{"title":"基于BERT-BIGRU-CRF *的邻接对预测核函数研究","authors":"Xin Chen, Zhanzhi Qiu","doi":"10.1145/3483845.3483866","DOIUrl":null,"url":null,"abstract":"Based on how to predict the core functions of the antecedent and the afterword belongs to the adjacency pairs according to different speaker's turn in the process of dialogue, So as to achieve overall control of the dialogue process, The core function prediction model based on BERT-BIGRU-CRF is put forward. The inputs of the model are the different turns of the conversationalist, Firstly, the model used BERT get sentence of vectorization, then it is input into BIGRU to make the two directional encoding, Finally, CRF is used to form the mark of its core function and the output of core function by S, R, Q, A and DO adjacent is completed. Experiments show that this prediction method has better performance than the previous sequential labeling model and its accuracy and F1 score are significantly improved.","PeriodicalId":134636,"journal":{"name":"Proceedings of the 2021 2nd International Conference on Control, Robotics and Intelligent System","volume":"69 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Research on Core Function of Adjacency Pairs Prediction Based on BERT-BIGRU-CRF∗\",\"authors\":\"Xin Chen, Zhanzhi Qiu\",\"doi\":\"10.1145/3483845.3483866\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Based on how to predict the core functions of the antecedent and the afterword belongs to the adjacency pairs according to different speaker's turn in the process of dialogue, So as to achieve overall control of the dialogue process, The core function prediction model based on BERT-BIGRU-CRF is put forward. The inputs of the model are the different turns of the conversationalist, Firstly, the model used BERT get sentence of vectorization, then it is input into BIGRU to make the two directional encoding, Finally, CRF is used to form the mark of its core function and the output of core function by S, R, Q, A and DO adjacent is completed. Experiments show that this prediction method has better performance than the previous sequential labeling model and its accuracy and F1 score are significantly improved.\",\"PeriodicalId\":134636,\"journal\":{\"name\":\"Proceedings of the 2021 2nd International Conference on Control, Robotics and Intelligent System\",\"volume\":\"69 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-08-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the 2021 2nd International Conference on Control, Robotics and Intelligent System\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/3483845.3483866\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 2021 2nd International Conference on Control, Robotics and Intelligent System","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3483845.3483866","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Research on Core Function of Adjacency Pairs Prediction Based on BERT-BIGRU-CRF∗
Based on how to predict the core functions of the antecedent and the afterword belongs to the adjacency pairs according to different speaker's turn in the process of dialogue, So as to achieve overall control of the dialogue process, The core function prediction model based on BERT-BIGRU-CRF is put forward. The inputs of the model are the different turns of the conversationalist, Firstly, the model used BERT get sentence of vectorization, then it is input into BIGRU to make the two directional encoding, Finally, CRF is used to form the mark of its core function and the output of core function by S, R, Q, A and DO adjacent is completed. Experiments show that this prediction method has better performance than the previous sequential labeling model and its accuracy and F1 score are significantly improved.