Presentation 2022-03-07
Generating the next action from sequence event pairs automatically acquired from Twitter
Haruka Enomoto, Kazutaka Shimada,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recently, natural language inference have been actively studied and it's mainly solved as a classification task. However, classification models can only select answers that are included in the given choices. Therefore, it doesn't truly solve the inference task. In this paper, we propose a method to generate the next actions from a previous action. Generally, we need a large number of training data to learn a generation model. However, there are no data for the generation task of commonsense reasoning. Therefore, we automatically collect event-sentence pairs from Twitter. and we conduct a generation experiment. We apply the extracted data to neural network-based models; seq2seq with attention and T5. One problem of this task is to evaluate the applicability of the generated sentences. In this paper, we also propose an evaluation measure based on Levenshtein distance. Through the measure, we compare the outputs from two generation methods.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Natural Language Inference / Commonsense Reasoning / Text Generation
Paper # NLC2021-39
Date of Issue 2022-02-28 (NLC)

Conference Information
Committee NLC
Conference Date 2022/3/7(1days)
Place (in Japanese) (See Japanese page)
Place (in English) Online
Topics (in Japanese) (See Japanese page)
Topics (in English) Information Processing of Tourism, etc.
Chair Kazutaka Shimada(Kyushu Inst. of Tech.)
Vice Chair Mitsuo Yoshida(Toyohashi Univ. of Tech.) / Takeshi Kobayakawa(NHK)
Secretary Mitsuo Yoshida(Univ. of Tokyo) / Takeshi Kobayakawa(Hiroshima Univ. of Economics)
Assistant Kanjin Takahashi(Sansan) / Ko Mitsuda(NTT)

Paper Information
Registration To Technical Committee on Natural Language Understanding and Models of Communication
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Generating the next action from sequence event pairs automatically acquired from Twitter
Sub Title (in English)
Keyword(1) Natural Language Inference
Keyword(2) Commonsense Reasoning
Keyword(3) Text Generation
1st Author's Name Haruka Enomoto
1st Author's Affiliation Kyushu Institute of Technology(Kyutech)
2nd Author's Name Kazutaka Shimada
2nd Author's Affiliation Kyushu Institute of Technology(Kyutech)
Date 2022-03-07
Paper # NLC2021-39
Volume (vol) vol.121
Number (no) NLC-415
Page pp.pp.62-67(NLC),
#Pages 6
Date of Issue 2022-02-28 (NLC)