OneRiot’s Trending Topics Engine leverages the company’s patented PulseRank technology and realtime index of the social web. It finds trending topics before anyone else because it looks at activity across wider social web and employs realtime machine learning to tune Trending Topics on the fly.
The OneRiot Trending Topics Engine delivers trending topics in realtime by employing a number of techniques:
1) Analyze the entire stream ingested from Twitter, Facebook, MySpace, Digg, and OneRiot’s own panel of users. This provides a raw picture of whatâ€™s resonating with people right now across the web.
2) Using natural language processing techniques, OneRiot extracts â€œinterestingâ€ and â€œemergingâ€ phrases, speech snippets and text blocks, which get tagged as â€œTrending Topic Candidates.â€
3) OneRiot then computes various weights for each candidate. Some of these are traditional Information Retrieval measurements (such as TFIDF Term Frequencyâ€“Inverse Document Frequency) and some refer specifically to the modern social web (such as viral velocity on Twitter).
4) Employ light filtering to reduce the data set - removing bad candidates.
5) Removing spam and porn. Apply filters to candidates with zero viral velocity.
6) Clustering. Specifically, OneRiot clusters together correlated candidates. For example, a cluster includes a number of words that appear together in text and tweets.
7) Compute weights for each cluster (which are now story-centered). This process inherits and aggregates the features of candidates inside each cluster. OneRiot can quickly capture if a topic is emerging by looking at factors like the rate of acceleration of the term across the social web.
8) Rank the topics utilizing OneRiot’s PulseRank algorithm. Topics can also be categorized.
There are a number of highly valuable uses for OneRiot’s Trending Topics technology including Content Discovery, Informing Editorial, Target Trending Ads, and Realtime Triggers for Mainstream Search.