The Freename Forum

All good domains are already registered and concise, short and brandable terms can only be purchased for large sums of money. Really all of them? David Barnett introduces the method of ‘phonotactic evaluation’, which can be used to unearth a few gems from the riverbed of domain noise.

In an article on circleid.com, Brand Protection Strategist David Barnett (Stobbs) presents a method for generating short, articulate and brandable domains. In his study, he conducts a ‘phonotactic evaluation’ of unregistered five-character .com domains. Phonotactics is a measure of the potential readability or similarity of candidate strings to other existing words (or brands) in the corpus of a language. It is orientated towards English-language terms, but in principle this can be carried out in any language. A phonotactically usable string does not necessarily have to be attractive and brandable. This is why the result of an automated analysis must be analysed manually.

There are various phonotactic calculators. Barnett refers to the one at the University of California, Irvine (UCI), where he uses Hayes' BLICK model as an algorithm for the analysis. This model outputs a score for any sequence of phonemes (word candidates), which is a measure of the extent of phonotactic ‘violation’; a lower score then represents a more credible potential name. The implementation used in this analysis required each string to first be converted to its phonetic representation using ARPABET syntax, in which each phonetic element is represented as a series of Latin characters (and in some cases a trailing digit).

For his specific study, Barnett used only those 478,369 of the 9 million unregistered 5-character .com domains that begin with ‘a’ and ‘b’. As a result, the algorithm sorted out 8,893 (1.9 per cent) domains with a score of ‘0’ as unusable. The five domains with the highest scores were awbzp (59.43), bctko (65.17), anwjf (65.94), apgdj (67.26) and bchji (67.92). The majority of 5-character combinations were collected in the range between 1 and 10 points, with 25,000 (1) to over 50,000 (2) data. Thanks to the points, the result can be usefully filtered again by using certain point values as a threshold. For Barnett, it seems sensible to include further criteria in order to master the data sets. It is also impossible to avoid sorting out terms or fragments that are similar to brands. As a result, Barnett sees this method as a way of filtering out a few useful domains that can be marketed as brandable. Some domains - manually selected by Barnett - were accepted as brandable at the domain marketplace atom.com and displayed with values over US$ 2,000.

The article by David Barnett, which is well worth reading and contains numerous sources, can be found at:
https://circleid.com/posts/20240903-unr ... c-analysis

The phonotactic calculator of the UCI can be found at:
https://phonotactics.socsci.uci.edu/