Building a NLP email analyzer to draw the relationship graph
I've been so annoyed to go through the inbox search result and jump back and forth to figure out what happened. (especially when there are multiple conversation threads and span across a long time period)
As a result, I'd like to build an add-in which use NLP to parse the titles and contents. Use semantic components to build the relationship graph. Each node in the graph will contain entity (email key words), features ( related information ) and connections to others. The final result will be drawed as a relationship graph, to give a direct clue leading to the context.
Currently, i'm not quite sure how the workload would be, when dealing with the model training.
Dump anything in your mind of this idea. Useful vs total bullshit? Require tons of work? Already sovled?
Essentially you just want a graph as a summary if an email chain? If I’m understanding you correctly that sounds pretty unique actually - and probably pretty appealing to people who are more visual based learners/retainers. Also could see something similar for summarizing Slack conversations that you are behind on!
I’d check out Spacy (https://spacy.io) it’s a fast python NLP library that has some thing built in that I think you’d need to build this- namely dependency parsing and word vectors.
Good luck if you decide to build this! Sounds interesting!
I think this is a cool idea but I also think it's not as easy. And, probably already implemented it. Have you done any search?
tried to search gmail and outlook add-in market, didn't see a good one
This comment was deleted 6 years ago