Features
[Search] Introduced Scoring Profile Execution Stages; query: Apply profiles on all matching items vs. rescore: rescore only top N ranked items (useful to apply more expensive operations on a smaller, most relevant subset)
[Search] Spellchecking Improvement:
Add full support for multilingual spell-checking (only works for languages that are actually used within a project aka having documents loaded)
Perform spellchecking only on User's input-terms and not on other Query Syntax components (filters, comparisons etc)
[Search] Forward query context to query activity log
Allow disabling old file removal in pdfconversion.
[Search] Introduced query classifier step that infers the type of the provided query. It detects whether the query is written in natural language form.
QA is then only performed on queries in natural language form
[Spellcheck widget] Make spellcheck suggestion application case insensitive
Updated libNLP documentation for step classes analyzer, classifier, debugger, embedder.
At the moment, we configure the number of community and community items in a newsletter in a config file. In a future release, it will be possible for a user to configure it in the UI. We ship the API for this in the current release.
We expose the functionality to automatically subscribe users to all the Accounts they own. An owner is responsible for an account and should always monitor assigned accounts automatically. Since the subscription is a continuous process, the auto-subscription also works continuously alongside community updates as new accounts might frequently be ingested as communities.
Add user_id as a parameter to the community subscription API which allows a user to subscribe others to different communities.
New Squirro Monitoring projects, except of activity tracking data, will now include data ingestion and ML monitoring data.
...