Open fedenanni opened 2 years ago
For each of these we should do an additional test and run them with broad search (see the example with #70 where we actually had mentions of Churchill). For the remaining ones we need manual checks to finally exclude possible bugs.
It seems a few of them might lead to results in other settings.
Current status:
dev
interface seems to be crashing when running broad entity search
queries. The initial problem was fixed in #79 and in fact the container doesn't crash, however it takes a very long time to perform the search. Next step:
Final summary of this issue:
broad_entity_search
(#70 )broad_entity_search
they retrieve way too many candidates to be processed:
Inquisition
Петр I
Russian literature
Kodály Zoltán
Kodály
LRS
For the moment we handle this situation by hardcoding cutoff on number of aliases ([d3db4c3](https://github.com/ArchivesPortalEuropeFoundation/Topic-Detection/commit/d3db4c39f141d5f778513acca9f5592a6c50b3e9)`)
- the rest are missing both using normal and broad search.
These should be double-checked to make sure they are not present in the full collection on APE:
Autres instruments de recherche
Familienwahlrecht
heritage and transport
Développement and Afrique and Europe
Samuelson
Beethoven music
steinthal
Pope Paul
greppi
Arhiv Republike Slovenije
Pope and Paul
heritage
Inquisición
Dalmine
First OR May
notaires
Slovenian
transport corridor Germany
Développement et Afrique et europe
Benedetto Croce
tenaris
Inquisition
franz ferdinand
paris conference
Kodály Zoltán
Kodály
Russian literature
rocca agostino
eric hobsbawm
German Democratic Republic
Greppi
Développement and Afrique
techint
keynes
Beethoven UND Musik
Петр I
Hortense de Beauharnais
eric and hobsbawm
LRS
Prešern
bretton woods
dalmine
Hospital Germany
Russian literature'
Marsaillaise
Thank you, @fedenanni. I probably won't get around to any testing in this regard during this week, but might be able to make some time for it next week.
@kerstarno no problem - this is just a final check before considering all tests completed. I believe the main issue for broad_search
is that timeout that we get when too many candidates are retrieved. That's why the method becomes so slow in certain settings
The following entities do not seem to appear in the dataset. Is it possible or is this still a bug and we know we have occurrences of them in the corpus?