Paper Reference

[pribadihcr]

Hi,

What is the paper reference for this repo. Thanks

[hannahburkhardt]

Hi there! The ESP paper is here: https://pubmed.ncbi.nlm.nih.gov/28284761/
Cohen T, Widdows D. Embedding of semantic predications. J Biomed Inform. 2017;68:150-166. doi:10.1016/j.jbi.2017.03.003
Thanks

[pribadihcr]

great. any example script, how to use ESP for document vector representation?

[treversec]

Morning all - we haven't used ESP to build document vector representations, although it should be possible to do this if you can represent the document as triples somehow. The closest we've come to this is this paper ( http://www.puttypeg.net/papers/expansion-by-analogy.pdf) in which we used vector representations of concepts derived from triples to infer connections between document and query vectors. The semvecpy doesn't (yet) support training of models - all of this is done with our Java package, Semantic Vectors ( https://github.com/semanticvectors/semanticvectors). The Wiki is a bit sparse unfortunately, and I see we've yet to add documentation for ESP - however the instructions to build a space are more or less the same as with our previous model of triples, PSI: https://github.com/semanticvectors/semanticvectors/wiki/PredicationBasedSemanticIndexing. A good place to start might be to run through these and substitute pitt.search.semanticvectors.ESP for pitt.search.semanticvectors.PSI. Apologies for the lamentable state of the ESP documentation, and thanks for the reminder to update the Wiki.

…

On Thu, Aug 6, 2020 at 12:17 AM Pribadi HCR ***@***.***> wrote: great. any example script, how to use ESP for document vector representation? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#19 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXSNCXCWCHZ6CH6ZXYW4LTR7JKIFANCNFSM4PVDYK5A> .

[hannahburkhardt]

To add to that, if you’re curious about embedding predications (subject-relationship-object triples), there’s an example in this notebook: https://github.com/hannahburkhardt/predicting_ddis_with_esp/blob/master/predicting_ddis_with_esp.ipynb
You can find the actual java commands I used in this work in the section titled “Train the model”.