Version 0.1, November 15, 2009.
- This service detects the product central to a page.
- Products currently include cameras, phones and other gadgets.
- More categories to be deployed: video games, media, ...
Interpreting the resulting JSON record:
For general topic classification see http://topics.speedi.ly/.
- language: the main language in the document.
Language detection covers main European languages, CJK, Russian, Thai...
No entity is detected when the language is not English.
- nsfw: number of distinct / total offensive terms.
- sentiment: happiness score from -1 (Zune review) to +1 (iPhone launch).
- product: the name of the product we believe is the most central to the page.
The name is followed by a confidence level (0 to 1) that the product is central to the story.
- categories: from most precise (say Phone or Camera) up to the top (Product).