8:00 – 6:00 SR
- Well, we have hoverboards, after a fashion.
- Discussed the new approach in some detail with Lenny. It turns out that the current system has around 140 tables and about 15 views spread across 5 databases. So getting it down to one DB with 5 tables makes some sense.
- Adding direct text interaction to the AlchempNLP code – Looking for a (good?) abstract
- So here’s kind of an interesting thing. Google has my citation history, and uses that to make recommendations. But since my previous work is surgical simulator heavy, I’m not getting results that are useful to me. I want to be able to adjust weights! I think I’ll call this ‘the self driving car problem’ 🙂
- Refactoring alchemyNLP pull into a base class and then an URL version and a text version
- Meeting on campus with Wayne, Shimei and Victor.
- Since by next week, something should be working, Wayne would like to use the system to try and find mentors. Sheimei suggests using the dblp as a source. It’s structured and can be downloaded
- I realized that items should(?) have their own(?) networks of items that are attached at a system level. For example, when AlchemyNLP parses a web page, it produces concepts, entities, author etc. Rather than re-running the analysis each time, the system should look to see if there is already an <itemGuid_Alchemy> network produced the first time the item was ingested. Then those items can be attached to the user network as needed.
- Talked with Victor about collaborating somehow. We also talked about how to present a list of search results in a ‘curated’ context. It could be a sharded display where each item listed has a slider/add/discard control, and hovering over the item shows it’s relationship to the curated collection on the right hand side of the page. Choosing the context could be in a dropdown at the top that is guessed at by context, but can be changed by the user.