garysieling

8 ms·

31.
by garysieling 8y ago
If you use Postgres, CTEs are optimization boundaries that make temp tables - you may end up inlining the CTEs or making them into regular temp tables so you can put indexes on them.
32.
by garysieling 8y ago
Do you know if there are common tools people use as a search engine backend for this? Vespa.ai seems promising, although I can't tell if there are many people using it.
33.
Using Postgres in Docker(garysieling.com)
1 points by garysieling 8y ago|0 comments
34.
by garysieling 8y ago
I built a search engine for talks ( http://findlectures.com/ ), and one of the explicit goals was not to collect user data for privacy. The content is selected by recommendations - e.g. someone on HN say a speaker was good, s
35.
by garysieling 8y ago
Two thing sthat really helped me with communication skills were technical blogging, and volunteering with Big Brothers, Big Sisters - trying to see the world through the eyes of a teenager is really illuminating.
36.
by garysieling 8y ago
SQL is incredibly valuable, it also allows you to do a lot of analysis to triage tickets or to get business insights (how many users does this feature affect, etc)
37.
by garysieling 8y ago
I found that blogging the steps as I do side projects made progress really helped with this - it makes it feel like I'm always completing something. It also allowed me to break a problem down into small steps (what be done in 30 minute
38.
by garysieling 8y ago
Oops, back up now!
39.
by garysieling 8y ago
I built something with a broader dataset but a different ranking technique, basically a bunch of cruder custom ranking rules. https://www.findlectures.com
40.
by garysieling 8y ago
I got frustrated with the Youtube recommendation system, and built a search engine of videos I scraped: https://www.findlectures.com I'm relying primarily on trusted recommendations - if someone I trust likes a speaker, I&#
41.
by garysieling 8y ago
A mix of Google Docs for larger research projects and notes on my blog - I use Google search to recover them. The blog entries let me save time if I return to a problem and tell me at what point I stopped.
42.
by garysieling 8y ago
Does Zeppelin work naturally with git? I've been struggling to get the right setup with just Jupyter
43.
by garysieling 9y ago
That particular speaker came from the Library of Congress, which I wrote a scraper for. The first level of categories come from either Youtube or the AlchemyAPI, depending on the talk. Levels 2+ are from AlchemyAPI or the scrape. The taxono
44.
by garysieling 9y ago
A mix of scraping and APIs - e.g. http://confreaks.tv/ has a great API where you can get the whole dataset in a json file
45.
by garysieling 9y ago
I got so frustrated with youtube discovery that I built my own search, although only for talks - https://www.findlectures.com/
46.
by garysieling 9y ago
I put time into my search engine for talks ( https://www.findlectures.com/ ). I did two conference talks on it, at Solr/Lucene Revolution ( https://www.youtube.com/watch?v=e_ia1DRz3l8 ) and AI With The Be
47.
by garysieling 9y ago
I ran about 800 from http://sacredwindowrescueproject.org/ through this. Some of the companies that do this work have been around for many decades and have tons of photographs / scanned images, so I'm investigatin
48.
by garysieling 9y ago
It's appropriate in search, if you use the data as features to improve recall
49.
by garysieling 9y ago
I ran a bunch of stained glass images through several APIs - Google's did by far the best, although there are a lot of issues (curved text, hand-written text in odd alignments or on a curve)
50.
by garysieling 9y ago
You only pay 33% federal income tax on the income over $191,650. If you corrected for that, the amount left over would still be close to the median income for a family in the U.S ($59k).
51.
Using Word2vec for Concept Search(garysieling.com)
5 points by garysieling 9y ago|0 comments
52.
by garysieling 9y ago
I make notes for each project in a google doc, i.e. collecting material as if I was going to do a write-up or talk on the project. That makes it easier to set things aside for a while, and if you discover something by accident while you&#x2
53.
by garysieling 9y ago
If you have this information, one thing that would be really helpful with this type of tutorial is to know how long it took to train the models when you tested them last (and what it cost).
54.
Using Word2vec for Concept Search(garysieling.com)
11 points by garysieling 9y ago|0 comments
55.
by garysieling 9y ago
Good to know, thanks!
56.
by garysieling 9y ago
If you wanted it to know that "machine learning" and "neural networks" were related, wouldn't you need to do some type of entity extraction first, since Word2vec is run on tokens?
57.
by garysieling 9y ago
For integrating into Solr, I've used Word2vec to improve rankings the synonyms it finds (boosting synonyms by how similar they are to the query). In English Word2vec tends to think plurals are synonyms. I did a talk on this with more d
58.
by garysieling 9y ago
Yeah, that makes a lot of sense. I started this from making an email list where I'm sending "lunch and learn" talks which I manually curate / write up descriptions. About half the people like that style, and half have sp
59.
by garysieling 9y ago
That's an interesting project. I've been exploring this from a different angle, in the form of generated email newsletters ( https://www.findlectures.com/form?type=alert ). This uses the contents of the articles as
60.
by garysieling 9y ago
I've experimented with building something like this but not at any scale. Is there a use case that doesn't involve writing code to do the personalization? I've been picking articles with some machine learning stuff but I&#x27

More ›