Infochimps and the scaling of dataset value

Image via CrunchBaseSure, a picture is worth a thousand words, but what is a thousand words worth? How about a million? If I had a dataset of the most recent trillion words spoken by humanity, (anonymized and randomized of course!) would that be worth any more than the set of words in this blog post?

These are real questions. A Texas company called Infochimps has datasets quite similar to these, ready for you to use. Some of the datasets are free, others you have to pay for. More interesting is that if you have a dataset you think other people might be interested in, or even pay for, InfoChimps will host it for you and help you find customers. (Infochimps just announced they had raised $1.2 million in its first round of institutional funding.)

One of the datasets you can get from Infochimps for free is the set of smileys used on twitter in tweets sent between March 2006 and November 2009. It's free. It tells you that the smiley ":)" was used 13,458,831 times, while ";-}" was only used 1,822 times.

If you're willing to fork over $300, you can get a 160MB file conatining a month-by-month summary of all the hashtags, URLs and smiley's used on twitter during the same period. That dataset wil tell you that during September of 2009, the hashtag #kanyeisagayfish was used 11 times while #takekanyeinstead was used 141 times.

If you're a scrabble player, you can spend $4 for a list of the 113,809 official words, with definitions. Or you can get them free, without the definitions.

courtesy of Infochimps, Inc. CC-BY-A

I had a great talk with Infochimps President and Co-Founder Flip Kromer a few weeks ago before his presentation to the New York Data Visualization Meetup. I fell in love with one of the visualizations he showed in his presentation, and he's given me permission to reproduce it here. (Creative Commons Attribution License) It's derived from the same Twitter data set you can get from Infochimps, and shows networks of characters that are found in the same tweet. So if ♠ and ♣ appear in the same tweet over and over again, the two characters will have a strong connection in the network of characters.

The character connection data was fed to a program called Cytoscape, which is an open source visualization program used in bioinformatics; Mike Bergmann has a nice article about its use for large RDF graphs. The networks are laid out using a force-directed algorithm (which is pretty much the simplest thing you can do). Coloring is applied arbitrarily.

As you might expect, the main character networks that show up are associated with languages, but there are some anomalies. For example, the katakana character ツ (tu) sticks out. Katakana is a set of phonetic characters used in Japanese for non-Japanese words. The reason "tu" is set apart from all the other katakana is that people use it on Twitter as a smiley.

The other anomalous character subnet is labeled "???" in the graph. A closer look reveals this to be the set of characters that look like upside down roman text.

Kromer has noticed that the price (or perhaps cost) of a partial data set follows a non-monotonic curve (see graphic). Small amounts of data are essentially free, but a peak value is reached when portions of the data set are extracted from the full data set. If we were discussing book metadata, for example, peak value might accrue for a set of the 100,000 top selling books.

There's much less value, according to Kromer, in having a large incomplete chunk of a data set. Data for 10,000,000 books, for example, would have less value than the 100,000 book data set, because it's not complete. Complete data sets become extremely expensive because of the logistics involved, and because of the value of having the complete set.

This pattern seems plausible to me, but I'd like to see some clearer examples. I've previously written about having too much data, but that article looked at the effect of error rates on data collection; Kromer's curve is about utility.

For me, the most interesting thing about Infochimps is the idea that the best way to make data flow in large volumes and create new types of knowledge is to provide the right incentives for data producers through the establishment of a market. This makes a lot of sense to me; however I'm not sure that the Infochimps market has also established incentives needed for data set maintenance; the world's most valuable and expensive data sets are one that change rapidly.

Kromer contrasted the Infochimps approach to that of Wolfram, whose Alpha service is produced by "putting 100 PhDs and data in a lab". He also feels that much of the work being put into the semantic web is a "crock" because its technology stack solves problems that we don't have. Humans are pretty good at extracting meaning from data, given a good visualization.

We can even recognize upside-down text.

Article any source

Who's the Boss, Steinbrenner or Springsteen?

As I started playing with Mathematica when it first came out (too late for me to use it for the yucky path integrals in my dissertation), I just had to try Wolfram|Alpha. The vanity search didn't work; assuming that's what most people find, its probably the death knell for W|A as a search engine. Starting with something more appropriately nerdy, I asked W|A about "Star Trek"; it responded with facts about the new movie, and suggested some other movies I might mean, apparently unaware that there was a television show that preceded it. Looking for some subtlety with a deliberately ambiguous query, I asked about "House" and it responded "Assuming "House" is a unit | Use as a surname or a character or a book or a movie instead". My whole family is a big fan of Hugh Laurie, so I clicked on "character" and was very amused to see that to Wolfram|Alpha, the character "House" is Unicode character x2302, "⌂". Finally, not really expecting very much, I asked it about the Boss.

In New Jersey, where I live, there's only one person who is "The Boss", and that's Bruce Springsteen. If you leave off the "The", and you're also a Yankees fan, then maybe George Steinbrenner could be considered a possible answer, and Wolfram|Alpha gets it exactly right. Which is impressive, considering that somewhere inside Wolfram|Alpha is Mathematica crunching data. The hype around Wolfram|Alpha is that it runs on a huge set of "curated data", so this got me wondering what sort of curated dataset knows who "The Boss" really is. To me, "curated" implies that someone has studied and evaluated each component item and somehow I doubt that anyone at Wolfram has thought about the boss question

The Semantic Web community has been justifiably gushing about "Linked Data", and the linked datasets available are getting to be sizable. One of the biggest datasets is "DBpedia". DBpedia is a community effort to extract structured information from Wikipedia and to make this information available on the Web. According to its "about" page, the dataset describes 2.6 million "things", and is currently comprised of 274 million RDF triples. It may well be that Wolfram Alpha has consumed this dataset and entered facts about Bruce Springsteen into its "curated data" set. (The Wikimedia Foundation is listed as a reference on its "the boss" page.) If you look at Bruce's Wikipedia page, you'll see that "The Boss" is included as the "Alias" entry in the structured information block that you see if you pull up the "edit this page" tab, so the scenario seems plausible.

Still, you have to wonder how any machine can consume lots of data and make good judgments about who is "The Boss". Wikipedia's "Boss" disambiguation page lists 74 different interpretations of "The Boss". Open Data's Uriburner has 1327 records for "the boss" (1676 triples, 1660 properties), but I can't find the Alias relationship to Bruce Springsteen. How can Wolfram|Alpha, or indeed any agent trying to make sense of the web of Linked Data, deal with this ever-increasing flood of data?

Two weeks ago, I had the fortune to spend some time with Atanas Kiryakov, the CEO of Ontotext, a Bulgarian company that is a leading developer of core semantic technology. Their product OWLIM is claimed to be "the fastest and most scalable RDF database with OWL inference", and I don't doubt it, considering the depth of understanding that Mr. Kiryakov displayed. I'll write more about what I learned from him, but for the moment I'll just focus on a few bits I learned about how semantic databases work. The core of any RDF- based database is a triple-store; this might be implemented as a single huge 3 column data table in a conventional database management software; I'm not sure exactly what OWLIM does, but it can handle a billion triples without much fuss. When a new triple is added to the triple store, the semantic database also does "inference". In other words, it looks at all the data schemas related to the new triple, and from them, it tries to infer all the additional triples implied by the new triple. So if you were to add a triples ("I", "am a fan of", "my dog") and ("my dog", "is also known as", "the Boss"), then a semantic database will add these triples, and depending on the knowledge model used, it might also add a triple for ("I", "am a fan of", "the Boss"). If the data base has also consumed "is a fan of" data for millions of other people, then it might be able to figure out with a single query that Bruce Springsteen, with a million fans, is a better answer to the question "Who is known as 'the Boss'" than your dog, who, though very friendly, has only one fan.

As you can imagine, a poorly designed data schema can result in explosions of data triples. For example, you would not want your knowledge model to support a property such as "likes the same music" because then the semantic database would have to add a triple for every pair of persons that like the same music- if a million people liked Bruce Springsteen's music, you would need a trillion triples to support the "likes the same music" property. So part of the answer to my question about how software agents can make sense of linked data floods is that they need to have well thought-out knowledge models. Perhaps that's what Wolfram means when they talk about "curated datasets".
Article any source