Topic Assistant

“A utility to map arbitrary text to the WLO/OEH topics vocabulary based on keyword matching.”

the only endpoint currently targeted in the Redaktionsumgebung

endpoint POST /topics

uses:

`Language Detection`

Uses langdetect of python to detect language - has been used in MetaQS as well.

Predicts discipline of content

Returns integer, corresponding to discipline of content

Purely detection of duplicates based on pre-trained text

uses MinHash algorithm

currently trainied for id, url and description of some data

Returns: "a list of scores and document ids relevant to the query document. Only the top ten items are retrieved, in descending order"

Model pretrained, details unclear - same model as classification?

Analyzes the OER object itself and yields categories, e.g., which categories title, description and keywords could belong, too

Needs training, "Kreissektor" is connected to "Allgemeine Psychologie"

Verschiedene Mappings, e.g., OER Kategorien zu Wikipedia

some outdated links

kein Rest-Endpunkt o.ä.