To prepare this analysis, we look at data that your followers have chosen to make public on Twitter (their names, locations, bios etc.), and process this iteratively, using a series of algorithms developed in-house at schmap.

These algorithms are based on the tenets of fuzzy set theory (as pioneered by the mathematician and computer scientist Lotfi A. Zadeh), a probabilistic logic that we use to reassess the likelihood of given classifications, based on often conflicting interpretations of multiple data signals. It’s the kind of approach that’s fundamental to the promised Web 3.0, a semantic web, where fuzzy logic will power genuinely intelligent search, customer interaction and much more.

Put simply, we make statistically sensible deductions based on multiple bits of data. The multi-signal statistical stuff is necessary, because computers, by and large, still suck at interpreting natural language, particularly the kind of culturally diverse and informal language your followers use on Twitter.

Here we provide some insight into this process and the challenges involved, with reference to the primary data signals for our analysis...

Working with location is bread and butter for us at schmap, as you might have guessed from our name. While some users on Twitter have chosen to geo enable their accounts (add longitude and latitude to their tweets), we rely primarily on an interpretation of your followers’ stated locations to place them by country, state and city or town.

Our main challenge here is that there are many colloquial ways folk use to indicate their location on Twitter. Here’s a sample to illustrate the diversity:

Multi-cultural semantics aside, other challenges include resolving national and international ambiguity. For example, if one of your followers writes “Birmingham”, is that person in the US or the UK? Our fuzzy set theory approach adapts nicely for cases like this, combining multiple signals to disambiguate.

Continuing in fuzzy vein, we combine knowledge gleaned from the names of your followers with other data signals to give you an accurate assessment of gender. Specifically, for some 100,000 different given names, name shortenings, nicknames and appellations commonly used on Twitter, we have established the percentage chance country by country and language by language as to whether the name implies the user is male, female, or neither (i.e., a company, brand or organization of some kind, rather than a person). For example, if one of your followers live in the US, uses English as a native language and goes by the given name “Tracy”, absent any other data signals, there’s an 87.2% chance that this person is female. Combining this knowledge with other signals further improves our assessment.

The bios of your Twitter followers are a treasure trove of information, revealing professions, likes and interests, family status and more. Our challenge is that for any given interest, profession etc. there are countless ways to indicate this in natural language. A diehard baseball fan, for instance, might write “baseball is my life”, or “I’m a baseball junkie”, or “- ボク、野球が大好きです” (it’s popular in Japan too), or “I ? Joey Votto”, or "Go Giants!", or ... the list goes on. To index every common permutation across languages and cultures for all likes, interests and professions would dim the lights in a major city. We therefore first convert natural English (or Spanish, Japanese etc.) into less expressive coded renderings, before crunching these semantically. Combining this knowledge with other data signals then gives us a probabilistic assessment of the raw facts of your followers’ lives.

While the Twitter bio is a great profiling resource, we have to recognize it as a social persona, written by someone who is consciously trying to present him/herself with a few carefully chosen words. The personas we present online are typically tailored to a perceived audience: for instance, our Twitter bios read very differently from our profiles on Linkedin. At schmap, we aim to get behind the persona, in part by looking at the accounts that your followers follow. For example, a Twitter bio might read "I'm a loving husband, churchgoer and charity worker". When we look at the accounts this man follows, we find they include @KendraWilkinson, @WWE and @Eminem, giving us a more rounded picture of him as an individual.

So, how accurate is all this? When making an assumption based on a given individual signal for an individual follower, this has a quantifiable and often not insignificant chance of being wrong. But when we take multiple signals for each follower, then extend this process over thousands (or even just hundreds) of followers, the aggregate analysis stands up very well indeed.

And while you’re tweeting away adding new followers, we’ll be continually refining our fuzzy logic... Please do come back and refresh your analysis from time to time.

Schmap is a location technology service provider and local publisher, with cutting-edge expertise at the intersection of local, social, commercial and the real-time web. We're best known for a suite of popular Twitter services and for our series of real-time city guides:

"Know your Twitter followers" is our most recent addition to a suite of services that support international brands, local merchants and event promoters as they understand and connect with local audiences on Twitter. These services include analysis, powerful marketing tools and targeted promotion to 450,000 local consumers following Schmap-branded local Twitter accounts throughout 200 wordwide cities.

We released our first Schmap City Guides in March 2006: initially revised twice a year, these guides are now updated real-time, with local buzz for restaurants and bars from Twitter users, reviews from local residents, local photos from more than 300,000 contributing photographers, plus events, activities and local deals via users of our popular Twitter service.

Schmap was founded in 2004: we're based in Carrboro, North Carolina, with a development office in Beijing, China. If you'd like to know more, please do send us an email, or tweet us @schmap.