|Teragram and Alta Vista|
| 4:40 am on Aug 21, 2000 (gmt 0)|
I recently noticed that the bottom of each search return page on AltaVista has this linked acknowledgement: "Linguistic programming by Teragram Corporation"
Alta is giving them a lot of impressions by doing this, so I figure they are making some significant contribution and I started poking around a bit. Teragram is a Boston, MA company specializing in computational linguistics, and one of the favored varieties of computational linguistics today called "optimality theory".
Now I don't know what effect Teragram's work has on the AltaVista search results, but I do know that optimality theory could account for some odd things. It could weight one search term in a query as being more important or meaningful than another. So, searching on A +B might be highly weighted toward A, whereas searching on A +C might be highly weighted toward C.
I realize my post here might be about nothing at all, but when I notice that much favor being given at a high corporate level, I get curious. I did manage to discover that Teragram helped AV develop their "Discovery" product in 1998, so they've been partnering for a while.
| 4:58 am on Aug 21, 2000 (gmt 0)|
I've also notices some names that seem similar between Teragram and DirectHit. I was wondering if Teragram is a DH inc production? New technology, new name.
| 7:31 am on Aug 21, 2000 (gmt 0)|
That's intersting, Brett. You mean some of the same people are working with both companies?
DH and Teragram do seem to be interested in the same questions: "What did that search query really mean? What answers do the searchers seem to like?" DH goes about it by tracking actual search records and Teragram works to solve it upfront by pulling language apart. The two approaches would be a good fit.
Teragram site says they do context-sensitive spelling programs, stemming software, and HTML parsing for text extraction at a rate of 50,000 words per second (whew!). I can see where Alta might have some use for this stuff.
If searches are moved into the realm of meaning, rather than just matching character strings, the search engine becomes a very different critter. SEO becomes much more like writing clear and direct copy and much less like playing "Guess the Algorithm".
I think there's a very good chance that this is evolution is well underway, and that this is what makes search results seem so inscrutible at times.
| 4:48 pm on Aug 23, 2000 (gmt 0)|
"It could weight one search term in a query as being more important or meaningful than another. So, searching on A +B might be highly weighted toward A, whereas searching on A +C might be highly weighted toward C."
This is already in affect on AV. IDF (inverse document frequency) Gives words found on fewer documents in the AV database more pull than common words. This is why I reccomend optimizing for words in your search phrase seperately.
| 11:42 pm on Aug 23, 2000 (gmt 0)|
IMHO: Teragram is probably suppling AV with the ability to search their index quickly. Search Algo's are a hot property in Computer Science and if they have got a method for parsing HTML that fast then they are worth their weight. They almost certainly have nothing to do with the ranking, other than maybe optimizing search Algo's or AV's specs.
Maybe they spell check your page.
| 8:47 pm on Aug 24, 2000 (gmt 0)|
Teragram is all about Linguistics. Spelling, stemming, and correction. Go try
[urlhttp://www.spellonline.com/]SpellOnline[/url]. It is the best online spell checker I have ever seen. Try it with a url once - the followup display is excellent.
Also, check out Teragrams [teragram.com] product offering. It is a what's-what of linguistic software.