An Enhancement Method for Chinese Environment Semantic Slot Filling Based on POS Tagging
Shaoyong Qu, Weifeng Liu, Jianning Li, Zhangming Peng · 2021 International Conference on Control, Automation and Information Sciences (ICCAIS) · 2021
To solve the problem of Chinese semantic understanding and extraction, this paper presents an enhancement method for Chinese environment semantic Slot Filling based on part-of-speech(POS) tagging. Firstly, construct a special dictionary with a given task domain where each word has its own part of speech and word frequency tagging; Secondly, based on the Chinese word segmentation technique called Jieba, Chinese sentences are partitioned to obtain a finite sequence with words as the basic elements. And then, a new finite sequence is obtained by replacing the words with the same POS tagging in the finite sequence with the same symbol. Finally, the Bidirectional Long Short-Term Memory with Conditional Random Fields(BLSTM-CRF) network training model is used with a new limited sequence set and its tag sequence set as input, and the slot filling criteria based on POS tagging are appended. Applying popular data set ATIS training model, it shows that the method has high accuracy and reliability through simulation comparison experiment.