What are your ideas about promoting a new product?

What initiatives are there in building a social graph representing the history of ideas?

• There needs to be a project on the scale of what has done for geographic mapping. Are there any systems already in place that allow one to browse idea / actor networks?  Something coming from the skunkworks of , , , etc.? Are there any web-sites or academic initiatives attempting to create a comprehensive map of the actors, relationships, and subjects involved in the history of ideas? Which of the following strategies are (or could be) most useful? Creating and promoting small, standardized data models for representation and exchange. from  , , the . people who are interested in the . Modifying systems. Repurposing systems designed for (or collaborating with) other social data-gathering initiatives - http://Intelligence%20Agencies,, , etc.

• Answer:

  I am currently doing a data mining project to reconstruct graph of influence in different topics in biology based on open access publications. There are two large problems to this, at least in my field:  - Map a flat text file to a concept / idea map, with a hierarchical tree of topics  and ideas exposed (the kind you would draw using IHMC cmaps). For every text, the generating rules (in terms of formal grammars on words) depend on the domain, journal, and the person who is writing (choice of some words over others). In Biology, texts are highly structured and generally follow a defined outline, which makes the task a little bit easier, but still far from trivial.  - Get a normalized representation of ideas. All concepts are strictly speaking specific to the individuals and are strongly influenced by their own experiences and context of idea/concept learning. To be simple, if you are a city child, when you think about concept "donkey", you most likely would picture a sad companion of Winnie the Pooh, whereas if you grew up in a hilly areas of middle east, this will be your uncle's stinky and noisy gray barrel on feet. In my domain it means that if you try to compare texts written by the immunologists v.s. developmental biologists, you're going to have a very, very bad time, since the same words represent different ideas and inversely, same ideas are represented by different words. I am not 100% sure, but I would say that these two problems get but worse with more general texts. Drop me a message if you're up for a more detailed discussion. As far as I know, for biggies, Google is building a "knowledge graph", and is most likely analyzing the data it pulls out from the corpus of the books it has scanned. If you haven't yet, get a view at http://books.google.com/ngrams. Google programmers are generally very good and have time for side projects, so I can't imagine that none of them went into pulling more information from the books they've scanned. Update: after looking at the http://www.aaai.org/Magazine/Watson/watson.php behind the IBM's Watson, it seems that they have a deep enough learning base to be able to carry over such a project. However, I am not sure they would do it, for it represents no business opportunity for them right now.