{"title":"多轮对话意图识别方法的研究与应用","authors":"Jie Song, Qifeng Luo, J. Nie","doi":"10.1109/CIS52066.2020.00036","DOIUrl":null,"url":null,"abstract":"In the existing dialogue system, there are numerous sentences in non-standardized verbal expression form, which usually is brief and vague. It is a challenging task to identify the intentions through the analysis of these sentences. Considering that the supervised learning approach is the mainstream on multi-intention recognition, an amount of public labeled multi-intention dialogue data is necessary. However, labeling work is costly and time-consuming. In this paper, we put forward a multi-label classification method based on existing mainstream classification algorithms and used for dialogue-level multi-intention recognition to reduce the cost of labeling work. We publish the Chinese Multi-Intention Dialogue (CMID-Transportation) dataset of transportation customer service, which is collected by us in an actual production project. We conduct a series of experiments on the CMID-Transportation corpus by using the mainstream classification algorithms and then produce the basic benchmark performance. We find that BERT achieves the best results. We hope that the CMID-Transportation dataset can promote the research and development of intent recognition tasks in multiple rounds of dialogue.","PeriodicalId":106959,"journal":{"name":"2020 16th International Conference on Computational Intelligence and Security (CIS)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Research and Application of Multi-Round Dialogue Intent Recognition Method\",\"authors\":\"Jie Song, Qifeng Luo, J. Nie\",\"doi\":\"10.1109/CIS52066.2020.00036\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In the existing dialogue system, there are numerous sentences in non-standardized verbal expression form, which usually is brief and vague. It is a challenging task to identify the intentions through the analysis of these sentences. Considering that the supervised learning approach is the mainstream on multi-intention recognition, an amount of public labeled multi-intention dialogue data is necessary. However, labeling work is costly and time-consuming. In this paper, we put forward a multi-label classification method based on existing mainstream classification algorithms and used for dialogue-level multi-intention recognition to reduce the cost of labeling work. We publish the Chinese Multi-Intention Dialogue (CMID-Transportation) dataset of transportation customer service, which is collected by us in an actual production project. We conduct a series of experiments on the CMID-Transportation corpus by using the mainstream classification algorithms and then produce the basic benchmark performance. We find that BERT achieves the best results. We hope that the CMID-Transportation dataset can promote the research and development of intent recognition tasks in multiple rounds of dialogue.\",\"PeriodicalId\":106959,\"journal\":{\"name\":\"2020 16th International Conference on Computational Intelligence and Security (CIS)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-11-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 16th International Conference on Computational Intelligence and Security (CIS)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/CIS52066.2020.00036\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 16th International Conference on Computational Intelligence and Security (CIS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/CIS52066.2020.00036","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Research and Application of Multi-Round Dialogue Intent Recognition Method
In the existing dialogue system, there are numerous sentences in non-standardized verbal expression form, which usually is brief and vague. It is a challenging task to identify the intentions through the analysis of these sentences. Considering that the supervised learning approach is the mainstream on multi-intention recognition, an amount of public labeled multi-intention dialogue data is necessary. However, labeling work is costly and time-consuming. In this paper, we put forward a multi-label classification method based on existing mainstream classification algorithms and used for dialogue-level multi-intention recognition to reduce the cost of labeling work. We publish the Chinese Multi-Intention Dialogue (CMID-Transportation) dataset of transportation customer service, which is collected by us in an actual production project. We conduct a series of experiments on the CMID-Transportation corpus by using the mainstream classification algorithms and then produce the basic benchmark performance. We find that BERT achieves the best results. We hope that the CMID-Transportation dataset can promote the research and development of intent recognition tasks in multiple rounds of dialogue.