Abstract: We have invented a process and method for creating a general-purpose adaptive or static machine-learning classifier using prediction by partial matching (PPM) language modeling. This classifier can incorporate homogeneous or heterogeneous feature types; variable-size contexts; sequential or non-sequential features. Features are ordered (linearized) by information saliency; and truncation of least-informative context is used for backoff to handle previously unseen events. Labels may be endogenous (from within the group) or exogenous (outside the group) of the feature types. Classification may generate labels and their probabilities; or only labels. Classification stores may be complete or minimized where redundant states are removed producing significant space savings and performance improvements. Classifiers may be static (unchanging) or online (adaptive or updatable incrementally or in batch). PPM classifiers may be incorporated in ensembles of other PPM classifiers or different machine learning algorithms.

Abstract: We describe here a system and method for creating, maintaining and using a semantic search engine environment for precise retrieval of curated answers to questions where the answers may be drawn from an authoritative document collection. The invention combines processing by human developers and software: semantic editing tools for creating, storing, maintaining queries and variants of queries, and query and document passage categories; links from queries to text passages that provide answers to these queries; a document retrieval store; means for matching user queries against stored queries; means for creating, storing, maintaining, and retrieving semantic and management metadata and categories about queries and documents and using these data for navigating the document collection; and means for finding information related to the user's information need by text and semantic similarity retrieval.

Abstract: The present invention relates to an apparatus system and method for creating a customizable and application-specific semantic similarity utility that uses a single similarity measuring algorithm with data from broad-coverage structured lexical knowledge bases (dictionaries and thesauri) and corpora (document collections). More specifically the invention includes the use of data from custom or application-specific structured lexical knowledge bases and corpora and semantic mappings from variant expressions to their canonical forms. The invention uses a combination of technologies to simplify the development of a generic semantic similarity utility; and minimize the effort and complexity of customizing the generic utility for a domain- or topic-dependent application. The invention makes customization modular and data-driven, allowing developers to create implementations at varying degrees of customization (e.g., generic, domain-level, company-level, application-level) and also as changes occur over time (e.g.

Abstract: We describe here a system and method for creating, maintaining and using a semantic search engine environment for precise retrieval of curated answers to questions where the answers may be drawn from an authoritative document collection. The invention combines processing by human developers and software: semantic editing tools for creating, storing, maintaining queries and variants of queries, and query and document passage categories; links from queries to text passages that provide answers to these queries; a document retrieval store; means for matching user queries against stored queries; means for creating, storing, maintaining, and retrieving semantic and management metadata and categories about queries and documents and using these data for navigating the document collection; and means for finding information related to the user's information need by text and semantic similarity retrieval.

Abstract: The present invention relates to an apparatus system and method for creating a customizable and application-specific semantic similarity utility that uses a single similarity measuring algorithm with data from broad-coverage structured lexical knowledge bases (dictionaries and thesauri) and corpora (document collections). More specifically the invention includes the use of data from custom or application-specific structured lexical knowledge bases and corpora and semantic mappings from variant expressions to their canonical forms. The invention uses a combination of technologies to simplify the development of a generic semantic similarity utility; and minimize the effort and complexity of customizing the generic utility for a domain- or topic-dependent application. The invention makes customization modular and data-driven, allowing developers to create implementations at varying degrees of customization (e.g., generic, domain-level, company-level, application-level) and also as changes occur over time (e.g.

Abstract: We describe here a system and method for creating, maintaining and using a semantic search engine environment for precise retrieval of curated answers to questions where the answers may be drawn from an authoritative document collection. The invention combines processing by human developers and software: semantic editing tools for creating, storing, maintaining queries and variants of queries, and query and document passage categories; links from queries to text passages that provide answers to these queries; a document retrieval store; means for matching user queries against stored queries; means for creating, storing, maintaining, and retrieving semantic and management metadata and categories about queries and documents and using these data for navigating the document collection; and means for finding information related to the user's information need by text and semantic similarity retrieval.

Abstract: We have invented a process and method for creating a general-purpose adaptive or static machine-learning classifier using prediction by partial matching (PPM) language modeling. This classifier can incorporate homogeneous or heterogeneous feature types; variable-size contexts; sequential or non-sequential features. Features are ordered (linearized) by information saliency; and truncation of least-informative context is used for backoff to handle previously unseen events. Labels may be endogenous (from within the group) or exogenous (outside the group) of the feature types. Classification may generate labels and their probabilities; or only labels. Classification stores may be complete or minimized where redundant states are removed producing significant space savings and performance improvements. Classifiers may be static (unchanging) or online (adaptive or updatable incrementally or in batch). PPM classifiers may be incorporated in ensembles of other PPM classifiers or different machine learning algorithms.

Abstract: We describe here a system and method for creating, maintaining and using a semantic search engine environment for precise retrieval of curated answers to questions where the answers may be drawn from an authoritative document collection. The invention combines processing by human developers and software: semantic editing tools for creating, storing, maintaining queries and variants of queries, and query and document passage categories; links from queries to text passages that provide answers to these queries; a document retrieval store; means for matching user queries against stored queries; means for creating, storing, maintaining, and retrieving semantic and management metadata and categories about queries and documents and using these data for navigating the document collection; and means for finding information related to the user's information need by text and semantic similarity retrieval.

Abstract: We have discovered a system and method for improving the quality of information extraction applications consisting of an ensemble of per-user, adaptive, on-line machine-learning classifiers that adapt to document content and judgments of users by continuously incorporating feedback from information extraction results and corrections that users apply to these results. The satellite classifier ensemble uses only the immediately available features for classifier improvement and it is independent of the complex cascade of earlier decisions leading to the final information extraction result. The machine-learning classifiers may also provide explanations or justifications for classification decisions in the form of rules, other machine-learning classifiers may provide feedback in the form of supporting instances or patterns.

Abstract: A semantic discovery and exploration system is disclosed where an environment enabling a developer or user to uncover, navigate, and organize semantic patterns and structures in a document collection with or without the aid of structured knowledge. The semantic discovery and exploration system provides techniques for searching document collections, categorizing documents, inducing lists of related concepts, and identifying clusters of related terms and documents. This system operates both without and with infusions of structured knowledge such as gazetteers, thesauruses, taxonomies and ontologies. System performance improves when structured knowledge is incorporated. The semantic discovery and exploration system may be used as a first step in developing an information extraction system such as to categorize or cluster documents in a particular domain or to develop gazetteers and as a part of a deployed run-time information extraction system.

Abstract: We have discovered a system and method for improving the quality of information extraction applications consisting of an ensemble of per-user, adaptive, on-line machine-learning classifiers that adapt to document content and judgments of users by continuously incorporating feedback from information extraction results and corrections that users apply to these results. The satellite classifier ensemble uses only the immediately available features for classifier improvement and it is independent of the complex cascade of earlier decisions leading to the final information extraction result. The machine-learning classifiers may also provide explanations or justifications for classification decisions in the form of rules, other machine-learning classifiers may provide feedback in the form of supporting instances or patterns.

Abstract: A semantic discovery and exploration system is disclosed where an environment enabling a developer or user to uncover, navigate, and organize semantic patterns and structures in a document collection with or without the aid of structured knowledge. The semantic discovery and exploration system provides techniques for searching document collections, categorizing documents, inducing lists of related concepts, and identifying clusters of related terms and documents. This system operates both without and with infusions of structured knowledge such as gazetteers, thesauruses, taxonomies and ontologies. System performance improves when structured knowledge is incorporated. The semantic discovery and exploration system may be used as a first step in developing an information extraction system such as to categorize or cluster documents in a particular domain or to develop gazetteers and as a part of a deployed run-time information extraction system.