Thomson Financial and Uptick Data have announced a partnership, leveraging text mining tools from Uptick to extract business and financial events from news.
Thomson and other financial content providers have experimented with text mining for the past few years. Most of the early efforts have focused on automating (or semi-automating) the manual tagging process, with the primary focus on cost-cutting and scaling and managing a workflow that often varies with filing and market cycles. For many, text mining has been viewed as an alternative to outsourcing. This partnership is focused on leveraging the Uptick technology to trigger news alerts based on business events, a potentially more lucrative endeavor.
Content providers can leverage text mining to generate new revenues through creation of new products, micro-chunking existing content, developing alerts and more. Sophisticated providers can then add analytics to the extracted information to create higher value solutions.
Conversely, one of the challenges to automated tagging is the level of accuracy. The absolutely best systems today achieve a break-even score (where precision=recall) in the 80-90% range. Even at the top of the range, you're still getting one out of ten wrong. And, unlike humans, when the engine gets it wrong it often looks ridiculous. An example I often used in my days at ClearForest was that, due to the vagaries of language, you might see an event tagged as "Graubart acquires Starbucks", when the underlying story merely said that I bought a cup of coffee.
It will be interesting to see where Thomson and Uptick take this partnership. And, it will be even more interesting to see how accepting the financial markets users are of the imperfect results.
Posted by: |