Sentiment Parts

OpenDover has several "search sentiment" options. Three main constituents of the sentiment/opinion are taken into consideration:

• sentiment/opinion word
• domain word
• object

The search can be performed by using a single constituent of the sentiment, a combination of constituents or the whole set of them.

Let's consider an example.

“Nevertheless, the GadgetY's lens is perfect. The colours and exposure are very good too.”

A sentiment word is a word or a short phrase which describes appreciation or judgment. “Good”, “bad”, “nice” are simple examples of a sentiment word. “Perfect” and “very good” are sentiment words in the example above.

A domain word is a word or a short phrase which describes a certain subject domain. “Lens” and “exposure” in our example are domain words describing the subject domain “Product – Camera”.

An object is just a word or a short phrase which OpenDover looks for in a sentence to compose a sentiment. “GadgetY's” could be an object in our example.

Different sentiment search methods require different sentiment parts to be present in a sentence. The table below summarizes the requirements.

searchSentiment and searchObjectSentiment look for an ontology-based collocation (i.e. a sentiment word with its accompanying words describing a particular subject domain). As a result, domain-specific sentiment words are extracted.

searchBareSentiment and searchBareObjectSentiment do not use the ontology. These methods output “emotional background” of an input text.