Ask HN: Please review my OpenCourseWare Search Engine | Hacker News Reader

Ask HN: Please review my OpenCourseWare Search Engine

17 pointspierrefar15 comments

Hell all,

I've been hacking at this for many weeks and I'd love your feedback and guidance on how to improve it.

URL: http://www.ocwsearch.com/

What is it: A search engine for OpenCourseWare. Right now only MIT courses are indexed, but more universities and textbooks are planned.

Edit: It indexes all the lecture notes, the PDF files in the zip files, not just the course descriptions.

The blog has a poll for you to vote for what should come next.

What I'd like feedback on: Anything you see broken/improvable/missing. HN readers are the perfect target market so your feedback is going to carry a lot of weight.

Please get in touch: I've seen a few HN posts about OCW projects, and I'm very happy to connect with like-minded people. The email is pierre@(the domain name).

For the curious: The technology stack warrants its own blog post, so briefly: PHP, MongoDB, Redis, and Sphinx.

Thanks!

View on HN

I've done some work in this area, although more aimed at the school sector than higher ed.

One thing we found was that providing alternate forms of navigation and filtering is important.

For example, from memory the OCW metadata contains subjects. If the query result pages displayed those as facets then it would allow users to refine results by subjects. "School" and "Year" would be other good facets.

Solr does faceting really well, BTW

Others have asked for an advanced search interface, and I think your idea of filters fits right in with that. Thanks!

What I'm imagining now is actually at least two new user tools to help them do their searches: a query builder that does the advanced operator searches for them (like Google's advanced search form) and filters on the search results.

You got my brain in over-drive :)

My view is that advanced search interfaces don't add a lot of value, but filtering does.

People like to be able to explore result sets interactively. Properly implemented filters give you all the functionality of advanced search, but with the benefit of seeing exactly what you are getting dynamically.

Have a look at the filters on a search like this: http://shopper.cnet.com/1770-5_9-0.html?query=Core+i5&ta...

It gives much better feedback than an advanced search does.

Very useful product. This is a service I'm actually going to use. By the way, If you change your name to something more generic, add a couple of other educational sources to you engine as well, you will be google as academic earth is youtube.

Good point, and one I debated without a good resolution. For the launch, I stuck with what I had just to get something in the hands of users and get feedback.

Any ideas for different names? A fresh perspective would be appreciated.

Cool project! How did you get Sphinx to index MongoDB?

Link: http://www.ocwsearch.com/

Thanks!

I used the xmlpipe2 data source for Sphinx. A PHP script reads MongoDB and prints out the XML. Works a treat.

nice! looks like a useful service! The search was pretty snappy! Was impressed.

regarding the overall design, I think that a new make-up might not hurt ;)

What's your business model (if there is one), or are u doing it just for the fun of it?

Either way, keep up the good work! cheers

Thanks!

Yes new make-up is needed. Clearly I am not a designer and I know that. If you have ideas, I'm keen to hear them. Also, if you have favorite search engines in terms of design and/or functionality, I'd like to see them for inspiration.

Business model: there are different directions this could go. Of course the Creative Commons license is the main constraint, and I have what I think are compatible ideas. For now, it's just for fun because I actually really wished I had an OCW search engine recently, so I built one.

(unfortunately) I am not a designer either! When I opened the page, it design just felt a bit "old", e.g. not very "web2.0" like, if you know what I mean.

I think, that a bit less of clutter and a better focus on the important parts of the site (the search bar, and a clear message what value the site provides) would definitively help.

Further, something like "example searches" or "last searches" might be beneficial as well, e.g. to showcase the application. Just my 2cts ...

hth, cheers

Indexing lectures instead of courses is more useful.

It indexes the PDFs in the zip files. It's not just the courses, but the full text of all the lectures in each course.

What I meant is that I would prefer to see where exactly my keyword surfaced in the course - in what lecture or assignment vs lists of all courses mentioning the keywords.

Ah, sorry I misunderstood. Yes I completely agree. The "snippets" shown right now are just the official course descriptions. Better snippet generation is something I need to research before I deploy anything. It's a complicated problem to solve in a way that will work across all searches. Google used the meta description tags for years before they figured out how to best show snippets from page contents.

Even though it only has the MIT OCW data so far this is a most impressive and user-friendly interface. Congratulations, I'm sure it will appeal to a wide audience. Sorry but I'm not in your market niche so I do not know what's available from other institutions.

Reply on news.ycombinator.com