Open AccessAn Intelligent Multi-turn Dialogue Classification Method Based on Resampling
Author(s) -
Ming Guo,
Yunju Zhang,
Qiang Yang,
Guangyou Shen
Publication year - 2022
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2218/1/012034
Subject(s) - resampling , computer science , redundancy (engineering) , focus (optics) , turn taking , artificial intelligence , core (optical fiber) , turn (biochemistry) , segmentation , machine learning , natural language processing , pattern recognition (psychology) , linguistics , telecommunications , conversation , physics , philosophy , biochemistry , chemistry , optics , operating system
Intelligent multi-turn dialogue classification uses the whole text as input to predict tags. As one of the most popular research topics in dialogue system, intelligent multi-turn dialogue classification has important research significance in academia and industry. In this paper, we focus on removing redundant information from dialogue text. Multi-turn dialogue classification is essentially a text classification problem. We selects Bert as the core model of multi-turn dialogue classification in view of its strong learning and redundancy removing ability. However, Bert limits the length of its input to 512 and the length of dialogues often exceeds 512. A long text segmentation method based on resampling is presented to solve the problem of input length limitation. The resampling mechanism can also further remove the redundant information in multi-turn dialogues. Experimental results show that the proposed method outperforms all the baselines and achieves an F 1 of 53.91%.