Sentiment Analysis
Idilia’s WSD technology can dramatically improve sentiment analysis and content filtering. WSD enables more accurate filtering of documents based on sentiment (positive/negative) or content (e.g. adult content, offensive content), because documents annotated with word senses contain much richer, more reliable information about sentiment and content. Sentiment analysis can be used to monitor real-time opinions for brand management, public relations and market research. Specifically,
- Words which have multiple meanings, only some of which connote offensive or adult content, are assigned their correct meaning based on the context. Similarly, words which can, but do not always carry pejorative meaning, are disambiguated accurately by Idilia’s WSD.
- Idilia’s massive knowledge base, the Language Graph, contains a wide variety of semantic information about each word meaning, including, for example:
- word meanings which are laudatory versus pejorative or very pejorative,
- word meanings that connote offensive or adult content.
- word meaning which are considered slang, vulgar, regional usages, or technical terminology
- Combing WSD results with information in Idilia’s semantic knowledge base allows applications to build a “map” of a document based on the type of terminology being used. This can be used directly to filter documents or fed to a classifier, in either case generating much more precise results than ordinary keywords
