A Dictionary-Based Approach to Identifying Aspects Implied by Adjectives for Opinion Mining

ABSTRACT One of the central problems of opinion mining is to extract aspects of entities or topics that have been evaluated in an opinion sentence or document. Much of the existing research focused on extracting explicit aspects which are nouns and nouns phrases that have ap‐ peared in sentences, e.g., price in The price of this bike is very high. (owever, in many cas‐ es, people do not explicitly mention an aspect in a sentence, but the aspect is implied, e.g., This bike is expensive, where expensive indicates the price aspect of the bike. Although there are some existing works dealing with the problem, they all used the corpus‐based approach, which has several shortcomings. )n this paper, we propose a dictionary‐based approach to address these shortcomings. We formulate the problem as collective classifica‐ tion. Experimental results show that the proposed approach is effective and produces sig‐ nificantly better results than strong baselines based on traditional supervised classification.