This application shows an overview of the major topics being
discussed all
over noblogs.org,
in the last few weeks.
The analysis process is automated and unsupervised. As such,
results are only approximate, and occasionally can even be
wildly inaccurate. If you disagree with how something has
been categorized, you are probably correct.
This is the first result of our exploration of modern
Machine Learning technologies from the perspective of
developing useful tools for our communities. We think it is
valuable to provide such examples as a counter-point to the
current discourse around ML, centered around fear so as to
discourage understanding, when instead greater understanding
of these technologies is necessary in order for the people
abusing them to be held accountable.
Technical details
The data is analyzed with a combination of ML and
traditional numerical techniques,
using BERTopic
with Llama2 for summarization.
The analysis runs on public posts only, and in any case the
data never leaves our control. The ML pipeline runs on our
own hardware and does not rely on third-party services.
