A Better, Brighter Future for Domain Name Suggestions Has Arrived
GoDaddy was granted a patent this week from the USPTO (United States Patent & Trademark Office) for suggesting domain names based on recognized user patterns.
The patent addresses a difficult problem experienced by many of the domain name suggestion tools available today. Many solutions currently rely on static language dictionaries and include information such as part-of-speech (nouns, adjectives, verbs) and synonyms (words that mean the same) but it's been a clunky experience for the most part with few exceptions.
The problem, outlined within the patent filing, is that the pure dictionary approach limits the number of possible synonyms causing an inability to rank the popularity of the terms (the synonyms) when considering name ranking and which is most appealing to the domain buyers. And the frequency of words (their popularity in effect) is a key value component in the acquisition of a domain.
GoDaddy is essentially taking an entirely new approach to domain name suggestion by identifying interchangeable terms along with data about their popularity based on frequency of use (as it analyzed from data within domain names search logs and domain name zone files reportedly ).
Language models for name filtering and machine-learned ranking are then applied to further improve the quality of name suggestions (avoiding false positives or spurious suggestions) and the data is then likely piped back in to GoDaddy's for continued machine learning advancements.
GoDaddy already does a rather excellent job today of name suggestion and introducing domain buyers to available inventory of premium domains and this will likely solidify its position as the leading registrar.
If you're in the market for a new Web setup, GoDaddy is offering a $1/mo hosting (and a free domain name package right now, so this is your first chance to test out that domain name suggestion tool!