C++ Sentiment Analysis Library [closed]

Question

I'm looking for a C++ sentiment analysis library that I could use in my own application. Something that would take a text written by a human as an argument and return information on its mood (positive, negative, neutral, angry, happy, ...). Any ideas?

A few remarks:

• I'm not looking for a library with "just" NLP tools (as text tokenization, PoS tagging etc.), but really something that does sentiment analysis / opinion mining / mood analysis. Of course an NLP library with sentiment analysis tool is great.
• Something very simple would be ok (e.g. just returning +1/-1/0)
• I don't care which underlying technique it may use (dictionaries, bayesian stuffs, SVMs, rule-based...)
• C++ only!

I know it probably does not exist just like that, but hey.

Tell

Answer 1

I'd suggest writing your own. You can capture 80% of mood correctly with a simple filter for positive and negative sentiment words, and you can scale your sign based on some common modal words and punctuation. If you want to squeeze juice out of that remaining 20% of ambiguity and you don't want to train your own SVM or whatever, I'd suggest finding papers at op.to/goog/scholar+ some of those papers will have authors who have written C++ libs and talking about where you can find / download them.

Also see this SO post among many others for more ideas.

Backpedal

A shame none of the links work. – Tangential

Answer 2

I'd avoid using lists of words that you try to predict - it simply can't account for the niche and varying language found in real human texts (assuming they're social stuff). I can't help you with a C++ library, but if you're happy calling a web service you can use ours for free Sentiment Analysis of social media.

Peaslee

Answer 3

You can also take a look at Bitext's API for sentiment analysis: https://www.bitext.com/text-analysis-api-and-faq/

It currently supports English, Spanish, Portuguese, French, Italian, German, Dutch and Catalan.

There is no need to do any training of data, since Bitext's language technologies are based on language grammar and dictionaries, not machine-learning or statistics.

Selflove