System and methods for rule-based sentiment analysis

摘要

Rule-based system and methods for determining sentiment values for a text unit comprising multiple terms. The sentiment type of an input text unit is unknown when received, but can be derived based on various attributes of the component words or phrases in the text unit. The component words or phrases in the text unit may or may not have a known sentiment type attached. When the component words has a sentiment value, the present system provides methods to determine whether the derived sentiment type of the text unit as a whole should be the same or different or opposite to the sentiment type of the component words. When the component words do not have a known sentiment type, the present system can still produce correct sentiment type for the text unit based on other attributes associated with the component words. For example, it will determine that the text unit of “high price” can have a negative connotation, and “high quality” can have a positive connotation, even though the words “high”, “price”, “quality” are not marked with a sentiment type. The methods provide a solution to the challenge of correctly or accurately determining the sentiment type of various linguistic structures under different context.