Fasttext: Question: How to analyze sentence similarity under fastText?
I have a requirement on hand, e.g. user input some text (Chinese) inside the input box of the app, and we need to redirect user to other pages according to user's semantic meaning.
So, can fastText do some helps on this?
Thanks!
leonardgithub
All 4 comments
You can analyze the sentence similarity by averaging the value of the word vectors and find the nearest neighbour according to a similarity measure (e.g. cosine distance). You might want to benchmark with simpler models like tf-idf, bag-of-words, etc.
amirothman
on 8 Sep 2016
Hi,
As this is not an issue per-se, I would recommend posting this kind of questions on the FB fasttext users group.
Please note that since commit d652288bad3d5e30a350c67599df4c92dc471960, there is an option for printing vectors for a paragraph after supervised training.
Hope this helps!
piotr-bojanowski
on 12 Sep 2016
@amirothman @piotr-bojanowski Thank you very much.
By the way, @piotr-bojanowski, Is it possible to commit a question issue on this github?
leonardgithub
on 13 Sep 2016
@piotr-bojanowski
Hello,
what's the different between 'print-vectors' and 'print-sentence-vectors' ?
'print-vectors' is Unusable now ?
Thank you for you reply!
letotefrank
on 29 Aug 2018
Related issues
a11apurva
路
3Comments
flybirp
路
4Comments
AhmedIdr
路
3Comments
hughbzhang
路
3Comments
danoneata
路
3Comments
Most helpful comment
Hi,
As this is not an issue per-se, I would recommend posting this kind of questions on the FB fasttext users group.
Please note that since commit d652288bad3d5e30a350c67599df4c92dc471960, there is an option for printing vectors for a paragraph after supervised training.
Hope this helps!