Extracting Product Features and Opinions from Reviews - Natural Language Processing and Text Mining - page 9

This chapter introduces opine, an unsupervised information extraction
system that embodies a solution to each of the above subtasks. Given a partic-
ular product and a corresponding set of reviews, opine outputs a set of product
features , accompanied by a list of associated opinions , which are ranked based
on strength.
Our contributions are as follows:
1. We describe opine's novel use of a relaxation labeling method to find
the semantic orientation of words in the context of given product features and
sentences.
2. We compare opine with the review mining system of Hu and Liu [2]
and find that opine's precision on the feature extraction task is 22% higher
than that of Hu and Liu, although its recall is 3% lower. We show that 1/3
of opine's increase in precision comes from the use of its feature assessment
mechanism on review data while the rest is due to Web statistics.
3. While many other systems have used extracted opinion phrases in order
to determine the polarity of sentences or documents, opine reports its preci-
sion and recall on the tasks of opinion phrase extraction and opinion phrase
polarity extraction in the context of known product features and sentences.
On the first task, opine has a precision of 79% and a recall of 76%. On the
second task, opine has a precision of 86% and a recall of 89%.
4. Finally, opine ranks the opinion phrases corresponding to a particular
property based on their strength and obtains an accuracy of 73%.
The remainder of this chapter is organized as follows: Section 2.2 intro-
duces the basic terminology; Section 2.3 gives an overview of opine,and
describes and evaluates its main components; Section 2.4 describes related
work; and Section 2.5 describes our conclusions and future work.
2.2 Terminology
A product class ( e.g. , Scanner) is a set of products ( e.g. , Epson1200). opine
extracts the following types of product features : properties , parts , features of
product parts , related concepts , parts and properties of related concepts (see
Table 2.1 in subsection 2.3.2 for examples in the Scanner domain). Related
concepts are concepts relevant to the customers' experience with the main
product ( e.g. , the company that manufactures a scanner). The relationships
between the main product and related concepts are typically expressed as
verbs ( e.g. , “the company manufactures scanners”) or prepositions (“scanners
from Epson”). Features can be explicit (“good scan quality ”) or implicit
(“good scans” implies good ScanQuality ).
opine also extracts opinion phrases , which are adjective, noun, verb or
adverb phrases representing customer opinions. Opinions can be positive or
negative and vary in strength ( e.g. , “fantastic” is stronger than “good”).
2.3 opine Overview
This section gives an overview of opine (see Figure 2.1) and describes its
components and their experimental evaluation.