ANTUSD: A Large Chinese Sentiment Dictionary

ANTUSD: A Large Chinese Sentiment Dictionary Shih-Ming Wang and Lun-Wei Ku Academia Sinica 128 Academia Road, Section 2, Nankang, Taipei 11529, Taiwan E-mail: ipod825@gmail.com, lwku@iis.sinica.edu.tw Abstract This paper introduces the augmented NTU sentiment dictionary, abbreviated as ANTUSD, which is constructed by collecting sentiment stats of words in several sentiment annotation work. A total of 26,021 words were collected in ANTUSD. For each word, the CopeOpi numerical sentiment score and the number of positive annotation, neutral annotation, negative annotation, non-opinionated annotation, and not-a-word annotation are provided. Words and their sentiment information in ANTUSD have been linked to the Chinese ontology E-HowNet to provide rich semantic information. We demonstrate the usage of ANTUSD in polarity classification of words, and the results show that a superior f-score 98.21 is achieved, which supports the usefulness of the ANTUSD. ANTUSD can be freely obtained through application from NLPSA lab, Academia Sinica: http://academiasinicanlplab.github.io/ Keywords: opinion dictionary, sentiment dictionary, NTUSD, ANTUSD 1. Introduction Sentiment analysis and opinion mining has been an important sub-discipline in natural language processing, information retrieval and machine learning. Related techniques have many applications, such as product/travel/movie recommendation, automatic opinion poll, melancholia detection, etc., so not only researchers try to build cutting edge systems to find sentiment information, but industries are searching for good solutions. An especially useful resource for both the research and industrial communities is a big and complete sentiment dictionary which can serve as the material for multiple purposes, e.g., markers of opinions or machine learning features. However, we are lacking this kind of resources for the Chinese language. In this paper we introduce the augmented NTU sentiment dictionary, abbreviated as ANTUSD, which is constructed by collecting sentiment stats of words in several sentiment annotation work. It collects 27,221 words. For each word, the CopeOpi numerical sentiment score and the number of positive annotation, neutral annotation, negative annotation, non- opinionated annotation, not-a-word annotation are provided. To extend the usage of ANTUSD, it has been connected to a large Chinese ontology, E-HowNet, to provide rich semantic information to the dictionary entries. In the last part of this paper, we show that using ANTUSD in sentiment word detection and word polarity classification both achieve good results. 2. Related Materials ANTUSD was built mainly by collecting manual annotation of words in the process of building several sentiment corpora in a long period of time from year 2006 to year 2010. Therefore, before we describe the construction of ANTUSD, materials involved in this process are first introduced: NTUSD, NTCIR MOAT task dataset, Chinese Opinion Treebank (L.-W. Ku, Huang, & Chen, 2010), AciBiMA (T.-H. Huang, Ku, & Chen, 2010; T.-H. K. Huang, Chen, & Kong, 2015), CopeOpi and E- 1 http://research.nii.ac.jp/ntcir/index-en.html HowNet, where NTUSD is a sentiment dictionary, NTCIR MOAT task dataset and Chinese opinion Treebank are two manual labeled opinion sentence datasets, CopeOpi is a Chinese opinion scoring system, and E-HowNet is a Chinese knowledge ontology, described as follows: NTUSD NTUSD (L.-W. Ku, Liang, & Chen, 2006) is the prototype sentiment dictionary of ANTUSD. It …

Content truncated to 500 words