What a week it's been!
Long days doing assignment hacking, holding meetings, attending classes, and also managing to fit in some R&D hacking in the midst of all that (*). Oh, and having my birthday fall in the middle of the chaos (BTW, it was a great day; with nice, sunny, warm weather; even the parking meters played nice and had free parking but just on that day only :D)
Unfortunately, that has meant that Blender development work has had to take a backseat for a short while. Hopefully in just over a week, I'll finally have time to work on more Blender related stuff again. Perhaps even starting work on some long postponed projects there...
(*) In case you're curious, this R&D hacking has been related to automatic classification/clustering of short text snippets in response to some question. By grouping these text snippets by primary topic, taking into account an abnormally high rate of typos imposed by the use of limited-size devices, we can thus derive a simplified overview of the response-set-space.
This allows us to gain insight into the significant trends (with frequency metrics included for comparison purposes), which a simple flat-list format may not provide, and also hopefully act to counteract concerted attempts to abuse the system with spam (aka "offensive material"), though traditional black-listing technques would likely be necessary.
It's actually turned out to be quite an interesting endeavor. Reading about various data mining techniques and algorithms, natural language processing and classification, and fuzzy pattern matching techniques. And that's only just scratching the surface!
Perhaps a good topic for another day, when I manage to solve some of the remaining tricky aspects for longer texts ;)
No comments:
Post a Comment