[Hivemind] Data processing improvements

[amindadgar]

We'd like to improve the data processing ETLs for hivemind in order for the hivemind-bot to give answers better.

Here are some of the task lists to be done

• Skip ingesting non-knowledgeable data ("gm", "hi", "good talk") Improvement
• Remove stop words (“the”, “a”, “an”, or “in”) Improvement
• Skip messages with less than 3 words (they are probably non-informative)