US 11,893,353 B2Granted

Vector generation device, sentence pair learning device, vector generation method, sentence pair learning method, and program

AssigneeNippon Telegraph and Telephone Corporation

InventorsKosuke Nishida, Kyosuke Nishida, Hisako Asano, Junji Tomita

Life of this patent

Estimating the term…

Loading patent…

Abstract

To make it possible to accurately generate a word vector even if vocabulary of a word vector data set is not limited. In a vector generating device 10 that generates vectors representing an input sentence P, when generating a series of the vectors representing the input sentence P based on vectors corresponding to words included in the input sentence P, a definition-sentence-considered-context encode unit 280 generates, based on a dictionary DB 230 storing sets of headwords y and definition sentences Dy, which are sentences defining the headwords y, concerning a word, which is the headword stored in the dictionary DB, among the words included in the input sentence P, the series of the vectors representing the input sentence P using the definition sentence Dy of the headwords y.

Claims20

Dependent claims

A computer-implemented method for processing texts, the method comprising:

receiving an input text, wherein the input text comprises a headword;
receiving, based on the headword, a definition sentence from the dictionary database, wherein the dictionary database comprises one or more sets of headwords and definitions of the headwords, and wherein the definition sentence defines the headword;
and
generating, based on the input text, a first series of vectors representing the input text;
generating, based on the definition sentence, a second series of vectors representing the definition sentence of the headword;
combining, based at least on a substitution of a vector for the headword in the input text with the second series of vectors, the first series of vectors and the second series of vectors according to an emphasis weight, wherein the emphasis weight is based on semantic similarly between a first set of words in the input text and a second set of words in the definition sentence;
generating, based on the combined first series of vectors and the second series of vectors, an output series of vectors, the output series of vectors representing the input text based on the definition sentence for words of the headword for improved accuracy of semantics of the input text;
training, using the generated output series of vectors for training, a neural network for estimating an implication relation between texts;
and
providing, based on the trained neural network, the implication relation between a plurality of texts.

7 more elements

A system for processing texts, the system comprises:

a processor;
and
a memory storing computer-executable instructions that when executed by the processor cause the system to:
receive an input text, wherein the input text comprises a headword;
receive, based on the headword, a definition sentence from the dictionary database, wherein the dictionary database comprises one or more sets of headwords and definitions of the headwords, and wherein the definition sentence defines the headword;
and
generate, based on the input text, a first series of vectors representing the input text;
generate, based on the definition sentence, a second series of vectors representing the definition sentence of the headword;
combine, based at least on a substitution of a vector for the headword in the input text with the second series of vectors, the first series of vectors and the second series of vectors according to an emphasis weight, wherein the emphasis weight is based on a semantic similarly between a first set of words in the input text and a second set of words in the definition sentence;
generate, based on the combined first series of vectors and the second series of vectors, an output series of vectors, the output series vectors representing the input text based on the definition sentence for words of the headword for improved accuracy of semantics of the input text;
train, using the generated output series of vectors for training, a neural network for estimating an implication relation between texts;
and
provide, based on the trained neural network, the implication relation between a plurality of texts.

10 more elements

A computer-readable non-transitory recording medium storing computer-executable instructions that when executed by a processor cause a computer system to:

receive an input text, wherein the input text comprises a headword;
receive, based on the headword, a definition sentence from the dictionary database,
wherein the dictionary database comprises one or more sets of headwords and definitions of the headwords, and wherein the definition sentence defines the headword;
and
generate, based on the input text, a first series of vectors representing the input text;
generate, based on the definition sentence, a second series of vectors representing the definition sentence of the headword;
combine, based at least on a substitution of a vector for the headword in the input text with the second series of vectors, the first series of vectors and the second series of vectors;
generate, based on the combined first series of vectors and the second series of vectors, an output series of vectors, the output series of vectors representing the input text based on the definition sentence for words of the headword for improved accuracy of semantics of the input text according to an emphasis weight, wherein the emphasis weight is based on a semantic similarly between a first set of words in the input text and a second set of words in the definition sentence;
train, using the generated output series of vectors for training, a neural network for estimating an implication relation between texts;
and
provide, based on the trained neural network, the implication relation between a plurality of texts.

8 more elements

Dependencies are inferred from the claim text.