Newsflow for Robots

An Australian design for a “virtual institution” in which autonomous automated agents negotiate deals on behalf of their human users.

Making computers newsreaders (London Times): Reuters invests in “machine-readable” news so that trading algorithms can react to newsflow.

What, I wonder, will it be like to write “news” for an audience of machines?

PLANS by Reuters to ensure a healthy future in the next ten years include teaching computers how to read news stories in order to help them trade more efficiently.

The company is under pressure to capitalise on the growth in “algorithmic trading”, where computers buy and sell commodities according to pre-set rules. Algorithimic [sic] trading generates huge volumes, but does not add to the numbers of terminal subscriptions — traditionally the way that Reuters earned its money.

Reuters is trying to find ways of charging per transaction, and making its news valuable to computers by creating what it calls “machine readable news”. The idea is that Reuters stories are broken down and presented in a consistent way for computers to understand. The activity helped to generate an estimated £2 million in the half.

“The idea is that you could analyse whether the coverage of a company has become more positive in the last week or so, and perhaps you could generate a trade on that basis,” Tom Glocer, chief executive, said.

See also News-to-data conversion (June 8).

Perhaps I am short on the “vision” thing, but philosophically, I just can’t make this compute. Can you?

Would you let a machine that can’t pass the Turing test trade for your portfolio? Let’s keep an eye peeled for what developers come up with by way of a bullshit-detection module for the system.

In a related, though less futuristic, development, rival Thomson Corp. (TOC) is teaming with ticker-farmer Up-Tick Data to “enhance” its news services with live market data.

Thomson Financial and UpTick Data Technologies Collaborate to Produce Enhanced News Service: Financial News – Yahoo! Finance:

SAN FRANCISCO, July 25 /PRNewswire/ — UpTick Data Technologies, a leader in customized communication solutions utilizing rules based text generation today announced an agreement with Thomson Financial to produce news stories built around Thomson market data. The multi-year agreement is aimed at developing new products and enhancements for Thomson News, the market news service launched by Thomson last year, utilizing proprietary Thomson data.ADVERTISEMENT
UpTick will utilize its proprietary text generation technology customized for specific rules and events that Thomson Financial analysts have determined to be important to Thomson News subscribers. “Using UpTick allows us to rapidly mine our vast repository of proprietary data and report on significant changes and anomalies as they happen in real-time,” said Matthew Burkley, Senior Vice President, Operations for Thomson. “Whether you are looking at earnings, insiders or M&A activity, there are events taking place every day which can only be uncovered through a comprehensive analysis of the data,” continued Mr. Burkley. “Some of course are obvious, but many more are very subtle and no less significant. Unique reporting like this is one of the differentiators of Thomson News.”