search

adjidieng / DETM

MIT License
131 stars 39 forks source link

Does DETM work on short texts like tweets and does it work on corpus in other languages？ #4

Open joeylige opened 4 years ago

joeylige commented 4 years ago

Thanks for your sharing. I wish to conduct an analysis on the topic changes on twitter. I wonder whether DETM is suitable for doing this.

mona-timmermann commented 4 years ago

You can certainly try! Maybe run ETM first to see how that performs as it runs much faster than D-ETM and the code is easier to understand. If that works well and if you then have a feeling of how to set num_topics, try D-ETM next :)

mona-timmermann commented 4 years ago

Other languages should be fine if you use pre-trained embeddings in that language or train Word2Vec yourself.

joeylige commented 4 years ago

Other languages should be fine if you use pre-trained embeddings in that language or train Word2Vec yourself. Thanks for the advice!

Emekaborisama commented 4 years ago

Thanks for your sharing. I wish to conduct an analysis on the topic changes on twitter. I wonder whether DETM is suitable for doing this.

Awesome. let me know how its goes

Emekaborisama commented 4 years ago

You can certainly try! Maybe run ETM first to see how that performs as it runs much faster than D-ETM and the code is easier to understand. If that works well and if you then have a feeling of how to set num_topics, try D-ETM next :)

Pls share the link to ETM

mona-timmermann commented 4 years ago

D-ETM was based on ETM but added the temporal evolution of topics as a feature. ETM: https://github.com/adjidieng/ETM

mona-timmermann commented 4 years ago

However, there are also models designed specifically for short texts which might work better

joeylige commented 4 years ago

However, there are also models designed specifically for short texts which might work better

Thank you for your reply. I find that there are models for short text analysis like Biterm Topic Model and models for topic changes like Dynamic Topic Model but fail to find an appropriate method to obtain topics changes in short texts, cuz I don't have much experience in NLP…If you could give me some advice, I will be very grateful!

mona-timmermann commented 4 years ago

Personally I don't really know of anything similar to D-ETM that works better on short texts. A different way to look at (e.g., popularity of) topics over time is doing a post-hoc analysis using the topics assigned to documents and their timestamps. Maybe check out the model topics over time, too. But these approaches answer different questions than D-ETM . Have you tried D-ETM and does it not work?

espoirMur commented 3 years ago

However, there are also models designed specifically for short texts which might work better

what are those model , can you share them?