search

jhu-bids / fhir-zulip-nlp-analysis

Ad hoc NLP (Natural Language Processing) analysis of HL7 FHIR's online Zulip chat streams.
MIT License
0 stars 0 forks source link

Bugfix: Incorrect results #5

Closed joeflack4 closed 2 years ago

joeflack4 commented 2 years ago

Updates

    - Bugfix: Bugfix: Counts report was incorrect. I wasn't taking into account what was already in the cache. The easiest way for me to handle this was to move the "keyword report" out of `query_keyword()` and into a new function, `create_report1()` which is now called near the end of the program execution, after the cache and new results have been combined.
    - Refactor: Misc comments, style changes, TODO updates, and other minor code changes.
    - Bugfix: The "no messages found" report was also buggy due to the same issue with the "counts report". The logic for it needed to be moved out of `query_keyword()` and into new `create_report1()` function, called near the end of the program execution.
    - Delete: I decided to remove the caching of JSON. It wasn't really useful at all, and was just taking up lines and making everything more complicated / harder to read.
    - Docs: Added links to releases.
joeflack4 commented 2 years ago

@rohaher @DaveraGabriel New results: zulip_report1.csv I've also uploaded this to releases.

There was a bug before (described in an a comment above), which is why so many keywords showed no results, and why the rest of the results looked off. Should be good now!

The file size for the raw messages was too big, so I uploaded to Google Drive.