Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397
Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397
Issue N11 2015 year
The article is devoted to the problem of automatic sentiment analysis also known as opinion mining. Aspect-based version of such analysis is given in detail; its main subtasks are listed with examples. The primary goal of the article is to show a way for opinion mining system implementation.
The first part of the article gives overall structure of a system for aspect-based sentiment analysis. The whole system can be broken into four key subsystems namely are preprocessing, subsystems of expert knowledge, linguistic resources and main control subsystem. Depending on the method being implemented the number of subsystems can vary slightly. For example, in this article there is additional word representation subsystem. It implements the method which uses semantic similarity to address key subtasks of aspect-based sentiment analysis.
The second part of the article shows how such system can be realized in practice. The system is implemented according to the object-oriented paradigm. Main classes and their relations are described with UML class diagram. Also the example of a user interface and a way of storing the results are proposed.
In the third part of the article the experimental results based on the text corpora of restaurant and car reviews, provided by seminar SentiRuEval-2015' organizers, are presented. This results show the superiority of the proposed system over similar ones.
The developed system can be used as a standalone application for document analysis. It also can be integrated within search engine or recommendation service to enrich their results with aspect-specific sentiment knowledge.