1. Introduction
Recently, with the explosive of social media, there is a high demand from brands and industries to analyse the customers’ comments on their products automatically so that they can know how consumers perceive their products as well as those of their competitors. This sentiment information is not only useful for marketing and product benchmarking but also useful for product design and product development [Liu, 2008].
Extracting opinion or sentiment from text can be defined as sentiment analysis or opinion mining. The task receives raw texts talking about brands or products as input, and outputs sentiment information, which presents the author’s opinions, comments, and evaluation about a specific brand, product, or about products’ features/attributes. Generally, we need to determine whether the sentiment is positive, negative or neutral.
While research in this field for English has been being grown, there is a little work for Vietnamese. This has motivated us to build a sentiment analysis system for Vietnamese texts.
2. Research
We are now focusing on two levels of sentiment analysis: (1) sentence level, and (2) entity and aspect level. The task at sentence level goes to sentences and determines each sentence expressed a positive, negative or neutral opinion. However, the sentence-level analysis does not discover exactly what people liked or did not like, which is a motivation for the entity-level task. The task directly looks at the sentiment of an opinion target–a specific entity or aspect, e.g., the call quality and the battery life of an iPhone. End users, such as a marketing analyzer of a brand, can use such results for all kinds of qualitative or quantitative analyses.
References:
[1] Bing Liu, Sentiment Analysis and Opinion Mining. 2012