Exploiting end of sentences and speaker alternations in language modeling for multiparty conversations
Exploiting end of sentences and speaker alternations in language modeling for multiparty conversations
复制标题
DOI：
10.1109/apsipa.2017.8282217
复制
发表时间：
2017-12
期刊：
影响因子：
--
通讯作者：
Hiroto Ashikawa;Naohiro Tawara;A. Ogawa;Tomoharu Iwata;Tetsunori Kobayashi;Tetsuji Ogawa
中科院分区：
文献类型：
--
作者：
Hiroto Ashikawa;Naohiro Tawara;A. Ogawa;Tomoharu Iwata;Tetsunori Kobayashi;Tetsuji Ogawa
本文研究了在递归神经网络语言模型（rnnlm）中对多方对话中经常出现的句子结束和说话人交替的有效处理。这种辅助信息被表示为上下文线索和特征向量。前一种表示可以将目录插入到转录中并作为单词标记，而后一种表示则作为神经网络的辅助输入。使用包括AMI会议语料库在内的多方对话数据进行的实验比较表明，两种表示都有助于改进rnnlm，并且处理句尾非常重要，特别是在多方对话中。
The effective handling of end-of-sentences and speaker alternations, both of which are frequently observed in multiparty conversations, in recurrent neural network language models (RNNLMs) is investigated. This kind of auxiliary information is represented as context cues and feature vectors. The former representation can be inserted directory into a transcription and treated as a word token, while the latter serves as auxiliary input to the neural networks. Experimental comparisons using multiparty conversation data, including the AMI meeting corpus, demonstrated that both representations contribute to improvement of the RNNLMs, and that dealing with the end-of-sentences is important, especially on the multiparty conversations.