Conference Paper

A feature dependent method for opinion mining and classification

DLSI, Univ. Alicante, Alicante
DOI: 10.1109/NLPKE.2008.4906796 Conference: Natural Language Processing and Knowledge Engineering, 2008. NLP-KE '08. International Conference on
Source: IEEE Xplore


Mining the web for customer opinion on different products is both a useful, as well as challenging task. Previous approaches to customer review classification included document level, sentence and clause level sentiment analysis and feature based opinion summarization. In this paper, we present a feature driven opinion summarization method, where the term ldquodrivenrdquo is employed to describe the concept-to-detail (product class to product-specific characteristics) approach we took. For each product class we first automatically extract general features (characteristics describing any product, such as price, size, design), for each product we then extract specific features (as picture resolution in the case of a digital camera) and feature attributes (adjectives grading the characteristics, as for example high or low for price, small or big for size and modern or faddy for design). Further on, we assign a polarity (positive or negative) to each of the feature attributes using a previously annotated corpus and Support Vector Machines Sequential Minimal Optimization machine learning with the Normalized Google Distance. We show how the method presented is employed to build a feature-driven opinion summarization system that is presently working in English and Spanish. In order to detect the product category, we use a modified system for person names classification. The raw review text is split into sentences and depending on the product class detected, only the phrases containing the specific product features are selected for further processing. The phrases extracted undergo a process of anaphora resolution, Named Entity Recognition and syntactic parsing. Applying syntactic dependency and part of speech patterns, we extract pairs containing the feature and the polarity of the feature attribute the customer associates to the feature in the review. Eventually, we statistically summarize the polarity of the opinions different customers expressed about the product on the -
web as percentages of positive and negative opinions about each of the product features. We show the results and improvements over baseline, together with a discussion on the strong and weak points of the method and the directions for future work.

40 Reads
  • Source
    • "Some researchers study opinion analysis related to specific domains, this need to collect a features of that domain (Balahur & Montoyo 2008) focus on where the term " driven " is used to describe special product class from the other product classes in general. Then determine the polarity (positive or negative) to each of the feature attributes using annotated corpus. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Social networks and users’ interactions are distinct features for the current Web. They constitute a fundamental part of Web 2.0, where people produce, disseminate, and consume information in new interactive forms where users are not only passive information receivers. Social media succeed to attract a large portion of online users, which explains the explosive growth of social media in terms of comments, reviews, blogs, microblogs, Twitters, and postings in social network sites. In this scope, sentiment analysis research field refers to the analysis of people’s sentiments, opinions, attitudes, and emotions towards events, products, companies, individuals, issues, sport teams ...etc. Facebook, and YouTube are within the top 3 sites used in many Middle Eastern (ME) countries, and the world. Therefore a huge volume of Arabic comments and reviews are generated daily about different aspects of life in this part of the world. Modern Standard Arabic (MSA) is used mainly in media (Newspapers, Journals, TV and Radio), academic institution, and to some extent in social media. While colloquial Arabic is used by the public in their conversations, chatting, etc.. Analysis of social networks in ME countries shows that both MSA and colloquial or slang languages are used. The aim of this study is to build a novel sentiment analysis tool called colloquial Non-Standard Arabic - Modern Standard Arabic-Sentiment Analysis Tool (CNSAMSA-SAT) dedicated to both colloquial Arabic and MSA. A large number of Arabic collected comments and reviews from social media were tokenized and analyzed to build polarity lexicons which constitute an essential part of CNSA-MSA-SAT. Each Arabic collected comment and review is manually assigned to one of the three polarity values: (positive, negative, and neutral). Further, each collected review or comment is added to CNSA-MSA-SAT and is then assigned to one of the three polarities values based on algorithms developed for this purpose.
    Full-text · Conference Paper · Apr 2013
  • Source
    • "Opinion mining discovers opinioned knowledge at different levels such as at clause, feature, sentence or document levels [13]. In the previous section we discussed a way to classify student opinion at document level. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The purpose of this paper is to show how opinion mining may offer an alternative way to improve course evaluation using students’ attitudes posted on Internet forums, discussion groups and/or blogs, which are collectively called user-generated content. We propose a model to mine knowledge from students’ opinions to improve teaching effectiveness in academic institutes. Opinion mining is used to evaluate course quality in two steps: opinion classification and opinion extraction. In opinion classification, machine learning methods have been applied to classify an opinion as positive or negative for each student’s posts. Then, we used opinion extraction to extract features, such as teacher, exams and resources, from the user-generated content for a specific course. Then we grouped and assigned orientations for each feature.
    Full-text · Conference Paper · Jan 2011
  • Source
    • "Opinion mining discovers opinioned knowledge at different levels such as at clause, feature, sentence or document levels [8]. This paper discusses how to extract opinions in feature level. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In academic institutions, student comments about courses can be considered as a significant informative resource to improve teaching effectiveness. This paper proposes a model that extracts knowledge from students' opinions to improve and to measure the performance of courses. Our task is to use user-generated contents of students to study the performance of a certain course and to compare the performance of some courses with each others. To do that, we propose a model that consists of two main components: Feature extraction to extract features, such as teacher, exams and resources, from the user-generated content for a specific course. And classifier to give a sentiment to each feature. Then we group and visualize the features of the courses graphically. In this way, we can also compare the performance of one or more courses.
    Full-text · Article · Jan 2011
Show more