Latest DomainsBot software starts to tackle challenge of hundreds of top level domains.
DomainsBot announced its new TLD Recommendation Engine today.
The service considers a number of data points to return the most relevant top level domain options for a given search. It considers content analysis of popular websites, frequency analysis of searched and registered domains, social media data, and geo targeting.
I was eager to try it out given the coming changes to domain name search. You can give it a spin on a demo version at demo.domainsbot.com.
Overall, I found it to be a good start. It seems to fall down a bit on showing TLD alternatives that I would consider relevant for a given search.
This demo version of the software only takes into consideration “association rules”.
DomainsBot CEO Emiliano Pasqualetti explains, “This means that TLDs are selected exclusively on the basis of the frequency of its co-occurrences with the searched keywords as a registered domain.”
Here are some example queries and results.
Query: jims pizza
Results:
jims pizza.eat
jims pizza.food
jims pizza.menu
jims pizza.restaurant
jims pizza.pizza
This one is good. It takes “pizza” and knows that other relevant TLDs will be related to food. If a registrar also integrates geo data, it might also offer a TLD such as .nyc, .berlin, etc., depending on how easy it is to register those domains.
I also think a good domain search engine will recommend domain hacks, as these will become much more relevant with longer TLDs. For example, Jims.pizza is a good result for this query.
Query: rogerscpa
Results:
rogerscpa.cpa
rogerscpa.org
rogerscpa.info
rogerscpa.biz
rogerscpa.us
DomainsBot correctly extracts CPA from the query. But it misses the boat on similar TLDs, including two obvious ones, .accountant and .accountants. .Tax might also be a good suggestion. Even if these don’t currently have a lot of co-occurrences with CPA in domains, I think they’re quite relevant.
Another example I ran was a company name followed by “sucks”. One of the returned results was .sucks, but .gripe was missing.
Query: auctions
auctions.auction
auctions.bid
auctions.estate
auctions.property
auctions.realestate
These are good results. Not only did it get two obvious ones, .auction and .bid, but it knows that many auctions are for real estate businesses.
Query: slam dunk
slam dunk.com
slam dunk.net
slam dunk.org
slam dunk.city
slam dunk.co
.basketball is an obvious choice. What’s interesting here is that if you search for slamdunk without a space, one of the results is .sport, which is a step in the right direction.
Query: Haircut
haircut.company
haircut.rocks
haircut.top
haircut.web
haircut.live
What about .salon? .hair? I’m not sure how .rocks gets in there.
DomainsBot’s latest release is a step in the right direction to a difficult problem, although it needs some refining.
This tool is joke… Testing results: Enter “a” (single character) and it will suggest you to register a.com – huh! reserved domain name. Enter “aa” (two characters) and it will suggest you to register aa.web – huh! reserved again, as new gTLD blocks two-character marks…
@ Mark – this isn’t live as new TLDs aren’t live. An availability check is run in the background when the technology is actually implemented.
So the cocktail party test is just out the window now. Once this stuff hits the streets even the best most premium .something won’t be memorable to anyone other than the one who bought it.
@Andrew – So then it is a good feedback for them to adjust their tools to allow only 3+ characters inputs…
Mark, I think the demo is just to show the kinds of result they’ll pull up, but when it’s actually integrated with a live registrar system it wouldn’t return the results you got.
The demo version focuses on exact match of entered keyword(s). If this becomes a trend, then current practice of showing entered keyword(s) + suggested keyword may disappear, effectively making long domain names unfavorable.