What are the most difficult issues in mood analysis (intellectual examination)?

Question

Opinion Mining / Sentiment Analysis is a somewhat recent sub-task of natural language processing. Some comparison with the classification of the text, some take a deeper position in relation to it. What do you think about the most difficult problems in the analysis of moods (intellectual development)? Can you name a few?

Answer 1

Key issues for mood analysis are: -

1) Named entity recognition - as a matter of fact, what a person is talking about, for example. 300 Spartans - a group of Greeks or a film?

2) Anapora Resolution - The problem of resolving what a pronoun or phrase means. "We watched a movie and went to dinner, it was awful." What does it mean"?

3) Analysis - What is the subject and object of the sentence to which the verb and / or adjective really refers?

4) Sarcasm. If you do not know the author, you have no idea whether this is bad or bad.

5) Twitter - abbreviations, lack of capital, poor spelling, poor punctuation, poor grammar, ...

Answer 2

I agree with Hightechrider that areas where precision sentiment analysis can see improvement. I would also like to add that mood analysis is usually done over a closed-area text in most cases. Attempts to do this in plaintext domain usually end with very poor accuracy indicators / F1 / that you have, or it is a pseudo-open domain, because it looks only at certain grammatical constructions. Therefore, I would say that a sensory analysis of feelings, which can determine the context and make decisions based on this, is an exciting area for research (and industry products).

I would also expand my fifth paragraph from Twitter to other social networking sites (e.g. Facebook, Youtube), where short, non-grammatical statements are commonplace.

Answer 3

I think the answer is language complexity, grammar errors and spelling. There are many ways to express people's opinions, for example, sarcasm can be mistakenly interpreted as extremely positive feelings.

Answer 4

The question may be too general, since there are several types of sentiment analysis (document level, sentence level, comparative sentiment analysis, etc.), and each type has some specific problems.

Generally speaking, I agree with @Ian Mercer's answer, and I will add 3 more questions:

• How to identify a deeper feeling / emotion. Positive and negative is a very simple analysis. One of the problems is how to extract emotions, for example, how much hatred is inside an opinion, how much happiness, how much sadness, etc.
• How to determine an object for which a positive opinion, and an object for which a negative opinion. For example, if you say: “She won it!”, This means for her a positive mood and a negative mood for him at the same time.
• How to analyze very subjective sentences or paragraphs. Sometimes, even for people, it is very difficult to accept the feeling of these high subjective texts. Imagine a computer ...