0
$\begingroup$

I'm currently doing a course project for a course at university and I'm tasked on doing multilabel user profile classification based on their posts in social media, for each user there's a set of posts on social media which combined make up the user's occupation, gender etc.

In the past I've done simpler classification, where each post would have its own label and you would have to predict only the post's class, whereas in my current task I would have to somehow take into account all the posts of a user, since I only have the labels for the user, and not for each individual post.

My first idea would be to concatenate all the posts for each user and create a tf-idf vector which I would use for a classifier, but that way I would lose information from each individual tweet.

My question would simple concatenation suit my needs for this task, or are there more elaborate techniques which would take into account each individual tweet's importance?

$\endgroup$
0
$\begingroup$

My first idea would be to concatenate all the posts for each user and create a tf-idf vector which I would use for a classifier, but that way I would lose information from each individual tweet.

This is the way I would do it too, I think there is no way to preserve the individual information of tweets. But it's more valuable to have the summarised information of tweets for an individual person.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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