PHWordSet: the Chinese Sentiment Lexicon for Emotional Analysis of Public Health Emergencies

In this paper, we propose a method to automatically construct a target-specific sentiment lexicon. The core of our method is a unified framework that incorporates three kinds of methods for sentiment lexicon construction, i.e., seed emotional words are extracted from the existing general sentiment lexicon, candidate emotional words are extracted by calculating cosine similarity based on the Word2vec model, and candidate emotional words are extracted from large-scale public opinion corpus based on syntactic relationship. To avoid random propagation of label information, the LeaderRank algorithm is introduced to calculate the emphasis degree between nodes and update the emotional information of words in a certain order. Finally, the PHWordSet (sentiment lexicon of public health public opinion) is obtained. A test set containing 2000 public health data is randomly constructed, and the experimental results on the constructed public health dataset prove the effectiveness of this method. © 2023, Tamkang University. All rights reserved.