International Business Machines Corporation

Siddharth A Patwardhan

Branimir Konstantinov Boguraev

Aditya Anand Kalyanpur

James W Murdock, Iv

Jennifer Chu-Carroll

David Angelo Ferrucci

Rosanna Mannan

Hernan Badenes

Murthy V Devarakonda

Vadim Sheinin

Petros Zerfos

David W Buchanan

Elahe Khorasani

Preethi Raghavan

David J McClosky

Youngja Park

Anthony Tom Levas

David Deidou Taieb

Adam P Lally

Kenneth James Barker

Mihaela Ancuta Bornea

Sugato Bagchi

David C Gondek

Sara Rosenthal

Adam R Faulkner

Suneil H Berajawala

Yanpeng Li

Generating textual entailment pair by a natural language processing (NLP) system. The NLP system receives two input texts, such as a question and a candidate answer. The NLP system queries a database and retrieves passages likely to include text that support the candidate answer. The NLP system generates parse trees and performs term matching on the passages and scores them according to the matching. The NLP system detects anchor pairs in the question and in the passage and aligns subgraphs (within the parse trees) of one to the other based on matching. The NLP system identifies aligned terms in the question and the passage that are not in the aligned subgraphs. The NLP system identifies text fragments, for the question and the passage, within the non-aligned segments of their respective parse trees, that connect the aligned term to the aligned portion of the subgraph.

Entailment knowledge base in natural language processing systems

In configuring a processing system with an application made up of machine learning components, where the application has been trained on a set of training data, the application is executed on the processing system using another set of training data. Outputs of the application produced from the other set of training data identified that concur with ground truth data are identified. The components are adapted to produce outputs of the application that concur with the ground truth data using the identified outputs of the application.

Adaptable processing components

A computer-implemented method includes detecting a first set and a second set of citations to a legal case in a plurality of legal documents and a first legal document distinct from the plurality of legal documents, respectively. The computer-implemented method further includes determining tones corresponding to each citation in the first and second sets of citations. The computer-implemented method further includes determining a score for each tone in the first and second sets of tones. The computer-implemented method further includes aggregating a first and subset and a second of the first and second sets of citations, respectively. The computer-implemented method further includes generating an average score for the first and second subsets. The computer-implemented method further includes determining a degree of similarity between the first and second subsets based, in part, on a comparison of average scores. A corresponding computer program product and computer system are also disclosed.

Tone analysis of legal documents

Provided herein are techniques for processing a context-situated inquiry to provide results satisfying the inquiry. An inquiry and its supporting context are processed using natural language processing to determine an interpretation for the inquiry and context, and the interpretation is presented to receive corrective information for the interpretation. The corrective information is applied to the interpretation and content is retrieved based on the interpretation of inquiry and context to produce candidate results for the inquiry. Supplemental information is iteratively received to generate updated candidate results.

Processing context-based inquiries for knowledge retrieval

Mechanisms are provided to implement a self-training engine of a question and answer system. The self-training engine pairs an unanswered natural language question with portions of an electronic document to generate an unlabeled data set. The self-training engine trains a model based on a labeled data set comprising a finite number of question and answer pair data structures and applies the model to the unlabeled data set to identify an answer from the portions of the electronic document to the unanswered natural language question. The self-training engine modifies the labeled data set to include the identified answer and corresponding unanswered natural language question as an additional question and answer pair data structure. The self-training engine then trains the model based on the modified labeled data set.

Self-training of question answering system using question profiles

A method of generating secondary questions in a question-answer system. Missing information is identified from a corpus of data using a computerized device. The missing information comprises any information that improves confidence scores for candidate answers to a question. The computerized device automatically generates a plurality of hypotheses concerning the missing information. The computerized device automatically generates at least one secondary question based on each of the plurality of hypotheses. The hypotheses are ranked based on relative utility to determine an order in which the computerized device outputs the at least one secondary question to external sources to obtain responses.

Generating secondary questions in an introspective question answering system

A computer-implemented method includes detecting a first and second set of citations to a plurality of legal cases cited in a plurality of legal documents and a first legal document distinct from the plurality of legal documents, respectively. The computer-implemented method further includes determining a first and second set of tones corresponding to each citation in the first and second sets of citations, respectively. The computer-implemented method further includes determining a score for each tone in the first and second sets of tones, respectively. The computer-implemented method further includes detecting a first set of annotations corresponding to the first set of citations. The computer-implemented method further includes building a training model, based, at least in part, from at least a subset of the first set of annotations. The computer-implemented method further includes determining a second set of annotations corresponding to the second set of citations.

Annotation of legal documents with case citations

A mechanism is provided in a computing device configured with instructions executing on a processor of the computing device to implement a timeline generation system, for automated timeline completion. The timeline generation system executing on the processor of the computing device identifies a plurality of events in documents in a corpus of information. The timeline generation system places the plurality of events in a partial timeline data structure. The timeline generation system selects an event progression from an event progression knowledge base. The timeline generation system aligns the selected event progression to the partial timeline data structure. The timeline generation system identifies a set of events missing from the partial timeline data structure. The timeline generation system maps the set of events missing from the partial timeline data structure to the partial timeline based on the selected event progression to form a completed timeline data structure.

Automated timeline completion using event progression knowledge base

An embodiment of the invention provides a method for including receiving a natural language query from a user with an interface, and generating multiple dependency parses of the natural language query with a parser device connected to the interface. The generating of the multiple dependency parses includes dividing the natural language query into multiple components, and creating a single dependency parse by connecting each component of the components with at least one other component of the components. A processor connected to the parser device applies rules to all of the multiple dependency parses to identify entities and relations in the natural language query.

Robust and readily domain-adaptable natural language interface to databases

An embodiment of the invention provides a method wherein a natural language query is received from a user with an interface. An ontological representation of data in a database is received with an input port, including names of concepts and names of concept properties. Template rules are received with the input port, the templates rules being language dependent and ontology independent, the template rules including widely used constructs of a language. Rules are automatically generated with a rule generation engine with the ontological representation of the data in the database and the template rules to identify entities and relations in the natural language query. Entities and relations are identified with a processor, the entities and relations being identified in the natural language query with the rules. The structured data language query is generated with a query generation engine from the entities and relations.

Natural language interface to databases

Collaborative annotation of a document. A GUI is generated by a host computer that displays a document to annotate. The GUI is transmitted by the host computer to a plurality of client computers. Real time communication modes are established by the host computer between the client computers, each including communication channels between all the client computers. The host computer receives annotations entered into the GUIs at the client computers, each annotation identifying a span of words in the document to annotate and including an associated annotation type. The host computer updates the GUI to include for display the received annotations and inter-rater agreement statistics indicating the level of agreement between annotations received from each client computer and annotations received from the other client computers. The updated GUI is then transmitted by the host computer to the plurality of client computers.

Collaborative creation of annotation training data

A system and a computer program product for verifying a statement are provided. The system is configured to receive a statement. The system is configured to decompose the received statement into one or more sets of question and answer pairs. The system is configured to determine a confidence value of each answer in the one or more question and answer pair sets. The system is configured to combine the determined confidence values. The combined confidence values represent a probability that the received statement is evaluated as true.

Growing community of inventors

Yorktown Heights, NY, United States of America

Siddharth A Patwardhan