Understanding and analyzing opinions
At a time when the Internet becomes the main means to communicate and express themselves (via blogs, forums, etc.), the concept of Sentiment Analysis or Opinion Mining tends to play an increasingly central role.
For most players in the economic market today it has become essential to understand the wishes of consumers, their tastes, desires, opinions, feelings, expectations and needs in order to satisfy them and remain competitive.
Although Sentiment Analysis is not an asset in itself, when integrated with a global marketing and technical policy it becomes a significant competitive advantage as it fosters a meaningful analysis of consumer expectations and so the definition of a targeted strategy.
The Sentiment Analysis tools developed by Ho2S intend to be as close as possible to the consumer. This is achieved by researching, collecting and analyzing all the information to decipher the nuances and the relevance, to order them, classify them and quantify them so as to:
- Understand the implications
- Expose the strengths and weaknesses, areas for improvement, areas that need to be created or modified
- Adapt the strategy according to these requirements and therefore:
- Ensure satisfaction and attract new consumers,
- Ensure the customer loyalty.
The automatic classification of an opinion text is an important task for Ho2S. Three different methods to perform this text classification have been developed. The first method is symbolic, the second one is statistical and the last one is hybrid (a combination of the first two).
The combination can reap the benefits of the two methods, namely the robustness of statistical machine learning and the symbolic manual configuration oriented with respect to the use of real applications. The Opinion Mining is at the level of sentences, then the entire text is classified according to its polarity as “positive” or “negative.”
The objective of Ho2S is to implement a graphical user interface to help non-experts to browse among the opinions, trends and documents they could find on the Internet. This is therefore a component that is capable of converting unstructured information into fully structured information. In practice this is a logical structure that allows the user to interact inside threads in newsgroups and forums (with questions, answers and citations).
The application area is typically multifaceted:
- Named entities (product names, locations, etc.)