1 Experiments

  • Tried making a speech to intent system based on phoneme decoding. An interesting result, which I was not expecting with low data, was that dense representation learning for phonemes help a lot compared to even higher order n-grams. My hunch is that looking only at top-1 phoneme transcription is going to have high brittleness.

2 Readings/Explorations

The charter also identified which projects were in scope and out of scope for our work, and which were required features and which were optional.

The time spent writing the charter later paid for itself many times over and it’s an exercise I recommend to any committee engaged in any large endeavor over a period of time.

  • Metaobject protocols: Why we want them and what else they can do (kiczales1993metaobject).
  • Explored custom jupyter notebook widgets. The widget framework is really useful. I am now trying to move a lot of EDA, tagging etc. systems to widgets.

3 Programming

Commits for week 43-2019 and 4 previous weeks.

4 Media


  • [kiczales1993metaobject] Kiczales, Ashley, Rodriguez, Vahdat & Bobrow. 1993. "Metaobject protocols: Why we want them and what else they can do." Object-Oriented Programming: The CLOS Perspective, , 101-118. link. doi.