Dissertation > Excellent graduate degree dissertation topics show

Research on Product Reviews Analysis Based on Appraisal Expression

Author: HuangYiHua
Tutor: YuanChunFeng
School: Nanjing University
Course: Applied Computer Technology
Keywords: Sentiment Analysis Review Emotional evaluation unit Syntactic tree Convolution tree kernel Noumenon
CLC: TP391.1
Type: Master's thesis
Year: 2011
Downloads: 125
Quote: 0
Read: Download Dissertation

Abstract


With the vigorous development of Web2.0, advocated by the Internet users participate in the site content manufacturing \makes the explosive growth of Internet information. Information, so text they contain a large number of reviews, and important information for users to buy merchandise before as the essential basis for decisions, but also businessmen user feedback analysis. A large number of such comments as text, content organization is not standardized, more spam. In order to help people quickly and efficiently analyze these text sentiment analysis technology came into being. The evaluation object extraction Lee polar analytes are two core tasks of sentiment analysis. These two basic tasks as object and product reviews text analysis tasks are divided into emotional evaluation unit extraction and polarity of two parts, as the characteristics of the various components of the commodities and various properties, analysis positive evaluation and the number of negative evaluation and statistics for each commodity characteristics, and finally presented to the user. The work of this paper include the following three aspects: (1) an affective evaluation unit based on the GMCT the syntactic structure indicates mode and automatic mode library construction method. The emotional evaluation unit is the basic unit of product reviews, which includes the object being evaluated and the evaluation of the word. Most of the emotional evaluation unit extraction method of pattern recognition, the flat syntactic features and modes, mode libraries built using manual methods. This paper proposes a new structure, it can retain the syntactic tree based GMCT mode structural information, and thus better able to distinguish the noise pattern, to obtain a better accuracy rate. At the same time, this paper presents an automatic build mode library method to avoid the time-consuming manual build process. (2) an approximate pattern matching method based on convolution tree kernel, and on this basis, given the emotional evaluation unit extraction algorithm. Exact match for the more complex tree structure matching more difficult to match, exact match used herein pattern matching recall rate is not high. Article will convolution tree nuclear methods used to calculate the degree of similarity of the tree, to approximate pattern matching, thereby improving the recall rate. Furthermore, this article convolution tree kernel method modified approximate convolution tree kernel method, greatly improving slightly sacrificing accuracy Zhao same rate. (3) Product Features body and its construction methods and building evaluation word Taxonomy. Finally count Fasiluli method of analysis based on the body and evaluation word Taxonomy polarity. The evaluation Dictionary is the basis of the analysis of the polarity. The existing evaluation Dictionary lists only the evaluation of the polarity of the word, without taking into account the evaluation words with different evaluation objects with polarity change. The automatically extracted disorganized and not easy summary display of the results of the analysis of product reviews. In this paper, the characteristics of the hierarchical characteristic body, and evaluation of the word classification, evaluation word Taxonomy., And the corresponding evaluation object and evaluation of the different permutations of the word, not only a good deal with the polarity changes situation, but also be able to use a representative and characterized by high degree of concern to show the polarity of analysis results. In this paper, the proposed algorithm through experimental verification, the experimental results show that the proposed algorithm is effective.

Related Dissertations

  1. Analysis on Registration System of Forest Right,D923.2
  2. Experiment and Research on Mathematics Layering Review for Junior High School Graduation Exam,G633.6
  3. High School Physics Scenario-based Practice Research Unit Refresher Course,G633.7
  4. Study on Company Registration Review,D922.291.91
  5. Applied Research of Ontology in Intelligent Residence Community,TP391.1
  6. Fuzzy stability coexist with random goods inventory policy research,F224
  7. A Study of the Research from Home and Abroad on "Intercultural Competence" and "Intercultural Communication Competence" (2001-2010),H319
  8. Data mining technology in the Expo traffic with social media Prediction Research and Application,TP311.13
  9. Endovascular Repair of Ruptured Abdominal Aortic Aneurysms: A Systematic Review and Meta-analysis,R654.3
  10. Research on Information Retrieval Technology Based on Semantic Web,TP391.3
  11. Compound Traditional Chinese Medicine with Qi_supplementing and Blood Activating for Chronic Heart Failure on Systematic Reviews of Randomised Controlled Trials,R259
  12. The Systematic Review of Shuxuetong Injection in Treating Acute Cerebral Infarction,R277.7
  13. Foreign patents to speed up the review mechanism and its reference to China,G306
  14. Review of History Paper of College Entrance Examination Reform of Three Decades,G633.51
  15. Improvement of Science & Technology Award Legal System,D922.17
  16. Dominance of demolition,D922.181
  17. On the business judgment rule applies,D922.291.91
  18. Construction of the System of Criminal seizure,D925.2
  19. The Study on Fu Lei’s Musical Thought,J609.2
  20. Northwest 20th century arts and crafts of ancient Chinese and foreign exchange research review,J509.2
  21. Brief Review on BarOn Emotional Intelligence Scale ( adolescent version ) Amendments and trial,B841.7

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