Microsoft Technology Licensing, LLC

Nuance Communications, Inc.

Palo Alto Research Center Inc.

Other

Xerox Corporation

Fuji Xerox Co., Ltd.

Richard Stamford Crouch

Giovanni Lorenzo Thione

Martin Henk Van Den Berg

Peter R Stubley

Marisa Ferrara Boston

Ali Erdem Ozcan

Livia Polanyi

Barney Pell

David D Ahn

Daniel Gureasko Bobrow

Ronald M Kaplan

Chad P Walters

Scott A Prevost

Olga Iosifovna Gurevich

Francis Pieraut

Sylvain Goulet

John Ogden Lamping

Franco Salvetti

Tracy Holloway King

Christopher D Culy

Michael Tepper

Mary Dalrymple

Hadar Shemtov

Emmanuel Rayner

Hannah Copperman

Techniques for question answering involve receiving, from a user, a text input expressing a question in natural language. In response to the question, a text output expressing an answer to the question may be generated. A plurality of documents comprising natural language text may be analyzed, involving mapping the question to one or more hypotheses, analyzing at least one passage of text in at least one of the documents to determine whether the passage entails at least one of the hypotheses, and in response to determining that the passage entails at least one of the hypotheses, identifying the passage as providing supporting evidence for the answer to the question. The answer and the at least one passage identified as providing supporting evidence for the answer may be presented to the user in response to the text input.

Question answering with entailment analysis

Methods and systems for indexing document passages are presented. In some embodiments, a computing device may identify a plurality of documents that comprise a plurality of passages. A passage index comprising a plurality of entries may be generated. Each entry may comprise keywords from a passage of the plurality of passages in one of the plurality of documents. Each entry may further comprise at least one annotation associated with the passage. A search query comprising at least one search keyword may be received. The passage index for each document of the plurality of documents may be analyzed using the at least one search keyword to identify at least one passage from the plurality of documents that matches the search query. In response to the query, the at least one passage may be presented.

Text indexing and passage retrieval

Techniques for question answering involve receiving, from a user, a text input expressing a question in natural language. The text input may be analyzed, including identifying in the question at least one first portion answerable from at least one structured data source, and at least one second portion answerable from at least one unstructured data source. At least one first query configured for the structured data source(s) may be constructed from the at least one first portion of the question and applied to the structured data source(s) to retrieve first answer information for the at least one first portion of the question. At least one second query configured for the unstructured data source(s) may be constructed from the at least one second portion of the question and applied to the unstructured data source(s) to retrieve second answer information for the at least one second portion of the question.

Question answering from structured and unstructured data sources

Techniques for product recommendation involve receiving, from a user, a request for recommendation of a product matching one or more characteristics specified in the request. In response to the request, a product may be identified for recommendation, at least in part by searching an ontology in which the identified product is represented as matching at least one of the characteristics specified in the request. The ontology may further include at least one link to at least one natural language product review providing supporting evidence that the identified product matches the at least one characteristic. At least a portion of the at least one natural language product review may be retrieved using the at least one link in the ontology, and presented to the user in support of a recommendation of the identified product, in response to the request.

Product recommendation with ontology-linked product review

Techniques for product recommendation involve receiving, from a user, a text input expressing a question asking for a product recommendation. In response to the question, an answer that identifies a product for recommendation to the user may be generated. A plurality of product reviews comprising natural language text evaluations of the product may be analyzed, involving identifying at least one passage of text in at least one of the product reviews as providing supporting evidence for the product in answer to the question. The answer and the at least one passage in the at least one product review identified as providing supporting evidence for the answer may be presented to the user in response to the text input.

Product recommendation with product review analysis

Methods and computer-readable media for associating words or groups of words distilled from content, such as reported speech or an attitude report, of a document to form semantic relationships collectively used to generate a semantic representation of the content are provided. Semantic representations may include elements identified or parsed from a text portion of the content, the elements of which may be associated with other elements that share a semantic relationship, such as an agent, location, or topic relationship. Relationships may also be developed by associating one element that is in relation to, or is about, another element, thereby allowing for rapid and effective comparison of associations found in a semantic representation with associations derived from queries. The semantic relationships may be determined based on semantic information, such as potential meanings and grammatical functions of each element within the text portion of the content.

Identification of semantic relationships within reported speech

Technologies are described herein for coreference resolution in an ambiguity-sensitive natural language processing system. Techniques for integrating reference resolution functionality into a natural language processing system can processes documents to be indexed within an information search and retrieval system. Ambiguity awareness features, as well as ambiguity resolution functionality, can operate in coordination with coreference resolution. Annotation of coreference entities, as well as ambiguous interpretations, can be supported by in-line markup within text content or by external entity maps. Information expressed within documents can be formally organized in terms of facts, or relationships between entities in the text. Expansion can support applying multiple aliases, or ambiguities, to an entity being indexed so that all of the possibly references or interpretations for that entity are captured into the index. Alternative stored descriptions can support retrieval of a fact by either the original description or a coreferential description.

Coreference resolution in an ambiguity-sensitive natural language processing system

Methods, systems and computer readable media for finding documents in a data store that match a natural language query submitted by a user are provided. The documents and queries are matched by determining that words within the query have the same relationship to each other as the same words in the document. Documents are semantically analyzed and words in the document are indexed along with the role the word plays in a sentence. The initial semantic role may be generalized using a role hierarchy and stored in the index along with the original role. A similar analysis may be used with the search query to find words used in the same role in both the query and the document.

Indexing role hierarchies for words in a search index

Technologies are described herein for generating a semantic translation rule to support natural language search. In one method, a first expression and a second expression are received. A first representation is generated based on the first expression, and a second representation is generated based on the second expression. Aligned pairs of a first term in the first representation and a second term in the second representation are determined. For each aligned pair, the first term and the second term are replaced with a variable associated with the aligned pair. Word facts that occur in both the first representation and the second representation are removed from the first representation and the second representation. The remaining word facts in the first representation are replaced with a broader representation of the word facts. The translation rule including the first representation, an operator, and the second semantic representation is generated.

Semi-automatic example-based induction of semantic translation rules to support natural language search

Tools and techniques are described that relate to iterators for applying term occurrence-level constraints in natural language searching. These tools may receive a natural language input query, and define term occurrence-level constraints applicable to the input query. The methods may also identify facts requested in the input query, and may instantiate an iterator to traverse a fact index to identify candidate facts responsive to the input query. This iterator may traverse through at least a portion of the fact index. The methods may receive candidate facts from this iterator, with these candidate facts including terms, referred to as term-level occurrences. The methods may apply the term occurrence-level constraints to the term-level occurrences. The methods may select the candidate fact for inclusion in search results for the input query, based at least in part on applying the term occurrence-level constraint.

Iterators for applying term occurrence-level constraints in natural language searching

Techniques are presented to determine user-interest sensitive notes. User selected passages, user interest information, condensation transformations and optional meaning distortion constraints are identified. User foci expressed by the selected passages are determined based on the similarity of the elements in the selected passages to elements in the user interest information. User sensitive notes are determined by selectively applying the condensation transformations to the selected passages to preferentially retain user foci while eliding less salient information. Meaning distortions constraints are optionally applied in conjunction with condensation transformations or in creating the condensation transformations in order to reduce the likelihood of distorting the meaning of the passage.

Systems and methods for user-interest sensitive note-taking

Techniques are presented to determine user-interest sensitive condensations of a passage. One or more passages are selected and user interest information, condensation transformations and optional meaning distortion constraints are identified. The foci of user interest within the selected passages are determined based the similarity of the elements in the selected passages to elements in the user interest information. The condensation transformations are applied to the selected passages to preferentially retain user foci while eliding less salient information. The resultant condensate provides signals the user-interest sensitive meaning of the passage. Meaning distortions constraints are optionally applied in conjunction with the condensation transformations or in creating the condensation transformations to reduce the likelihood of distorting the meaning of the passage.

Systems and methods for user-interest sensitive condensation

Techniques are provided for segmenting text into categorized discourse constituents and attaching discourse constituents into a structural representation of discourse. Techniques for determining hybrid structural and non-structural summaries of a text are also provided. A text is segmented based on a theory of discourse analysis into at least a main discourse constituent containing spatio-temporal information about a single event in a possible world view. The discourse constituents are then inserted into a structural representation of discourse. Non-structural techniques are used to determine relevance scores and important discourse constituents are determined. Relevance scores are percolated through the structural representation of discourse to determine supporting preceding discourse constituents that preserve grammaticality. A hybrid text summary is then determined based on the structural representation of the discourse and relevance scores.

Systems and methods for hybrid text summarization

Techniques are provided for detecting entailment and contradiction. Packed knowledge representations for a premise and conclusion text are determined comprising facts about the relationships between concept and/or context denoting terms. Concept and context alignments are performed based on alignments scores. A union is determined. Terms are marked as to their origin and conclusion text terms replaced with by corresponding terms from the premise text. Subsumption and specificity, instantiability, spatio-temporal and relationship based packed rewrite rules are applied in conjunction with the context denoting facts to remove entailed terms and to mark contradictory facts within the union. Entailment is indicated by a lack of any facts from the packed knowledge representation of the conclusion in the union. Entailment and contradiction markers are then displayed.

Growing community of inventors

Cupertino, CA, United States of America

Richard Stamford Crouch