Proceedings of the International Conference on
Software, Knowledge, Information Management and Applications (SKIMA 2013)
Advanced Technology Solutions and Applications in Higher Education and Enterprises |
(from Chapter 5: Optimisation, Planning and Decision Support)
| Full Citation and Abstract
|
Title: |
Product reputation evaluation: the impact of conjunction on sentiment analysis |
| |
Author(s): |
Umar Farooq, Antoine Nongaillard, Yacine Ouzrout, Muhammad Abdul Qadir |
| |
Address: |
DISP Laboratory, University Lumiere Lyon 2, 160 Bd de l’Université 69676 Bron Cedex, France
DISP Laboratory, University Lumiere Lyon 2, 160 Bd de l’Université 69676 Bron Cedex, France
DISP Laboratory, University Lumiere Lyon 2, 160 Bd de l’Université 69676 Bron Cedex, France
Department of Computer Science, Mohammad Ali Jinnah University, Islamabad Expressway, Kahuta Road, Zone-V, Islamabad, Pakistan
farooq.umar @ univ-lyon2.fr, nongaillard.antoine @ univ-lyon2.fr, ouzrout.yacine @ univ-lyon2.fr, aqadir @ jinnah.edu.pk |
| |
Reference: |
SKIMA 2013 Proceedings pp. 242 - 250 |
| |
Abstract/
Summary |
The two main issues in sentiment analysis are 'word sense disambiguation' and 'conjunction analysis'. The polarity identification of sentences is a challengeable task, especially when one or more conjunctions appear. In the presence of conjunction, it is required to identify which part of the sentence contributes to the overall sentiment expressed. This work analyzes the role of conjunction and investigates the problem of determining the polarity of compound sentences, i.e., when conjunction appears in a sentence. A classification of the most frequent conjunctions is introduced and rules for each class are proposed to determine sentiment at both sentence level and feature level. The method we propose in this paper is also able to handle negation. Two sets of experiments are performed to test the effectiveness. The accuracy of this method at sentence level and feature level is better than the accuracy of existing methods. |
| |
|
| |
We welcome your comments about this Article  |
|
