International Business Machines Corporation

Alexander Charles Tonetti

Edward Graham Katz

Sean Thomas Thatcher

Charles Evan Beller

John A Riendeau

Allen Ginsberg

William G Dubyak

Julie T Yu

Sean L Bethard

Andrew Doyle

A system, computer program product, and method are provided for use with an intelligent computer platform to identify intent and convert the intent to one or more physical actions. The aspect of converting intent includes receiving content, identifying potential variants, and statistically analyzing the variants with a confidence assessment. The variants are sorted based on a protocol associated with the confidence assessment. A variant from the sort is selected and applied to a physical device, which performs a physical action and an associated hardware transformation based on the variant.

Identifying intent in dialog data through variant assessment

A controller accesses an initial taxonomy for a domain comprising one or more existing terms for the domain identified in a hierarchical structure. The controller analyzes a corpus documents for a domain to identify a selection of one or more documents with glossaries. The controller extracts, from the glossaries, one or more pairs each comprising a term and a definition. The controller attempts to map a respective term of each of the one or more pairs into the initial taxonomy for the domain based on text of a respective definition of each of the one or more pairs to generate an updated taxonomy for the domain.

Automatically extending a domain taxonomy to the level of granularity present in glossaries in documents

Method and apparatus improves the quality of responses from an automatic dialogue system by dynamically adjusting response thresholds. More particularly, the automatic dialogue system may dynamically determine response threshold values in response to user feedback. The response threshold values may be used to evaluate a confidence value. The confidence value may be assigned to or otherwise associated with an input class, or user intent. The system may automatically adjust the response threshold values to provide a better user experience as the amount of user-interaction with the system increases.

Dynamic adjustment of response thresholds in a dialogue system

A system and a computer program product are provided for evaluating question-answer pairs in an answer key by generating a predicted answer to a test question based on the answer key modification history for comparison matching against a generated answer that is generated in response to the test question, and then comparing the predicted answer and generated answer to determine an accuracy score match indication therebetween so as to present an indication that the answer key may have a problem if there is a match between the predicted answer and generated answer.

Exploiting answer key modification history for training a question and answering system

An approach is provided in which an information handling system analyzes a message tone vector corresponding to a message against a target tone vector corresponding to a recipient of the message. The message tone vector includes a set of message tone attributes and the target tone vector includes a set of target tone thresholds. The information handling system, in response to determining a difference between the message tone vector and the target tone vector, creates a variant message by substituting words in the message with one or more similar words based on the target tone vector. In turn, the information handling system sends the variant message to the recipient.

Cognitive system for declarative tone modification

A device includes a processor configured to, in response to determining that an input phrase includes a first term that is included in a term hierarchy, generate a second phrase by replacing the first term in the input phrase with a second term included in the term hierarchy. The processor is configured to determine that interactive response (IR) training data indicates that the input phrase is associated with a user intent indicator. The processor is configured to determine that user interaction data indicates that a first proportion of user phrases received by an IR system correspond to the user intent indicator. The processor is configured to update speech-to-text training data based on the input phrase and the second phrase so that a second proportion of training phrases of the speech-to-text training data correspond to the user intent indicator. The second proportion is based on the first proportion. A speech-to-text model is trained based on the speech-to-text training data.

Speech-to-text training data based on interactive response data

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

According to one embodiment, a method, computer system, and computer program product for retraining a classifier-based automatic dialogue system with recorded user interactions is provided. The present invention may include receiving recorded interactions, where the interactions are between a user and an automatic dialogue system; determining, based on the recorded interactions, whether to pair a given input with one or more classes; pairing inputs with one or more classes; assessing the reliability of the paired inputs and classes; determining whether the reliable paired inputs and classes can be consistently mapped; and merging all consistently mapped reliable pairs with an initial training set.

Automatic identification of retraining data in a classifier-based dialogue system

The temporal stability of an answer from a deep question answering system is predicted using a natural language classifier. A training corpus is divided into time-ordered slices having uniform granularity. A series of candidate answers to a training question is generated based on the slices, and a temporal profile for the series is identified by associating candidate answers with respective temporal intervals. The temporal profile is translated to a temporal stability value (representing a time period) using a temporal stability model. The classifier is trained using such training questions correlated with respective temporal stability values. Thereafter, when a user submits a natural language query to the deep question answering system, the query is also applied to the classifier which determines its temporal stability. The temporal stability is presented to the user with the answer to give a sense of how long the answer can be deemed reliable.

Predicting the temporal stability of answers in a deep question answering system

According to one embodiment, a method, computer system, and computer program product for retraining a classifier-based automatic dialog system with recorded user interactions is provided. The present invention may include receiving recorded interactions, where the interactions are between a user and an automatic dialog system; determining, based on the recorded interactions, whether to pair a given input with one or more classes; pairing inputs with one or more classes; assessing the reliability of the paired inputs and classes; determining whether the reliable paired inputs and classes can be consistently mapped; and merging all consistently mapped reliable pairs with an initial training set.

A mechanism is provided in a data processing system having a processor and a memory storing a store of semantic types and instructions for implementing a natural language processing engine for generating semantically equivalent variants of a natural language term. The mechanism receives an input term for variant analysis. The natural language processing engine executing on the data processing system identifies a semantic type of the input term based on a store of semantic types. The natural language processing engine performs a type-specific series of rule-based expansions of the input term based on the identified semantic type of the input term to form a set of semantically equivalent variants of the input term. The natural language processing engine performs a natural language processing operation using the input term and the set of semantically equivalent variants of the input term.

Type-specific rule-based generation of semantic variants of natural language expression

A mechanism is provided in a data processing system having a processor and a memory storing instructions for implementing a natural language processing engine, a store of semantic types, and a store of units, conversions among units, and variants of unit names, for generating semantically equivalent variants of a natural language term. The mechanism receives an input term for variant analysis. The natural language processing engine executing on the data processing system identifies a semantic type of the input term based on the store of semantic types. The natural language processing engine extracts a quantity and a unit from the input term based on the store of units, conversions among units, and variants of unit names. The natural language processing engine populates type-specific templates at a level of specificity based on the input term based on the identified semantic type of the input term and the extracted quantity and unit to form a set of semantically equivalent variants of the input term. The natural language processing engine performs a natural language processing operation using the input term and the set of semantically equivalent variants of the input term.

Generating semantic variants of natural language expressions using type-specific templates

A mechanism is provided in a data processing system having a processor and a memory storing a store of semantic types and instructions for implementing a natural language processing engine for generating a question/answer pair list with semantically equivalent variants. The mechanism generates a user interface for generating a question/answer pair list. The mechanism receives user input in the user interface specifying a question and an answer term and specifying an answer type from a list of previously created answer types. The input term comprises the answer term. The natural language processing engine executing on the data processing system identifies a semantic type of the answer term based on the store of semantic types. The natural language processing engine performs a type-specific series of rule-based expansions of the answer term based on the identified semantic type of the answer term of the answer term. The natural language processing engine adds at least one semantically equivalent variant from the set of semantically equivalent variants of the answer term in association with the specified question to the question/answer pair list to form an expanded question/answer pair list. The natural language processing engine trains a question answering machine learning model for a question answering cognitive system using the expanded question/answer pair list as ground truth.

Growing community of inventors

Washington, DC, United States of America

Alexander Charles Tonetti