Open stefangrotz opened 10 months ago
Also discussed in the revo repo: https://github.com/revuloj/revo-fonto/issues/61
I managed to get a result after 20 seconds, still pretty slow.
Todo:
The sql file also works (45 seconds). This is my prompt for it:
Reta Vortaro - revo-inx.db
Use this file to search for Esperanto words and their translations.
Database Structure:
nodo
: Contains main entries (Esperanto words). Key columns aremrk
(unique marker) andkap
(the word).traduko
: Holds translations. Key columns aremrk
(linking tonodo
) andlng
(language code of the translation).var
: Stores variations of words. Key columns aremrk
andvar
(variation).- Other tables like
referenco
,uzo
,malong
,bildo
,artikolo
,vortspeco
,agordo
may contain additional information but may not always have relevant data.Finding Words:
- To locate an Esperanto word, query the
nodo
table using thekap
column.- Example SQL:
SELECT mrk FROM nodo WHERE kap = 'desired_word';
Getting Translations:
- Once you have the
mrk
fromnodo
, use it to find translations in thetraduko
table.- Example SQL:
SELECT txt FROM traduko WHERE mrk = 'obtained_mrk' AND lng = 'language_code';
Word Variations:
- To find variations of a word, use the
mrk
in thevar
table.- Example SQL:
SELECT var FROM var WHERE mrk = 'obtained_mrk';
Additional Information:
- For more details like usage or references, use the
mrk
to query tables likeuzo
orreferenco
.
You can search through the revo.xdxf file using code interpreter:
Used prompt
I used this prompt with a minified version of an example python code and some information about the data structure:
Conclusion
IMO right now it is too slow to include it into EsperantoGPT. It takes almost one minute to look up a word. I tried to make it use minfied code to speed things up, but this hasn't worked until now.