4
$\begingroup$

I want to detect tweet text agreement. Suppose someone posts some subjective opinion in twitter. Other users will post reply either agreeing or opposing the original tweet. I want to estimate the amount of agreement. Is there any algorithm/library in any language to do that or any labeled dataset?

$\endgroup$

2 Answers 2

1
$\begingroup$

Not sure there is anything for that, you could check

  • is it a verbatim retweet
  • does it have the same sentiment
  • is the edit distance low

Or you can train your own model, where you label (agree) by hand and then build features.

$\endgroup$
1
$\begingroup$

Sure there can be more complicated approach but if you are dealing with raw tweets, I think problem is twofolds;

  1. Topic Discovery

You first need to find out what the tweet is talking about. It will be much easier task if you can skip this, given that you are looking at tweets with specific tags; that you know what the tweet is pertaining to. Otherwise, you can use LDA or gensim library in Python.

  1. Sentiment Analysis

This is much easier task. For each topic, for all tweets, associate tweets with probability of positive / negative and also, you could scale this by certainty. This could be using out of box solution such as from nltk.

This github repo seems to be doing what you are intending to do, and could get some inspiration.

https://github.com/nagarmayank/twitter_sentiment_analysis

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.