Lexical Domain Adaptation

Many sentiment-bearing expressions are ambiguous in that they can exhibit positive, neutral, or negative polarity depending on the context, domain, genre, or topic in question.

One easy way to disambiguate sentiment is to fine-tune specific sentiment carriers in the underlying sentiment lexicon.

Using sentiment lexicon entries, it is possible to refine the sentiment model to be better tuned to your domain(s), genre(s), and key topic(s). PreCeive REST API documentation provides details about how to create such entries. However, what do they do?

Consider the following sentence with an ambiguous sentiment carrier (“beast“):

He was regarded as a political beast.

By default, this sentence would be attributed with negative polarity as beast is negative in most contexts, domains, and genres.
It is possible to create an entry in the sentiment lexicon for the ‘beast‘ lexeme and assign positive polarity to it. In this way, we change change the sentiment score of the above sentence to be positive.

To achieve this, you can POST the following JSON object to the sentiment lexicon endpoint at /resources/lexicons/sentiment/lexicon after which the customised lexicon entry exists on your account:

{
    "text":     "beast",
    "polarity": "pos",
    "reverse":  "equ"
}
Simon Guest

Simon Guest