Hacker News Reader

Comment by jseliger | Hacker News Reader

This is low-hanging fruit not just for online stores, but for many kinds of sites, especially those that don't specialize in search. For example, I often have to use the site Grants.gov, which you'll recognize if you've ever had to submit or deal with research grants. And the search engine is terrible and frequently doesn't work at all.

The last time this happened, it inspired me to write a blog post about how to use Google for site-specific search: http://blog.seliger.com/2010/10/24/google-faster-than-grants... . Most of our readers aren't technical and probably don't know how to do this, but it can make search much, much better.

But that's for Grants.gov, and grants.gov dispenses money, which highly motivates its users. If you're trying to get users to pay you, on the other hand, you'll need to pay more attention to their experience.

View on HN

I had the same first instinct. Search is almost always an after thought unless it's the primary focus. If anyone wants a startup idea, the search feature on forums is something that people constantly suggest using, but generally provides awful results. Furthermore, a lot of forum owners restrict search usage for guests because it eats memory. If you make a replacement for vBulletin, phpBB, and Invision I wouldn't be surprised if you could sell it for $50+ to more active communities.

If you make a replacement for vBulletin, phpBB, and Invision I wouldn't be surprised if you could sell it for $50+ to more active communities.

The problem with search for forums is that there is no getting around the fact that you have to either search through the text of the posts (standard search) or have a massive index that gets out of date quickly (Sphinx). You can't get more creative than that and have your software still be compatible with most hosting (including $5/mo shared hosting that most people will have). When I worked at Internet Brands (makers of vBulletin), we used Sphinx on all of our forums, which worked, but was far from perfect. It's just the nature of the beast for forums...the interesting stuff is in the text fields (hard to index), compared to an online store where you can sort by category, tags, price, title, etc.

Why is the text inside the document hard to index? Isn't that what Lucene/Solr is for? Not sure what you mean...

Why is the text inside the document hard to index? Isn't that what Lucene/Solr is for? Not sure what you mean...

You just proved my point...if I have to install a standalone Java full-text engine running on Tomcat, I've already gone beyond what most forum owners can do on shared hosting.

Imagine searching through all posts on HN for the phrase "proved my point" anywhere in the post. There is no way other than iterating through all of the millions of posts.

For commonly-searched phrases such as "P vs NP", "scalability", or "rate my startup" you can create an index for faster searching, but it is impossible to anticipate every possible search, and in particular, long tail searches which can be the most useful will be the least likely to be indexed.

Imagine searching through all posts on HN for the phrase "proved my point" anywhere in the post. There is no way other than iterating through all of the millions of posts.

This is not true. Rather, in addition to the document ID, the inverted index stores the position information for each term occurrence, and can very quickly find any phrase of your choosing. While it's possible that this isn't always the default behaviour, all modern general purpose search engines support this.

I concur with nkurz here. Inverted index are the basis for most full-text search engines and they can handle this quite well.

Another issue with searching a large forum like HN though can be keeping the index up-to-date through crawling new posts constantly to add the terms they contain to the inverted index. However, you can usually build a simple script that adds news posts URLs to a list of priority crawl targets.

Free HTDig (http://www.htdig.org/) for instance could do the task pretty well and run on a budget hosting plan described above. It can index up to 100ks of documents easily and may require disk space upgrades for large sites but not much more.

There are many other inverted index search engines available and some now easily index without crawling through XML feeds from your CMS or forum software.

See my original post...I acknowledge that an inverted index can do it, but when you have to make it compatible with a wide range of hosting and technical ability as in the case of vBulletin, you are more limited. The typical forum owner will not install a standalone search engine, especially if his $5 hosting doesn't support it

Yes, but perhaps larger forums with dedicated servers can handle it? If this is true, selling a Search Replacement for forum software could really be a viable business.

yes but you could do it with a hosted search solution which, can index without crawling if it can be fed xml feeds of old and new posts...

Reply on news.ycombinator.com