Probably the easiest and straightforward first-pass is not an AI tactic. A semantic text analysis of content areas to generate metadata. I suspect what you will find is that within certain areas the keywords and phrases bubbling to the top will be what people are working on recently, or what people talk about a lot, making the results skewed. For example, if we did a semantic text analysis of this AI Forum, it would reflect much of what I posted, because for a while I was regularly posting to this forum weekly, so naturally the skew would reflect my biases. Even an ML system would detect similar. If 100% of the forum data split into 70% training data for modeling, and 30% reserved to apply to the model - the results could very well be skewed due to the extreme smallness of the data set and the amount of material I posted in the past.
I think that a problem in any focused technical community, is circularity and circular reasoning. We have to accept as true the very thing we are trying to prove - we begin with where we are trying to end. People bring in topics that they have an emotional attachment too (a problem, a solution, a motivation, a bias, a call to action, etc.) and the community acts like an echo chamber where the conversation gravitates to the mean and amplifies the message. A ML algorithm would discover what everyone is already taking about, because its in the data set (this is the problem with recommender systems amplifying extremist messaging by recommending views of additional extremist messaging). I suspect what you ask for is a NN (Neural Network) where the network discovers patterns that are not apparent by filtering the data back through n-layers of itself.
It maybe simpler to do a reflective study after semantic text analysis, to appreciate what people are
not talking using industry standard TOC's as a guide. Areas of little discussion could represent neglected, stalled, or under-developed topics in need of further investigation.
------------------------------
Mark Yanalitis
------------------------------
Original Message:
Sent: Nov 12, 2020 01:50:37 PM
From: Jim Reavis
Subject: AI and the CSA knowledge base
Hi All,
This post might be a little off-the-wall, but bear with me.
My theory is that in the 11+ years that CSA has been in existence, we have created so much content in a variety of ways that we have solved many issues that organizations are grappling with. The problem is that this information is scattered among whitepapers, webinars, working group meetings, Circle discussion threads, chapter conferences, podcasts, etc., and there isn't an easy way to connect the dots.
The idea I would like to explore would be to see if some type of AI system could ingest CSA content in all of its formats and data structures and help us organize this information and gain insights into the industry problems and solutions we are facing. I have had initial discussions with a company about this, but I would like to broaden the discussion and perhaps get some expertise from the Circle community to weigh in.
Thanks!
------------------------------
Jim Reavis CCSK
Cloud Security Alliance
Bellingham WA
------------------------------