The software tools that convert news into machine-readable data put numbers on, among other things, sentiments about companies mentioned in news stories and quick assessments on how the news will affect the firms’ prospects in the markets they operate in. This gives traders quantifiable values on which they can make better and often instantaneous decisions.
In the example here, Thomson Reuters’ NewsScope Analytics uses a natural language processor (NLP) to turn human langauge into terms and values a computer can understand and process. Other players in the nascent machine-readable data category are Psydex Corp., Dow Jones & Co. and a new entry Selerity, Inc. Consider that there’s about a zetabyte of new information every year. That’s a 1 followed by 21 zeros ... and 80 per cent of that data is unstructured, according to Rob Usey, CEO and founder of Psydex.
NewsScope Analytics breaks down stories with unstructured data where there are few or no inherent values (slide 3). Other NewsScope products handle stories with structured and semi-structured data (slides 1, 2).
NewsScope Analytics starts out by identifying key sentiment words (slide 3) and assigns values to them which correlate to short and long term prices changes (slides 4, 5, 6).
Average speed for analyzing unstructured stories for things like sentiment, relevance, and novelty is 150 milliseconds, according to Richard Brown, Thomson Reuters Global Business Manager, Machine Readable News.
“The natural language processor parses and categorizes text which produces metadata regarding any number of factors including item relevance, whether it’s substantive for the company, repetition and tone,” Brown explains. NewsScope Analytics can assign values in up to 41 categories.
How would a trader or machine act upon the assigned values?
“If it passes the relevance smell test, it might move into looking at the tone. If it’s highly positive you or the machine might decide to buy,” Brown explains.
Thomson Reuters’ has only “a few dozen” customers for NewsScope Analytics so far, but Brown says software that converts news to actionable digits is used by both high-frequency and long-term traders.
“Wide adoption is low because of the price tag and people’s [lack of] understanding of what to do with the data and how to interpret unstructured data signals,” he says. The monthly cost is typically “$15,000-$50,000” for real-time access and less for overnight updates, all depending on how many companies client wants to follow.
This article can also be found at SecuritiesIndustry.com.
John Dodge is a contributing author for Securities Industry News.
Be the first to comment on this post using the section below.