1
1
mirror of https://github.com/github/semantic.git synced 2024-11-29 02:44:36 +03:00
semantic/weekly/2016-08-22.md
2016-08-22 10:32:48 -05:00

1.9 KiB

August 22nd, 2016 weekly

What Went Well?

@joshvera: A lot of pull requests, a lot of changes. Hard to remember everything we covered, but covered a lot of bug fixes, a lot of testing and improving term productions in preparation for staff shipping diff summaries.

@robrix: We covered a lot of ground especially in RWS.

@rewinfrey: Great to dive into the meat of RWS, and also make so much progress on diff summaries. Good week of work.

What Went Less Well?

@joshvera: Would have liked to get error reporting into diff summaries. Also on platform support the week before staff shipping so in hindsight, timing could have been better.

@robrix: Some concern that changes to RWS have overfitted it to our fixtures rather than maintaining general applicability. I have changes in mind, but those changes are on the magnitude of weeks, not days. Also apologies for pushing back reviews.

@rewinfrey: Programming by coincidence to improve our RWS output.

What did we learn?

@joshvera: Learned more about RWS. Also updates to RWS and test output allows us to verify if our models are correct in a loose machine learning analogous way. The pruning step before running RWS was interesting to learn about.

@robrix: Learned about biases RWS has and that our implementation has, and why the paper didn't run into them or describe them as much. Want to introduce a proportional feature vector rather than a stochastic feature vector so we can make comparisons more meaningful between tree nodes.

@rewinfrey: Learned about xml diff subscription and querying.

Other items

Staff Ship Plan:

  • Write blog post

  • Test github/github locally

  • Unless something major happens, we'll staff ship this afternoon (probably around 2pm EST)

  • Follow up with people in the #javascript for repo suggestions diff summaries

  • Reprioritize / add issues in the milestone for post staffship

  • Look at improving error messages and refining the diff summary productions