprising one or more to-be-vetted QA pairs; vetting each of the one or more to-be-vetted QA pairs in the to-be-vetted QA pair set through a first-pass automatic vetting procedure; if a vetted QA pair passes the first-pass automatic vetting procedure, adding the vetted QA pair to the training set; and retraining one or more ranking models based upon the training set.

Harvesting question/answer training data from watched hypotheses in a deep QA system

Embodiments can provide a computer implemented method, in a data processing system comprising a processor and a memory comprising instructions which are executed by the processor to cause the processor to implement an enhanced corpus management system, the method comprising: identifying one or more functional domain categories; ingesting one or more incoming documents to form an open-domain corpus; for each functional domain category, identifying one or more representative documents to establish a seed sub-corpus; calculating a degree of fit score between each of the one or more incoming documents and the one or more established functional domain category seed sub-corpora; and assigning one or more of the incoming documents to one or more of the functional domain categories based upon the degree of fit score to create an enhanced corpus.

Corpus management by automatic categorization into functional domains to support faceted querying

Embodiments can provide a computer implemented method, in a data processing system comprising a processor and a memory comprising instructions which are executed by the processor to cause the processor to implement an improved search query generation system, the method comprising inputting a natural language question; parsing the natural language question into a parse tree; identifying argument positions comprising one or more argument position terms; for each argument position: comparing a head term's discriminator score against a threshold discriminator score; and if the head term surpasses the threshold discriminator score, adding the head term as a required term to an improved search query; and outputting the improved search query.

System and method for generating improved search queries from natural language questions

A method of retrieving information includes obtaining access to a plurality of documents comprising document terms, receiving a lexicon including a plurality of lexical terms, receiving a query, identifying a plurality of search terms in the query by decomposing the query, determining at least one match of the search terms to the lexical terms, determining a plurality of matches of the search terms to the document terms, and scoring each of the documents based on the at least one match to the lexical terms and the matches to the document terms.

Weighting of unobserved terms in a language model for information retrieval using lexical resources

An embodiment of the invention may include a method, computer program product and system for generating follow-up questions based on machine learning utilizing a computing device. The embodiment may include receiving an input question from a user. The embodiment may include parsing the received input question to extract input question components. Parsing utilizes natural language processing techniques. The embodiment may include executing trained question component models to predict follow-up question components. The extracted input question components are utilized as inputs to the trained question component models. The embodiment may include combining the predicted follow-up question components to generate one or more follow-up questions. The embodiment may include returning the one or more follow-up questions to the user.

Suggesting follow up questions from user behavior

Formulating a response to a natural language query based on user intent

A method for recommending responses to emergent conditions is provided. The present invention may include receiving a query from a user. The present invention may also include retrieving a plurality of recommended responses for the received query from a plurality of entities and a plurality of relations stored in a graph-based knowledge resource. The present invention may further include presenting the retrieved plurality of recommended responses to the user.

Recommending responses to emergent conditions

A computer-implemented method, a computer program product, and a computer processing system are provided. The method includes identifying, by a processor using a topic identification system, topic information for a source topic-domain in a corpus. The method further includes extracting, by the processor, an entity from the source topic-domain. The method also includes tagging, by the processor, the entity with the topic information to obtain a tagged entity that includes a tag specifying the topic information. The method additionally includes storing the tagged entity in a memory device. The method further includes performing, by the processor, downstream processing of the tagged entity in a natural language processing pipeline using the tag of the tagged entity as an additional data point.

Tagging named entities with source document topic information for deep question answering

A method, computer system, and a computer program product for recommending responses to emergent conditions is provided. The present invention may include receiving a query from a user. The present invention may also include retrieving a plurality of recommended responses for the received query from a plurality of entities and a plurality of relations stored in a graph-based knowledge resource. The present invention may further include presenting the retrieved plurality of recommended responses to the user.

Embodiments are directed to an entity extraction and filtering method that enables a close search of documents to build filters necessary for near real-time monitoring of streaming sources of information. According to an embodiment, the entity extraction and filtering method operates based on the following parameters. First, a detection of an entity of interest warrants flagging an arriving article for analyst attention. Nothing more than a match may be required. The list of entities may be derived by an entity extractor from a corpus of data. Secondly, automatic updates may be utilized, so that exports are automatically updated to the filters. Thirdly, information flowing past the filters may update a static corpus whether or not they are flagged for an analyst or user. This allows for new relationships to be detected and extracted, and the filters subsequently updated.

Growing community of inventors

Severna Park, MD, United States of America

William G Dubyak