Show HN: Search engine for lectures | Hacker News Reader

Show HN: Search engine for lectures

findlectures.com

200 pointsgarysieling38 comments

The idea is really good, but there is still a long way to go.

Searching for "python simulation" (a suggested search, btw) yields results such as "introduction to programming lecture 23" - short of watching the lecture, how am I supposed to judge it's relevancy? There is no description, and the title contains neither "python" nor "simulation".

Also, the email sign up modal is enough to not make you want to return.

Thanks for the feedback. The email thing should be better now.

Agree on the relevancy issue, I'm working on ways to improve that. If you do a search that matches the closed captions, it will tell you where in the video the match starts.

On mobile,

The email modal is over most of screen and can't be closed...

Seconded. It's super annoying and while it can be worked around, I don't like having to deal with it following me around as I scroll.

Really sorry about that. I pushed a change to hide it when it doesn't fit.

That area of the screen is supposed to be an informational card, like you see in Google search results. When I have the data, it shows books a speaker has written, and I'm working on getting more in there.

I tried to look for some math topics, such as "harmonic analysis", "lie algebras" and "model theory" - unfortunately none of the matches were relevant for any of the queries

Thanks, I'll look at adding some math classes.

Hey this is pretty cool! I looked for a specific CS topic and the result was definitely better than Google's. I'll probably keep using this!

Looks interesting. As someone going through all of Dr Jordan Peterson's lectures, I put him in the search and only one of his talks (with Joe Rogan) came up. Please add his stuff if you get the chance (he has a YT channel). It'd also be nice to be able to find like-minded talks.

Also, when pressing the back button through some searches, the URL changed with me but the results stayed the same.

Thanks for the feedback!

As someone who does a lot of learning online, I can see the usefulness of this! It'd be a lot easier for me to use if the filters on the side of the screen if they matched my search queries more closely. Currently, searching for "react.js vs. angular.js" leads to Ronald Reagan and Bill Clinton being options in the "speaker"filter.

Thanks! I think you'll get a better experience if you filter to "Technology" first, if you're using it that way. I'd like that to be more seamless, but haven't found a good way yet.

Just curious, is this using React/Searchkit/Elasticsearch under the hood?

It's a custom React UI + Semantic UI + D3 + Solr. If I had found Searchkit when I started this a year and a half ago, I would have used it!

It shows ApplicationError. Got a HN hug of death, or are you updating?

Sorry, looks like a memory issue on the Solr server. I shut everything else down, so hopefully it holds!

Good initiative. Suggestion: The content could be categorized.For example - the lecture of MOOCs should be at one place.

Problem: Say, I am browsing page 5 for some query and I change the query, it shows page 5 of the current query rather than page 1 and if the content of current query is not enough to reach page 5, it shows "No results found, sorry!".

Good feedback, thanks.

Great project Gary. Thanks for sharing (and building). I would love to see this kind of content get some love.

Thanks! I'm getting a ton of good feedback and lots of speaker recommendations, so the next version should be even better!

Interesting project, but as op said email modal on mobile can't be navigated out of.

Sorry about that, it should be fixed now.

Great project, thank you for your time!

(minor bug: Searching for "top 20%" results in "URIError: URI malformed /public/lecture_search.bundle.js?v=1.0.22:79346")

Thanks, I will look a this!

Nice! That's going to be very useful when learning a new topic/technology

How are the hits ranked? I imagine just time? Do you have a list of all sources you used?

There are about 20 factors right now, which I've been tuning over time. They are applied as I'm able to get the data for the talk. Here is a selection-

1. Is the publisher of the talk good? (whitelist)

2. Is the talk referenced in social media (e.g. HN)

3. Safe for work (i.e. you could use in a lunch and learn)

4. Length (15-50 minutes preferred)

5. Is the speaker a keynote level speaker?

6. Are there closed captions? Can the video be embedded?

7. How often does the speaker say 'um' (or similar)

8. Is the talk in English (there are some in Russian and Arabic, but only small collections)

9. Are there notable audio quality issues (channel balance)

10. If it's a historic talk, older talks get higher ranking

11. If it's a technical talk, these get lower ranking, unless you filter to that

12. Is the speaker influential enough to have a wikipedia article?

13. Has the speaker written books?

14. What is the role of the institution that sponsored the talk (conferences are neutral, think tanks, etc lose some ranking)

I think a language filter would still be good.

Great resource but your filtering needs some serious work. I managed to filter myself into a corner with no way to get back.

Thanks for the feedback - I'll look into fixing that.

Pagination doesn't seem to be working, I need to change the URL to go to page 2 and so on when filtering a cathegory.

e.g.: https://www.findlectures.com/?p=1&class1=Technology&category...

Sorry about that, I will look into this.

Thanks! Great site!

Really interesting, are you searching through Youtube? Wonder what other sources that you are getting the source data out of

I'm collecting lists of speakers/conferences and doing very targeted crawling. There are a couple essays on the site (in progress) if you're interested in more detail-

https://www.findlectures.com/articles/2017/05/15/Building-a-...

this is awesome. thanks!!

obviously a huge undertaking probably not in your wheelhouse- but finding a way to map all these topics together would be amazing.

I'm not sure if this is what you mean, but looking at ways to map concepts together with word2vec. E.g. so you could search "writing NOT code" and get literature, or facets more conceptually related to search results.

interesting but i was thinking more of a Khan-style mindmap: http://2.bp.blogspot.com/-ChwlFfoI4TE/TtEl5UA1AkI/AAAAAAAAAe...

Reply on news.ycombinator.com