Here’s a relatively new web service for publishers and developers that’s a different spin on the semantic web. Amplify is mostly a semantic web service, although it tries to differentiate itself from other semantic services by saying it focuses on understanding content rather than classifying content — which is what the semantic web has been all about up until now. It’s an interesting spin on things. As an aside, perhaps you could argue they’re indulging in some semantic footwork of their own, because in a sense they’re really just classifying content too.
It’s fascinating stuff — and the real power of the service will emerge as more sites sign on and build applications that use their service. At the moment they provide a free API (for up to 1,000 requests per day), thereafter you would negotiate something further with them.
It works like this: You feed content through Amplify, it analyses the content and returns back a set of “signals” in structured XML. These signals take the form of:
- The Topic Analysis: returns topical Signals about the text, including polarity (positive/negative perception of each topic) and guidance (degree to which guidance is sought or offered about each topic). Topics may be related to each other, and Amplify will attempt to detect those relationships (for example, that pizza and salad are both foods) and report them, ranked by likelihood. In addition, proper names and referenced locations are returned by the topic analysis.
- The Action Analysis: returns Signals related to actions detected in the text. Associated with each action will be a measure of decisiveness (how likely the action is to be taken), guidance (whether guidance is sought or offered on taking the action) and temporality (when the action may take place).
- The Style Analysis: returns stylistic Signals about the text. Initially, this analysis will return flamboyancy (a measure of how “flowery” the writing style is), and slang (degree to which slang vocabulary is used).
- The Demographic Analysis: returns the likely age, gender and education level Signals of the text’s author or audience.
There are some interesting applications. For example, network online advertisers could go further than merely identifying keywords related to their ads — but ajudge whether those keywords are being mentioned in a positive or negative context. (This may then affect a decision on whether to show the ad or not). It would also allow an advertiser to make some assumptions about the type of content and quality of content, together with its intended target market, to ensure more accurate placement. (Do I sense a Google acquisition here??)
It would also help aggregators filter and classify content to deliver personalised, relevant content to users (based on your age for example). Online publishers that rely heavily on user generated content communities, could also use the service to help classify and monitor content on their sites, as well as monetise the content with more accurate ad placement. It could be potentially used to analyse Twitter feeds to get a more accurate sense of whether negative or positive things are being said about a brand, site or person. Here are some thoughts on more applications, can you think of any more?