CODEWITHSUNDEEP
text-miningsentiment-analysis
KS1
KS1

Reputation: 165

Differentiating sentiments for subjects and predicates in text mining

I'm trying to search for literature on text mining specifically for determining the sentiment with respect to the subject and object in a sentence. For example, I could have a sentence "Alice scored much better than Bob", or "Alice's gain is Bob's loss", or "Alice's problems are an opportunity for Bob".

In each of these, the sentence has different sentiments for Alice and Bob.

My question is: what is the general term I should use for searching for literature on determining sentiments with respects to the subject or object? Is there a good reference for this?

Question 2: Many such sentences have the use of apostrophes. Is there a standard one-pass technique to change sentences from "Alice's good work is appreciated" to "Good work of Alice is appreciated"? Generally, the tokenizer and lemmatizer phases seem to remove apostrophes. I could see a two-pass technique that does POS,determines the relation between "Alice" and "good work" and perform the transformation, and then again does POS. I was wondering if there's an easier way.

Upvotes: 0

Views: 162

Answers (1)

vpekar
vpekar

Reputation: 3355

The general term used in the context of sentiment analysis is "comparative sentences". Here is a blog post looking at how different sentiment analysis APIs handle comparative sentences: https://blog.bitext.com/comparing-apis-example-2

Re. Question 2. You are better off not changing the original sentence at all. You can use a dependency parser, such as the Stanford parser. It will give you the syntactic relations between words based on which you will be able to recognize the similarity between "X's Y" and "Y of X", and to find out which words fill out the pattern "X is any-comparative-adjective than Y".

Upvotes: 1

Related Questions