A Chinese short text semantic similarity computation model based on stop words and TongyiciCilin

Abstract

Short text similarity computing plays an important role in natural language processing, and it can be applied to many tasks. In recent years, there are lots of researches getting important results on natural language processing. Although there are some good results in English, there is no major breakthrough in Chinese. Different from the proposed methods, we reserve the Stop words in the training dataset of word vector for Chinese characteristics, and add the TongyiciCilin to the training data of the short text semantic similarity computation model. We compared the effect of Word2vec and Glove methods in our model. We use the Chinese short text semantic similarity dataset which is designed by Chinese grammar experts. The results show that the accuracy of the model is improved by 2%–3% by retaining Stop words in word vector training data and adding TongyiciCilin to training data. The accuracy of our model is better than Baidu short text similarity calculation platform on the same testing dataset.