Dissertation > Excellent graduate degree dissertation topics show

Extraction Collocation Relations of Product Opinion Target and Sentiment Words

Author: XuLiBo
Tutor: ZhouYanQuan
School: Beijing University of Posts and Telecommunications
Course: Computer technology
Keywords: Evaluation of collocation extraction Conditional RandomFields Word Alignment Random Walk
CLC: TP391.1
Type: Master's thesis
Year: 2013
Downloads: 37
Quote: 0
Read: Download Dissertation

Abstract


Product reviews mining is to use natural language processing techno-logy from network in the comment text extract the user product evaluation information, which help users to purchase products or services from the decision, provide valuable information to business or producers to improve products and improve service quality for decision-making. Product evaluation targets and opinion word collocation extraction is a very important research topic in the mining of product reviews, is a very important part of the basis of sentiment analysis tasks, but also emotional information extraction. This paper makes use of the natural language processing, machine learning, statistical machine translation technology, in view of the network Chinese review information, extract the evaluation target and evaluation word collocation relations. We carried out the study and comparison of the two methods, according to the special features of the evaluation collocation relationship in the Chinese text and based on basis of previous studies.1. This article will evaluate collocation extraction problem is transformed into a sequence labeled problem, constructing the word itself, part of speech, dependence relationship, word location and other characteristics, based on conditional random field model extract comment text opinion target and emotional word with relationships. This method is suitable for comparison rules, length short text sentences, and the need to manually label part of the training corpus. 2. The other way is make opinion target and emotional words with relationship problems transformed into word alignment problem, the use of machine learning methods based on the statistics of the source and channel corresponding relationship between the words and the words extracted from the parallel corpus, alignment probability is greater than the threshold value in the two words is considered to be a candidate evaluation collocation. After get the candidate evaluation collocation relationship, using the principle of random walk after several times of iteration calculation for each candidate words convergence probability, set up separately evaluation words and the evaluated object’s threshold, filtering the words which value is less than the threshold value. The rest of the words between candidate evaluation collocations, that they are the final evaluation with the relationship. This method is suitable for large-scale evaluation of corpus, without manual label the corpus.Finally, in this paper, using two different corpora respectively on the two methods of experiment, the experimental results show that the proposed two methods in the evaluation of product collocation extraction is feasible, and obtain better accuracy rate and recall rate. Research of these methods to other tasks of sentiment analysis has an important reference value.

Related Dissertations

  1. Research on Word Alignment Based on Statistics and Linguistics and Correlation Fusion Strategy,TP391.2
  2. The Research on English-Chinese Name Entity Translation,TP391.2
  3. Based on the random walk model personalized information recommendation,TP391.3
  4. Graph model based on statistical and machine learning algorithm and its application of certain,TP181
  5. Some Topics on Two-parameter Random Walk,O211.6
  6. A Special Model of Random Walk on the Plane,O242.1
  7. Comparative Study of Noise Trading in Sino-American Stock Markets,F832.51;F831.51
  8. Study on Word Alignment Technology and Construction of Statistical Machine Translation Platform,TP391.2
  9. Research on P2P Trust Model and Search Algorithm Based on the Incentive Mechanism,TP393.02
  10. A Two-stage Collaborative Filtering Algorithm Based on Random Walking and Cluster-based Smoothing,TP391.3
  11. Research on Machine Translation System Combination Based on Confusion Network,TP391.2
  12. Research of Image Annotation and Tag Recommendation on Shared Resource Websites,TP393.092
  13. Research on Chinese-uyghur Word-alignment for Statistical Machine Translation,TP391.2
  14. The Research on Hownet Based Word Alignment Technology,TP391.2
  15. Based on feature fusion method of single- word aligned,TP391.2
  16. Identification and Extraction of Phrasal Paraphrase,TP391.1
  17. Based on analysis and auxiliary music composer complex networks,J619
  18. Rule-based and statistical - based combination the bilingual parallel to sentence the phrase alignment method,TP391.2
  19. Research on Collaborative Filtering by the Integration of Social Tags,TP301.6
  20. Multi-word Expression Extraction Based on Chinese-English Bilingual Corpus,TP391.1
  21. Interactive Aided Translation Technology Research Based on Internet,TP391.2

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Text Processing
© 2012 www.DissertationTopic.Net  Mobile