Hello everyone,
On Twitter, I'm well aware that topic modeling and TF*IDF methods can be used to extract a user's interests from tweets he/she posted or consumed. However, what is the best and easiest way to do that? Are there other efficient methods?
Thanks.
Thank you, Mr. Mataoui. Your answer is clear.
Should we consider N-grams in Frequency-based approaches?
Sami, there is no issue with applying both methods (N-gram and tf-idf) in one process.
HTH.
Samer
You may take a look at this work that may help you to deal with topic extraction :
http://mediamining.univ-lyon2.fr/people/guille/tom.php
Depends what is definition of the Topic, if it's pre defined set of topics catered for your business and product needs then you might want to check our work.
[1] & [2] has comprehensive catalog of features and their performance in respect of identifying features for task of interest and expertise detection. Both of them relie that you have engine that from short from text returns topics. This can be accomplished by relying on the system mentioned by @Nicolas Dugué above, or using entity extraction linking system (where entities you care about are basically topic... or they may map to set of topics).
[1] Mining Half a Billion Topical Experts Across Multiple Social Networks
https://arxiv.org/pdf/1608.09002.pdf
[2] LASTA: large scale topic assignment on multiple social networks
http://dl.acm.org/ft_gateway.cfm?id=2623350&ftid=1495182&dwn=1&CFID=936924916&CFTOKEN=35331393
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size