Pinyin Based Fuzzy Matching for Semantic Class Labeling in Chinese Spoken Understanding
Yanling Li, Yonghong Yan · International Conference on Business Computing and Global Informatization · 2013
In this paper we propose an approach to solve the words' variation induced by automatic speech recognition (ASR) and missing field of key information by users in the process of spoken language understanding (SLU). Considering the characteristics of Chinese, fuzzy matching based on Chinese pinyin is utilized to correct the semantic concepts in a natural language query. The approach is in two stages: first, conditional random field (CRF) model is trained for building probabilistic models to segment and label entity names from an input sentence. Second, similarity measure is conducted through Chinese pinyin of both the named entities labeled by a CRF model and lexicon of dictionary. The experimental results report that the approach based on Chinese pinyin is better than the one based on Chinese characters in actual human-computer interaction systems.