International Business Machines Corporation

Hiroshi Kanayama

Tetsuya Nasukawa

Hideo Watanabe

Akira Koseki

Issei Yoshida

Toshiro Takase

Risa Kawanaka

Ryosuke Kohita

Daisuke Takuma

Shoko Suzuki

Radu Florian

Miki M Enoki

Yuta Tsuboi

Laura Chiticariu

Hiroshi Nomiyama

Robert T Ward

Koichi Takeda

Takuma Murakami

Shiho Negishi

Akihiro Kosugi

Kentarou Fukuda

Takuya Ohko

Efsun Kayi

Yousef El-Kurdi

An approach for generating synthetic treebanks to be used in training a parser in a production system is provided. A processor receives a request to generate one or more synthetic treebanks from a production system, wherein the request indicates a language for the one or more synthetic treebanks. A processor retrieves at least one corpus of text in which the requested language is present. A processor provides the at least one corpus to a transformer enhanced parser neural network model. A processor generates at least one synthetic treebank associated with a string of text from the at least one corpus of text in which the requested language is present. A processor sends the at least one synthetic treebank to the production system, wherein the production system trains a parser utilized by the production system with the at least one synthetic treebank.

Treebank synthesis for training production parsers

A method is provided for dictionary expansion. The method acquires an object from a user and adds the object to a set of objects previously acquired from the user that form an expandable dictionary. The method calculates a centroid based on the set. The method calculates a similarity score of each of a plurality of objects relative to the centroid for each of a plurality of object features to calculate a weighted sum of similarity scores for each of the plurality of objects. The method presents candidate objects selected among the plurality of objects based on the weighted sum. The method acquires, from the user, a preferred candidate object among the candidate objects. The method updates weights of the plurality of features to maximize the weighed sum of similarity scores for the preferred candidate object. The method expands the dictionary by adding the preferred candidate object to the expandable dictionary.

User-centric optimization for interactive dictionary expansion

A computer-implemented method for learning a policy for selection of an associative topic, which can be used in a dialog system, is described. The method includes obtaining a policy base that indicates a topic transition from a source topic to a destination topic and a short-term reward for the topic transition, by analyzing data from a corpus. The short-term reward may be defined as probability of associating a positive response. The method also includes calculating an expected long-term reward for the topic transition using the short-term reward for the topic transition with taking into account a discounted reward for a subsequent topic transition. The method further includes generating a policy using the policy base and the expected long-term reward for the topic transition. The policy indicates selection of the destination topic for the source topic as an associative topic for a current topic.

Learning of policy for selection of associative topic in dialog system

A computer-implemented method for extracting semantic relations is disclosed. In the method, a plurality of hierarchal structures that originates from a corpus of documents is obtained. Each hierarchal structure includes a plurality of elements having respective recitations included in a corresponding document. In the method, for each predetermined relationship between ancestor and descendant elements in the hierarchal structures, a first keyword list is extracted from the ancestor element and a second keyword list is extracted from the descendant element. A statistical index is calculated for each pair of first and second keywords using the first keyword lists and the second keyword lists. The index indicates a strength of association between the first and second keywords. In the method, a candidate list of keyword pairs having semantic relationships is output using the statistical index calculated for each pair.

Extraction of semantic relation

A computer-implemented method is provided. The method includes acquiring a seed word; calculating a similarity score of each of a plurality of words relative to the seed word for each of a plurality of models to calculate a weighted sum of similarity scores for each of the plurality of words; outputting a plurality of candidate words among the plurality of words; acquiring annotations indicating at least one of preferred words and non-preferred words among the plurality of the candidate words; updating weights of the plurality of models in a manner to cause weighted sums of similarity scores for the preferred words to be relatively larger than the weighted sums of the similarity scores for the non-preferred words, based on the annotations; and grouping the plurality of candidate words output based on the weighted sum of similarity scores calculated with updated weights of the plurality of models.

Word grouping using a plurality of models

A computer-implemented method, computer program product, and system are provided for providing a recommendation. The method includes receiving, by a conversation system, a query from a user. The method also includes inputting the query into an intent-entity model to obtain a first output. The method further includes inputting the query into a question-answering system to obtain a second output. The method additionally includes analyzing the first output from the intent-entity model to calculate a value of degree of ambiguity of the first output to the query. The method also includes analyzing the query and the second output from the question-answering system to calculate a value of extent of satisfaction to which the second output satisfies the query. The method further includes determining one or more recommendations using the degree of ambiguity and the extent of satisfaction. The method additionally includes outputting the one or more recommendations to the user.

Recommendation technique using automatic conversation

A method of question answering from multilingual information sources is disclosed. The present invention discloses a method, a computer system and a program product for selecting an information source language of an information source, the method includes: receiving a question; analyzing the question to obtain a category information of a word included in the question; obtaining a word included in the category information as estimated topic or region related to the question; determining a candidate for an information source language using the estimated topic or region; and selecting the information source language and corresponding information sources for retrieving documents to generate an answer of the question.

Question answering system using multilingual information sources

A method, a computer system, and a computer program product are provided for calculating a degree of similarity of users. Each user may be a preparer and an answerer of one or more quizzes. The method includes collecting one or more quizzes from one or more users as preparers and providing the collected quizzes to one or more user as answerers. The method further includes collecting one or more answering activities of one or more answerers to the provided quizzes from the one or more answerers, wherein a pair includes a preparer and an answerer. The method additionally includes calculating, for each pair, a degree of similarity of knowledge background, based on the one or more answering activities of the one or more answerers.

Calculation of a degree of similarity of users

A method of question answering from multilingual information sources is disclosed. The present invention discloses a method, a computer system and a program product for selecting an information source language of an information source, the method comprising; receiving a question; analyzing the question to obtain a category information of at least one word included in the question; obtaining a word included in the category information as estimated topic or region related to the question; determining a candidate for an information source language using the estimated topic or region; and selecting the information source language and corresponding information sources for retrieving documents to generate an answer of the question.

A computer-implemented method, program product, and system, for extracting pieces of text from a plurality of pieces of text. The method includes: primarily evaluating a measure of positive expressions and a measure of negative expressions included in each of pieces of text; secondarily evaluating each of the pieces of text on the basis of a plurality of evaluation functions, where certain evaluation functions among the plurality of evaluation functions include, as variables, the measure of positive expressions and the measure of negative expressions; and extracting a piece of text having an evaluation result with a higher rating in preference to a piece of text having an evaluation result with a lower rating, where the individual evaluation results are based on the same evaluation function among the plurality of evaluation functions.

Method for classifying pieces of text on basis of evaluation polarity, computer program product, and computer

A method, device, and computer program product for detecting negative opinions in social media, computer program product, and computer. Negative opinions in social media can be precisely detected at an early stage. A method for processing, with a computer, a plurality of messages sent by a plurality of users over time includes the following steps: obtaining a plurality of messages, each including a specific proper noun; determining a politeness level of each of the plurality of messages, each including the specific proper noun; and calculating a proportion of messages having a politeness level lower than a certain threshold with respect to the plurality of messages, each including the specific proper noun.

Method for detecting negative opinions in social media, computer program product and computer

Provided are two types of translation engines. One of the translation engines translates a text written in a source language X into an intermediate language C, and the other translates the intermediate language C into a target language Y. The X-C translation engine translates an original text written in the source language X into a text in a predetermined intermediate language C as well as acquires and adds annotations obtained by this translation to the translated text written in the intermediate language C. The C-Y translation engine references annotations, which are obtained by the X-C translation engine and added to the text in the intermediate language C, and translates the text in the intermediate language into a text in the target language Y sentence.

Machine translation device, method of processing data, and program

An apparatus which manages confidentiality of information. This apparatus includes: a recording unit operable to record information in association with a history of users having accessed the information, or, with access rights defining users able to access the information; a generating unit operable to generate management information indicating whether the information should be managed confidentially from users not permitted to access the information; a selecting unit operable to select, based on the history or access rights, users able to access the information; and a notifying unit operable to notify the selected users of the generated management information in association with identification information of the information.

Apparatus for managing confidentiality of information, and method thereof

A system and method for detecting preference expressions indicating evaluators' likes and dislikes of a product from evaluations of the product and stores text describing evaluation of the product in association with an attribute of the text. The method extracts an evaluating expression describing evaluation of the specific object from each of the texts, determines whether the extracted evaluating expression has positive or negative polarity, where the positive indicates favorable evaluation and the negative indicates unfavorable evaluation. The system inputs a text attribute that is designated as an object for detecting the preference expression and detects an evaluating expression, which is detected from a text having an input attribute from the extracted evaluating expressions as one of the preference expressions and outputs the preference expressions in association with a frequency of the preference expressions being determined to have the positive or negative polarity in the text having the attribute.

Growing community of inventors

Tokyo, Japan

Hiroshi Kanayama