sTarget is another sentiment analysis system. How is it different from other such systems? The main difference is that this system requires sentiment target to define sentiment. Many sentiment analysis systems use bag-of-words approach and simply count the number of positive and negative words in a text. But words do not exist without a context. Even positive ones like "perfect" may be used in a negative sentence, e.g.: It's a perfect example of how it shouldn't be done.
But context is very difficult to capture. We decided to reduce the context to a sentiment target, a word (or a phrase) that is evaluated in a context. For example, a sentence "I love tomatoes but hate cucumbers" is positive toward tomatoes and negative towards cucumbers. The "overall" sentiment is useless here. But if we specify "tomatoes" as a target, we'll get a correct sentiment - positive. And negative will be a correct sentiment for cucumbers.
Please, do not forget to try different classifiers changing parameter Classifier to either 'p' ("precision") for a more cautions approach to classification (reports only clear and reliably classified sentiment) or to 'r' ("recall") to find more sentiment (less accurately though).
Feel free to share your ideas and observations with us))
The version presented here is a limited version and is not as accurate as the full version being developed. If you need more detail please contact the author.
- No information
Sentiment analysis type: r - to maximize recall; p - to maximize precision (default)
Specify a word or phrase (must be present in a text) that is commented about. Sentiment target is usually a noun or a pronoun.
the text that needs to be processed
Example: context=I love the car.