Looking for techniques for building a KG using Slack as a source.
Both open source and proprietary solutions.
Any input appreciated!
Really interesting @daria, and at Derwen we’ve been looking at the Slack SDK regarding potential areas of integration for kglab
. FWIW, our Slack board is set up more toward being a dev/test platform.
What would you want to build as a KG from a Slack board? Some ideas come to mind:
- channels
- members
- member relations to channels, how much they post, date ranges
- topics per channel, from the raw text
- links posted (e.g., to external resources online)
FIWW, several years ago I did a Scala tutorial for Apache Spark called exsto
for Scala Days 2015, etc. The ASF had a formal requirement that open source projects run developer discussions through particular email forums, as part of the project governance standards. We used exsto
on the Spark developer forums and surprisingly found through graph analysis that the top-ranked contributor was from a competitor firm. Databricks later hired that person into a lead role
Similar approaches could work now with Slack. If you’d like to collab on this with the kglab
dev team, we welcome you!
Thank you @paco - this question sprung up from a conversation with a friend.
I actually tried to play around with a set of archived Slack channels (full, mostly, of Q&As). The idea was to have a clearer picture of a few days of intense conversations and knowledge exchanges.
I personally lack - yet - a rigorous background on KGs, but I personally like working on semistructured data (like text) and was looking for how this is being done.
There’s some good material from RASA about parsing text for conversations, and Vincent Warmerdam provides really excellent tutorials, AMAs, etc. That’s mostly about building conversational agents.
Another approach that I’ve seen for modeling conversation (not agent response) is to leverage the “ontology” and metadata typically used in film scripts. That may not be far from modeling Slack convo? From what I’ve seen, handle a Slack channel as if it were a scene in a film - with characters, director’s notes, etc.