This is quite a big change, so it's hard to list all the changes, but a quick sum up:
Main Features
New way of adding topics to articles - based on Keywords from URLs, which set "Gold" Labels which can then used in combination of SemanticEmbeddings + KNN from ESv8 to classify articles which have no labeled keywords.
Upgraded ElasticSearch to V8 which comes with new features, which we could look into implement into some of our existing features
Numerous Tools to monitor the performance of the Inference Process
Added an API endpoint to return a semantic embedding for a document (using distiluse-base-multilingual-cased-v2)
Added some tests to the new API endpoint + Keyword Topic assignment
NOTE
This is meant as an initial attempt as deployment requires slowly re-indexing the existing data in our database, so it shouldn't be deployed directly.
This is quite a big change, so it's hard to list all the changes, but a quick sum up:
Main Features
NOTE
This is meant as an initial attempt as deployment requires slowly re-indexing the existing data in our database, so it shouldn't be deployed directly.