Identification of Evaluation Collocation Based on Maximum Entropy Model

  • LingYun Zhao
  • FangAi Liu
  • Zhenfang Zhu
Conference paper
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 269)


In the process of analyzing the orientation of hotel comment, some opinion-bearing words may cause ambiguity. In this paper a method of evaluation collocation identification based on maximum entropy is proposed. This method designed a sentiment word table, mined the category of opinion-bearing words as semantic feature, combined this feature with lexical, part-of-speech, position and negative adverbs to construct a compound template, and then employed maximum entropy model to implement evaluation collocation identification. Experimental results show that the accuracy is higher when using the compound template constructed in this paper to identify evaluation collocation.


Orientation Evaluation collocation Maximum entropy Sentiment word table Semantic feature 


  1. 1.
    Popescu AM, Etzioni O (2007) Extracting product features and opinions from reviews. In: Natural language processing and text mining, Springer, London, pp 9–28 Google Scholar
  2. 2.
    Liu B et al (2005) Opinion observer: analyzing and comparing opinions on the web. In: Proceedings of the 14th internationa1 conference on World Wide Web. ACM. New York. pp 342–351Google Scholar
  3. 3.
    Zhao YY et al (2010) A survey of sentiment analysis. J Softw 21(8):1834–1848CrossRefGoogle Scholar
  4. 4.
    FAN N et al (2010) Extraction of subjective relation in opinion sentences based on maximum entropy model. Comput Eng 36(2):4–6Google Scholar
  5. 5.
    Della Pietra S et al (1992) Adaptive language modeling using minimum discriminant estimation. In: HLT ‘91 proceedings of the workshop on speech and natural language. pp 103–106Google Scholar
  6. 6.
    Darroch JN, Ratcliff D (1972) Generalized iterative scaling for log-linear models. Ann Math Statist 43(5):1470–1480CrossRefzbMATHMathSciNetGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2014

Authors and Affiliations

  1. 1.Business SchoolShandong Jianzhu UniversityJinanChina
  2. 2.School of Information Science and EngineeringShandong Normal UniversityJinanChina
  3. 3.School of Information Science and Electric EngineeringShandong Jiaotong UniversityJinanChina

Personalised recommendations