ANSWER CLASSIFIER AND REPRESENTATION GENERATOR FOR QUESTION-ANSWERING SYSTEM USING GAN, AND COMPUTER PROGRAM FOR TRAINING THE REPRESENTATION GENERATOR | Matter42 Literature
Patent
Atlas literature
Patent
US 12,099,801 B2
ANSWER CLASSIFIER AND REPRESENTATION GENERATOR FOR QUESTION-ANSWERING SYSTEM USING GAN, AND COMPUTER PROGRAM FOR TRAINING THE REPRESENTATION GENERATOR
Jonghoon Oh, Kazuma Kadowaki, Julien Kloetzer, Ryu Iida et al.
National Institute of Information and Communications Technology, Tokyo (JP)·Sep. 24, 2024·US
Drawings
Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1
FIG. 1 is a block diagram schematically showing a configuration of a why-type question-answering system dis- closed in Patent Literature 1.
FIG. 2
FIG. 2 shows a question and a positive example of an answer passage to the question.
FIG. 3
FIG. 3 shows a question and a positive example of an answer passage to the question.
FIG. 4
FIG. 4 shows a question and a negative example of an answer passage to the question.
FIG. 5
FIG. 5 shows a question, a positive example of an answer passage to the question, and a core answer prepared based on the positive example.
FIG. 6
FIG. 6 is a schematic illustration showing a scheme of typical generative adversarial network.
FIG. 7
FIG. 7 is a schematic illustration showing a scheme of a virtual system that trains a generator for forming a core answer similar to a core answer manually …
FIG. 8
FIG. 8.
FIG. 9
FIG. 9.
FIG. 10
FIG. 10 is a flowchart showing a control structure of a 10 routine for parameter training of the fake representation generator in the program shown in
FIG. 11
FIG. 11 is a block diagram showing a configuration of an answer classifier determining whether a passage is a correct answer to the question or not, using the …
FIG. 12
FIG. 12 is a block diagram showing a basic configuration of an encoder for forming the fake representation generator 20 shown in
FIG. 13
performance graph
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
FIG. 14
FIG. 14 is the same table as
FIG. 15
FIG. 15 is the same table as
FIG. 16
FIG. 16 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a second embodiment of the …
FIG. 17
FIG. 17 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a third embodiment of the …
FIG. 18
FIG. 18 shows, in the form of a table, accuracy of answer 45 classifiers adopting the fake representation generators of the first, second and third embodiments …
FIG. 19
FIG. 19.
FIG. 20
FIG. 20 is a block diagram showing a configuration of the Open QA system shown in
FIG. 21
FIG. 21 is a schematic illustration showing a process by the Open QA system adopting the fake representation gen- erator in accordance with an embodiment of …
FIG. 22
FIG. 22 is a block diagram showing a configuration of the Open QA system adopting the fake representation generator in accordance with an embodiment of the …
FIG. 23
FIG. 23. B₂
FIG. 24
FIG. 24 is a block diagram showing a hardware configu- 65 ration of the computer system of which appearance is shown in
FIG. 25
FIG. 25 4 has a text fragment 102 related to question 90, other text fragments, particularly those underlined 104 are not related to the question nor to the …
Claims
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
1 independent · 4 dependent
1
Independentanswer classifier
An answer classifier, comprising: a representation gen-erator trained by a computer program configured to cause a computer to operate as: a first representation generator, upon receiving a question in natural language and an input forming a pair with the question, outputting a first representation vector repre-senting the input; a second representation generator, upon receiving the question and an answer to the question, outputting a second representation vector representing said answer in a format same as said first representation vector; a discriminator responsive to said first representation vector or said second representation vector received at an input for determining whether an input representa-tion vector is the first representation vector or the second representation vector; and a generative adversarial network unit for training said discriminator and said first representation generator by said generative adversarial network unit such that error determination of said first representation vector is maximized and error determination of said second representation vector is minimized, the representation generator responsive to receipt of a question and a passage including an answer to the question at an input, for outputting a first representation vector obtained from the passage, representing an answer to said question, wherein the answer classifier further comprises: a passage encoder responsive to receipt of said passage, said first representation vector and said question at an input, for outputting a representation vector encoding said passage, having an attention by said first repre-sentation vector and said question added; a question encoder responsive to receipt of said question and said passage, for outputting a representation vector of said question having an attention by said passage added; and a determiner trained beforehand such that upon receiving said first representation vector, the representation vec-tor of said passage and the representation vector of said question, said determiner classifies said passage as a correct answer or an erroneous answer to said question.
2
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences including an answer to the question, for outputting said first representa-tion vector representing said answer to said question from said passage.
3
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences selected at random for outputting said first representation vector representing said answer to said question from said passage and said question.
4
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a random vector consisting of random elements for outputting said first representation vector representing said answer to said question from said random vector and said question.
5
Dependent← claim 1
A representation generator trained by the computer program according to claim 1. ∗ ∗ ∗ ∗ ∗
Device structures
Layer stacks claimed or described, ordered top of device to substrate.
answer classifier
determinerdeterminer
questionencoderquestion encoder
passageencoderpassage encoder
generativeadversarialnetworkunitgenerative adversarial network unit
discriminatordiscriminator
Characterization
Measurements and analyses referenced in the patent, with their drawing references.
device performance measurement
Device Performance Measurement
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
ANSWER CLASSIFIER AND REPRESENTATION GENERATOR FOR QUESTION-ANSWERING SYSTEM USING GAN, AND COMPUTER PROGRAM FOR TRAINING THE REPRESENTATION GENERATOR
Jonghoon Oh, Kazuma Kadowaki, Julien Kloetzer, Ryu Iida et al.
National Institute of Information and Communications Technology, Tokyo (JP)·Sep. 24, 2024·US
Drawings
Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1
FIG. 1 is a block diagram schematically showing a configuration of a why-type question-answering system dis- closed in Patent Literature 1.
FIG. 2
FIG. 2 shows a question and a positive example of an answer passage to the question.
FIG. 3
FIG. 3 shows a question and a positive example of an answer passage to the question.
FIG. 4
FIG. 4 shows a question and a negative example of an answer passage to the question.
FIG. 5
FIG. 5 shows a question, a positive example of an answer passage to the question, and a core answer prepared based on the positive example.
FIG. 6
FIG. 6 is a schematic illustration showing a scheme of typical generative adversarial network.
FIG. 7
FIG. 7 is a schematic illustration showing a scheme of a virtual system that trains a generator for forming a core answer similar to a core answer manually …
FIG. 8
FIG. 8.
FIG. 9
FIG. 9.
FIG. 10
FIG. 10 is a flowchart showing a control structure of a 10 routine for parameter training of the fake representation generator in the program shown in
FIG. 11
FIG. 11 is a block diagram showing a configuration of an answer classifier determining whether a passage is a correct answer to the question or not, using the …
FIG. 12
FIG. 12 is a block diagram showing a basic configuration of an encoder for forming the fake representation generator 20 shown in
FIG. 13
performance graph
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
FIG. 14
FIG. 14 is the same table as
FIG. 15
FIG. 15 is the same table as
FIG. 16
FIG. 16 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a second embodiment of the …
FIG. 17
FIG. 17 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a third embodiment of the …
FIG. 18
FIG. 18 shows, in the form of a table, accuracy of answer 45 classifiers adopting the fake representation generators of the first, second and third embodiments …
FIG. 19
FIG. 19.
FIG. 20
FIG. 20 is a block diagram showing a configuration of the Open QA system shown in
FIG. 21
FIG. 21 is a schematic illustration showing a process by the Open QA system adopting the fake representation gen- erator in accordance with an embodiment of …
FIG. 22
FIG. 22 is a block diagram showing a configuration of the Open QA system adopting the fake representation generator in accordance with an embodiment of the …
FIG. 23
FIG. 23. B₂
FIG. 24
FIG. 24 is a block diagram showing a hardware configu- 65 ration of the computer system of which appearance is shown in
FIG. 25
FIG. 25 4 has a text fragment 102 related to question 90, other text fragments, particularly those underlined 104 are not related to the question nor to the …
Claims
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
1 independent · 4 dependent
1
Independentanswer classifier
An answer classifier, comprising: a representation gen-erator trained by a computer program configured to cause a computer to operate as: a first representation generator, upon receiving a question in natural language and an input forming a pair with the question, outputting a first representation vector repre-senting the input; a second representation generator, upon receiving the question and an answer to the question, outputting a second representation vector representing said answer in a format same as said first representation vector; a discriminator responsive to said first representation vector or said second representation vector received at an input for determining whether an input representa-tion vector is the first representation vector or the second representation vector; and a generative adversarial network unit for training said discriminator and said first representation generator by said generative adversarial network unit such that error determination of said first representation vector is maximized and error determination of said second representation vector is minimized, the representation generator responsive to receipt of a question and a passage including an answer to the question at an input, for outputting a first representation vector obtained from the passage, representing an answer to said question, wherein the answer classifier further comprises: a passage encoder responsive to receipt of said passage, said first representation vector and said question at an input, for outputting a representation vector encoding said passage, having an attention by said first repre-sentation vector and said question added; a question encoder responsive to receipt of said question and said passage, for outputting a representation vector of said question having an attention by said passage added; and a determiner trained beforehand such that upon receiving said first representation vector, the representation vec-tor of said passage and the representation vector of said question, said determiner classifies said passage as a correct answer or an erroneous answer to said question.
2
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences including an answer to the question, for outputting said first representa-tion vector representing said answer to said question from said passage.
3
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences selected at random for outputting said first representation vector representing said answer to said question from said passage and said question.
4
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a random vector consisting of random elements for outputting said first representation vector representing said answer to said question from said random vector and said question.
5
Dependent← claim 1
A representation generator trained by the computer program according to claim 1. ∗ ∗ ∗ ∗ ∗
Device structures
Layer stacks claimed or described, ordered top of device to substrate.
answer classifier
determinerdeterminer
questionencoderquestion encoder
passageencoderpassage encoder
generativeadversarialnetworkunitgenerative adversarial network unit
discriminatordiscriminator
Characterization
Measurements and analyses referenced in the patent, with their drawing references.
device performance measurement
Device Performance Measurement
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
ANSWER CLASSIFIER AND REPRESENTATION GENERATOR FOR QUESTION-ANSWERING SYSTEM USING GAN, AND COMPUTER PROGRAM FOR TRAINING THE REPRESENTATION GENERATOR
Jonghoon Oh, Kazuma Kadowaki, Julien Kloetzer, Ryu Iida et al.
National Institute of Information and Communications Technology, Tokyo (JP)·Sep. 24, 2024·US
Drawings
Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1
FIG. 1 is a block diagram schematically showing a configuration of a why-type question-answering system dis- closed in Patent Literature 1.
FIG. 2
FIG. 2 shows a question and a positive example of an answer passage to the question.
FIG. 3
FIG. 3 shows a question and a positive example of an answer passage to the question.
FIG. 4
FIG. 4 shows a question and a negative example of an answer passage to the question.
FIG. 5
FIG. 5 shows a question, a positive example of an answer passage to the question, and a core answer prepared based on the positive example.
FIG. 6
FIG. 6 is a schematic illustration showing a scheme of typical generative adversarial network.
FIG. 7
FIG. 7 is a schematic illustration showing a scheme of a virtual system that trains a generator for forming a core answer similar to a core answer manually …
FIG. 8
FIG. 8.
FIG. 9
FIG. 9.
FIG. 10
FIG. 10 is a flowchart showing a control structure of a 10 routine for parameter training of the fake representation generator in the program shown in
FIG. 11
FIG. 11 is a block diagram showing a configuration of an answer classifier determining whether a passage is a correct answer to the question or not, using the …
FIG. 12
FIG. 12 is a block diagram showing a basic configuration of an encoder for forming the fake representation generator 20 shown in
FIG. 13
performance graph
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
FIG. 14
FIG. 14 is the same table as
FIG. 15
FIG. 15 is the same table as
FIG. 16
FIG. 16 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a second embodiment of the …
FIG. 17
FIG. 17 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a third embodiment of the …
FIG. 18
FIG. 18 shows, in the form of a table, accuracy of answer 45 classifiers adopting the fake representation generators of the first, second and third embodiments …
FIG. 19
FIG. 19.
FIG. 20
FIG. 20 is a block diagram showing a configuration of the Open QA system shown in
FIG. 21
FIG. 21 is a schematic illustration showing a process by the Open QA system adopting the fake representation gen- erator in accordance with an embodiment of …
FIG. 22
FIG. 22 is a block diagram showing a configuration of the Open QA system adopting the fake representation generator in accordance with an embodiment of the …
FIG. 23
FIG. 23. B₂
FIG. 24
FIG. 24 is a block diagram showing a hardware configu- 65 ration of the computer system of which appearance is shown in
FIG. 25
FIG. 25 4 has a text fragment 102 related to question 90, other text fragments, particularly those underlined 104 are not related to the question nor to the …
Claims
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
1 independent · 4 dependent
1
Independentanswer classifier
An answer classifier, comprising: a representation gen-erator trained by a computer program configured to cause a computer to operate as: a first representation generator, upon receiving a question in natural language and an input forming a pair with the question, outputting a first representation vector repre-senting the input; a second representation generator, upon receiving the question and an answer to the question, outputting a second representation vector representing said answer in a format same as said first representation vector; a discriminator responsive to said first representation vector or said second representation vector received at an input for determining whether an input representa-tion vector is the first representation vector or the second representation vector; and a generative adversarial network unit for training said discriminator and said first representation generator by said generative adversarial network unit such that error determination of said first representation vector is maximized and error determination of said second representation vector is minimized, the representation generator responsive to receipt of a question and a passage including an answer to the question at an input, for outputting a first representation vector obtained from the passage, representing an answer to said question, wherein the answer classifier further comprises: a passage encoder responsive to receipt of said passage, said first representation vector and said question at an input, for outputting a representation vector encoding said passage, having an attention by said first repre-sentation vector and said question added; a question encoder responsive to receipt of said question and said passage, for outputting a representation vector of said question having an attention by said passage added; and a determiner trained beforehand such that upon receiving said first representation vector, the representation vec-tor of said passage and the representation vector of said question, said determiner classifies said passage as a correct answer or an erroneous answer to said question.
2
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences including an answer to the question, for outputting said first representa-tion vector representing said answer to said question from said passage.
3
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences selected at random for outputting said first representation vector representing said answer to said question from said passage and said question.
4
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a random vector consisting of random elements for outputting said first representation vector representing said answer to said question from said random vector and said question.
5
Dependent← claim 1
A representation generator trained by the computer program according to claim 1. ∗ ∗ ∗ ∗ ∗
Device structures
Layer stacks claimed or described, ordered top of device to substrate.
answer classifier
determinerdeterminer
questionencoderquestion encoder
passageencoderpassage encoder
generativeadversarialnetworkunitgenerative adversarial network unit
discriminatordiscriminator
Characterization
Measurements and analyses referenced in the patent, with their drawing references.
device performance measurement
Device Performance Measurement
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
ANSWER CLASSIFIER AND REPRESENTATION GENERATOR FOR QUESTION-ANSWERING SYSTEM USING GAN, AND COMPUTER PROGRAM FOR TRAINING THE REPRESENTATION GENERATOR
Jonghoon Oh, Kazuma Kadowaki, Julien Kloetzer, Ryu Iida et al.
National Institute of Information and Communications Technology, Tokyo (JP)·Sep. 24, 2024·US
Drawings
Patent drawings and their descriptions. Click a drawing to enlarge it.
FIG. 1
FIG. 1 is a block diagram schematically showing a configuration of a why-type question-answering system dis- closed in Patent Literature 1.
FIG. 2
FIG. 2 shows a question and a positive example of an answer passage to the question.
FIG. 3
FIG. 3 shows a question and a positive example of an answer passage to the question.
FIG. 4
FIG. 4 shows a question and a negative example of an answer passage to the question.
FIG. 5
FIG. 5 shows a question, a positive example of an answer passage to the question, and a core answer prepared based on the positive example.
FIG. 6
FIG. 6 is a schematic illustration showing a scheme of typical generative adversarial network.
FIG. 7
FIG. 7 is a schematic illustration showing a scheme of a virtual system that trains a generator for forming a core answer similar to a core answer manually …
FIG. 8
FIG. 8.
FIG. 9
FIG. 9.
FIG. 10
FIG. 10 is a flowchart showing a control structure of a 10 routine for parameter training of the fake representation generator in the program shown in
FIG. 11
FIG. 11 is a block diagram showing a configuration of an answer classifier determining whether a passage is a correct answer to the question or not, using the …
FIG. 12
FIG. 12 is a block diagram showing a basic configuration of an encoder for forming the fake representation generator 20 shown in
FIG. 13
performance graph
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
FIG. 14
FIG. 14 is the same table as
FIG. 15
FIG. 15 is the same table as
FIG. 16
FIG. 16 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a second embodiment of the …
FIG. 17
FIG. 17 is a schematic illustration showing a scheme of a system for training a fake representation generator in accor- dance with a third embodiment of the …
FIG. 18
FIG. 18 shows, in the form of a table, accuracy of answer 45 classifiers adopting the fake representation generators of the first, second and third embodiments …
FIG. 19
FIG. 19.
FIG. 20
FIG. 20 is a block diagram showing a configuration of the Open QA system shown in
FIG. 21
FIG. 21 is a schematic illustration showing a process by the Open QA system adopting the fake representation gen- erator in accordance with an embodiment of …
FIG. 22
FIG. 22 is a block diagram showing a configuration of the Open QA system adopting the fake representation generator in accordance with an embodiment of the …
FIG. 23
FIG. 23. B₂
FIG. 24
FIG. 24 is a block diagram showing a hardware configu- 65 ration of the computer system of which appearance is shown in
FIG. 25
FIG. 25 4 has a text fragment 102 related to question 90, other text fragments, particularly those underlined 104 are not related to the question nor to the …
Claims
Claims define the patent's legal scope. Independent claims stand alone; dependent claims (nested) narrow them. Click a claim to expand its dependents.
1 independent · 4 dependent
1
Independentanswer classifier
An answer classifier, comprising: a representation gen-erator trained by a computer program configured to cause a computer to operate as: a first representation generator, upon receiving a question in natural language and an input forming a pair with the question, outputting a first representation vector repre-senting the input; a second representation generator, upon receiving the question and an answer to the question, outputting a second representation vector representing said answer in a format same as said first representation vector; a discriminator responsive to said first representation vector or said second representation vector received at an input for determining whether an input representa-tion vector is the first representation vector or the second representation vector; and a generative adversarial network unit for training said discriminator and said first representation generator by said generative adversarial network unit such that error determination of said first representation vector is maximized and error determination of said second representation vector is minimized, the representation generator responsive to receipt of a question and a passage including an answer to the question at an input, for outputting a first representation vector obtained from the passage, representing an answer to said question, wherein the answer classifier further comprises: a passage encoder responsive to receipt of said passage, said first representation vector and said question at an input, for outputting a representation vector encoding said passage, having an attention by said first repre-sentation vector and said question added; a question encoder responsive to receipt of said question and said passage, for outputting a representation vector of said question having an attention by said passage added; and a determiner trained beforehand such that upon receiving said first representation vector, the representation vec-tor of said passage and the representation vector of said question, said determiner classifies said passage as a correct answer or an erroneous answer to said question.
2
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences including an answer to the question, for outputting said first representa-tion vector representing said answer to said question from said passage.
3
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a passage including one or more sentences selected at random for outputting said first representation vector representing said answer to said question from said passage and said question.
4
Dependent← claim 1
The computer program according to claim 1, wherein said first representation generator includes a vector output-ting means responsive to receipt of said question and a random vector consisting of random elements for outputting said first representation vector representing said answer to said question from said random vector and said question.
5
Dependent← claim 1
A representation generator trained by the computer program according to claim 1. ∗ ∗ ∗ ∗ ∗
Device structures
Layer stacks claimed or described, ordered top of device to substrate.
answer classifier
determinerdeterminer
questionencoderquestion encoder
passageencoderpassage encoder
generativeadversarialnetworkunitgenerative adversarial network unit
discriminatordiscriminator
Characterization
Measurements and analyses referenced in the patent, with their drawing references.
device performance measurement
Device Performance Measurement
FIG. 13, indicating that the classification performance of the answer classifier using the fake representation generator in accordance with the first embodiment …
International Search Report for corresponding Application No. PCT/JP2020/026360, mailed Sep. 24, 2020.
Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts. Jong-Hoon Oh et al., “Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts”, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 4227-4237, Florence, Italy, Jul. 28-Aug. 2, 2019. Suzan Verberne et al., “What is not in the Bag of Words for Why-QA?”, Association for Computational Linguistics, pp. 229- 245, 2010. Jong-Hoon Oh et al., “Why Question Answering using Sentiment Analysis and Word Classes”, Proceedings of the 2012 Joint Con- ference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 368-378, Jeju Island, Korea, Jul. 12-14, 2012. Jong-Hoon Oh et al., “Why-Question Answering using Intra- and Inter-Sentential Causal Relations”, Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pp. 1733-1743, Sofia, Bulgaria, Aug. 4-9, 2013. Jong-Hoon Oh et al., “A Semi-Supervised Learning Approach to Why-Question Answering”, 2016, Association for the Advancement of Artificial Intelligence, pp. 3022-3029. Jong-Hoon Oh et al., “Multi-col. Convolutional Neural Networks with Causality-Attention for Why-Question Answering”, WSDM 2017, Feb. 6-10, 2017, Cambridge, United Kingdom, pp. 415-424.
International Search Report for corresponding Application No. PCT/JP2020/026360, mailed Sep. 24, 2020.
Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts. Jong-Hoon Oh et al., “Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts”, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 4227-4237, Florence, Italy, Jul. 28-Aug. 2, 2019. Suzan Verberne et al., “What is not in the Bag of Words for Why-QA?”, Association for Computational Linguistics, pp. 229- 245, 2010. Jong-Hoon Oh et al., “Why Question Answering using Sentiment Analysis and Word Classes”, Proceedings of the 2012 Joint Con- ference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 368-378, Jeju Island, Korea, Jul. 12-14, 2012. Jong-Hoon Oh et al., “Why-Question Answering using Intra- and Inter-Sentential Causal Relations”, Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pp. 1733-1743, Sofia, Bulgaria, Aug. 4-9, 2013. Jong-Hoon Oh et al., “A Semi-Supervised Learning Approach to Why-Question Answering”, 2016, Association for the Advancement of Artificial Intelligence, pp. 3022-3029. Jong-Hoon Oh et al., “Multi-col. Convolutional Neural Networks with Causality-Attention for Why-Question Answering”, WSDM 2017, Feb. 6-10, 2017, Cambridge, United Kingdom, pp. 415-424.
International Search Report for corresponding Application No. PCT/JP2020/026360, mailed Sep. 24, 2020.
Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts. Jong-Hoon Oh et al., “Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts”, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 4227-4237, Florence, Italy, Jul. 28-Aug. 2, 2019. Suzan Verberne et al., “What is not in the Bag of Words for Why-QA?”, Association for Computational Linguistics, pp. 229- 245, 2010. Jong-Hoon Oh et al., “Why Question Answering using Sentiment Analysis and Word Classes”, Proceedings of the 2012 Joint Con- ference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 368-378, Jeju Island, Korea, Jul. 12-14, 2012. Jong-Hoon Oh et al., “Why-Question Answering using Intra- and Inter-Sentential Causal Relations”, Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pp. 1733-1743, Sofia, Bulgaria, Aug. 4-9, 2013. Jong-Hoon Oh et al., “A Semi-Supervised Learning Approach to Why-Question Answering”, 2016, Association for the Advancement of Artificial Intelligence, pp. 3022-3029. Jong-Hoon Oh et al., “Multi-col. Convolutional Neural Networks with Causality-Attention for Why-Question Answering”, WSDM 2017, Feb. 6-10, 2017, Cambridge, United Kingdom, pp. 415-424.
International Search Report for corresponding Application No. PCT/JP2020/026360, mailed Sep. 24, 2020.
Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts. Jong-Hoon Oh et al., “Open-Domain Why-Question Answering withAdversarial Learning to EncodeAnswer Texts”, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 4227-4237, Florence, Italy, Jul. 28-Aug. 2, 2019. Suzan Verberne et al., “What is not in the Bag of Words for Why-QA?”, Association for Computational Linguistics, pp. 229- 245, 2010. Jong-Hoon Oh et al., “Why Question Answering using Sentiment Analysis and Word Classes”, Proceedings of the 2012 Joint Con- ference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 368-378, Jeju Island, Korea, Jul. 12-14, 2012. Jong-Hoon Oh et al., “Why-Question Answering using Intra- and Inter-Sentential Causal Relations”, Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pp. 1733-1743, Sofia, Bulgaria, Aug. 4-9, 2013. Jong-Hoon Oh et al., “A Semi-Supervised Learning Approach to Why-Question Answering”, 2016, Association for the Advancement of Artificial Intelligence, pp. 3022-3029. Jong-Hoon Oh et al., “Multi-col. Convolutional Neural Networks with Causality-Attention for Why-Question Answering”, WSDM 2017, Feb. 6-10, 2017, Cambridge, United Kingdom, pp. 415-424.