Mining comparative opinions from customer reviews for Competitive Intelligence

Xu, K., Liao, S. S., Li, J., & Song, Y. (2011). Mining comparative opinions from customer reviews for Competitive Intelligence. Decision Support Systems, 50(4), 743—754.

Key Concepts

1. companies apply CI methods that
• monitor the reception of their own products and services
• analyze information on their competitors activities
2. comparative opinions are valuable information sources for identifying the relative weaknesses and strengths of products.
3. the authors distinguish between rule-based relation extraction methods with automatically and manually defined rules. The first group is further separated into (i) feature-based methods, that rely on features such as POS tags and entity types for relation extraction, and (ii) kernel-based methods that represent examples and define kernels to compute similarities in a high-dimensional space.
4. comparative relation extraction obtains directed relation with the following four arguments: $$Relation Type and Direction(Product_1, Product_1, Attribute, Sentimental Phrase)$$
5. Examples relations:
• >(Nokia N95, iPhone, camera, better)
• ~(Pearl, Curve, camera, high resolution)
• >(iPhone, Curve, screen, bigger

Method

The authors use conditional random fields to model (a) the relation between relations and entities, and (b) between relations and words.

Linguistic Features

1. Capitalization information (identify product names), Word type, POS tags
2. Prefixes and suffixes: recognize sentiment phrases (bett-er, quick-er, fast-est, ...)
3. Indicator phrases for comparisons (in contrast to, unlike, compare with, ...)
4. Syntactic paths (identify comparisons)
5. Grammatical roles (products are often subjects, attributes entities are objects, ...)