Thanks to Curt Monash for calling attention to an article by Seth Grimes about a presentation that Richard Brown of Thomson Reuters made at Gartner’s Event Processing Summit.
Brown’s talk was about the Reuters NewsScope Sentiment Engine, which “processes a stream of Reuters English language news items, producing sentiment data for a list of customer determined target companies.” The list of collaborators is intruiging:
- Technology from Infonic to assess author sentiment.
- Stream processing from Streambase.
- Quantitative research from AlphaSimplex, a firm chaired and founded by Andrew W. Lo, director of MIT’s Laboratory for Financial Engineering.
- And of course Reuters’ own Calais Web service.
I have no idea how valuabe the results are, but I intrigued at the ambition of putting all of these pieces together to support financial-market trading, Given the rocky history of quantitative trading, I’m curious to hear about their results.
The Noisy Channel 
3 replies on “Event Processing Meets Text”
Hello Daniel. there’s one other piece: ClearForest technology, installed client side (as opposed to server-side as part of NewScope) to provide information-extraction capabilities.
As for the results, these are described in the Reuters white paper I mentioned, which you can obtain from Richard W. Brown of Thomson Reuters. I’m reluctant to post his e-mail address, but it combines his and his company’s full names. If you can’t figure it out, let me know and I’ll provide it. Reuters and AlphaSimplex applied difference-of-means testing and got statistically significant t-test scores to establish an effect of news-sentiment-scoring. This is my recollection from Tuesday’s presentation and a very quick scan through the paper, which I haven’t had a chance to read at all closely. I would’ve pasted in a chart and some quotations in my blog article but Rich asked me not to use the materials directly.
LikeLike
Seth, thanks for the clarification. I’d assumed that all of the ClearForest technology was now part of Calais. I’ll follow up with Rich and hopefully learn more.
LikeLike
And another correction from Seth: Reuters is encouraging information publishers to use Calais to tag the content they publish, and they’re deploying ClearForest client side for extraction, but this isn’t a core part of the solution.
LikeLike