How to incorporate timestamp into word embeddings?

[santosh.b...@gmail.com]

Hello community, I was wondering if there was a way to incorporate time stamp into word embeddings trainings. 

Essentially, I am looking at ways to examine whether cosine similarity between two word vectors changed from the period 2000 to 2023. I have text corpus from the same source for each year. 

I understand that training 24 separate word2vecs and comparing the cosine similarity scores may not be the most accurate way of doing it. 

Is that right? If yes, are there creative ways to incorporate timestamp? Would doc2vec be a able to handle this?

Many thanks!

sbs

[Andrey Kutuzov]

Hi,

There is a large field of research aimed at lexical semantic change
detection using word embeddings. You might want to have a look at these
papers:

https://aclanthology.org/C18-1117/

https://aclanthology.org/P19-1044/

https://aclanthology.org/2020.semeval-1.1/

> --
> You received this message because you are subscribed to the Google
> Groups "Gensim" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to gensim+un...@googlegroups.com
> <mailto:gensim+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/gensim/8ae0f4c3-dc62-4c90-8fc6-f65334d30c8fn%40googlegroups.com <https://groups.google.com/d/msgid/gensim/8ae0f4c3-dc62-4c90-8fc6-f65334d30c8fn%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Solve et coagula!
Andrey

[santosh.b...@gmail.com]

Thank you so much, Andrey. I will peruse the articles you shared - beginning with your co-authored 2018 piece.

Thanks!

sbs