What does search by rank do?

Tim Engwer

I have been using the basic search by rank rather than title as a substitute for topic searching for quite a while and I certainly see a difference.  Search by rank seems more topically oriented and yet I know that true topic searching is not yet available in L4. 

Exactly what happens when the search hits are sorted by rank rather than title?

Dave Hooton

I tried searching for threads as this has been raised before. The Help  states "search hits in  order of relevance". I think it is the number of hits per word of the immediate text.

Kevin Becker

Ranked results is when the search engine puts whatever it believes is the most relevant hits at the top of the list, regardless of source. I haven't come across any explanation of what logic Logos' algorithm uses to provide this order. It's quite possible that Dave's speculation is right that it is a reflection of the density of that reference in a resource.

Bob Pritchett

Search ranking is a complicated process; we started with "standard" algorithms, but may have some subtle tweaks, and plan more in the future. (Like Google -- it's always changing.)

In simple terms, it's a measure of how similar the target "document" (an article in a book, in our case) is to the query. Now there's a big difference in document length and query length, but the basic idea is that you rank higher documents in which rarer query terms appear in higher frequencies. 

So if I search for 

the huge dog

you'll get a higher rank for documents in which "huge" and "dog" appear frequently, but the frequent appearance of "the" in a document won't promote its rank much, because "the" is not at all rare. 

If you want to go further, start with reading about "tf-idf" http://en.wikipedia.org/wiki/Tf–idf

Beyond that, every search engine starts introducing special hacks and tricks to improve ranking, by usually subjective criterica. (Search "pizza san jose" while you're in San Jose, California, and Google gives higher rank to pizza restaurants near year. Search for the same thing while in Costa Rica, and it'll promote San Jose, Costa Rica restaurants higher in the list, etc.)

 

Tim Engwer

https://community.logos.com/discussion/comment/111267#Comment_111267

Thanks for the reply.  Is the intention to eventually come up with a complex algorithm that comes up with results that are increasingly topic-like?

Dave Hooton

https://community.logos.com/discussion/comment/111563#Comment_111563

TimEngwer said:

Thanks for the reply.  Is the intention to eventually come up with a complex algorithm that comes up with results that are increasingly topic-like?

I doubt that. If you run a Basic Search you will often see the Topic section gets populated with results from dictionaries and other sources. Logos are making a concerted attempt to improve topics compared to the results L3 was able to provide. See http://community.logos.com/forums/t/12424.aspx