New Google Service Will Decide For You

Google is very good at figuring out what text is about. It can pick out ads that are relevant to the content of any Web page, its search engine anticipates the meaning of search terms rather than just finding them in Web sites, and Google's Chrome browser can tell when you're visiting a site in a foreign language and helpfully offer to translate.

Now some of that language-processing power is available to everyone in the form of an API that developers can use to have Google distinguish between different categories of text. The service is only available to registered testers and it requires some coding to use, so I've built a modest, easy-to-use demonstration here.

When a user uploads training text along with labels to the Prediction API, Google's servers search for patterns in the text and assume that they're linked to the label. For instance, it might find that text labeled as English includes lots of instances of the word "the," and text labeled as German includes lots of "der." Later, when Google is asked to figure out whether a blob of text is in English or German, it looks for those telltale "the"s and "der"s to help it come to a conclusion

Here is Jon Bruner demo of separating liberal from conservative opinions

(Gregory PS: In my informal test, passages from Krugman blog "Conscience of a liberal" and from Sarah Palin Facebook page were both scored as liberal by this API)

Read more.