Anyone who’s used Twitter will be familiar with the concept of Trending Topics – those hash-tags and phrases that appear on the right side of the screen, letting you know that, for one reason or another, they’re what’s popular in the world of Twitter at any one point in time.
OneRiot, a real-time search engine, likes the idea of offering trending topics to users, but they first explained the need for a new system by examining the flaws with Twitter’s. According to the official OneRiot blog, Twitter’s trending topics setup fails due to three main reasons:
- A Trending Topics set that rarely deviates from a list of mass-culture popular phrases (aka the Justin Bieber effect)
- A list that’s skewed by the particular demographics of a single service (aka the Apple fan Boy effect)
- A list that is lagging behind what people are really interested in right now. This is especially true if a list is created based on search query volume. In the main, search query volume will spike because people already know about the topic.
The OneRiot Trending Topics Engine is set to change all of that, by analyzing user data from not only Twitter, but MySpace, Facebook statuses, Digg, and its own toolbar. Using “natural language processing techniques,” the system will pick out words or phrases that are appearing in mass amounts, and will remove the spam and adult content from these potential trending topics before working with them further.
Terms can be grouped into like categories, in the case of a natural disaster or political event, as examples. In the case of the January earthquakes in Haiti, the terms “Haiti” and “earthquake” would’ve been combined, to create a more all-encompassing topic.
As OneRiot says, the system creates a sort of central hub of one particular topic, and then creates a web surrounding it of all things that are important and related. In the case of the recent “Gizmodo iPhone leak” story, the terms “Gizmodo” and “iPhone” appear in the central hub, while stories concerning the police raids, and general searches for terms like “iPhone prototype” would appear as tendrils branching out from this center.
Using this new system, OneRiot’s partners will receive different trending topics based on their own particular needs, and users will be able to see just why a particular topic is trending, rather than just learning that it is.