Alexandria Builds Relevancy Indicators into Sentiment Scores

eugene-shirley-alexandria

Currently, Alexandria's proprietary Contextual Text Analytics engine (ACTA) generates either positive, negative or neutral classifications for articles published by Dow Jones Newswires' feeds, after striking a deal to provide sentiment scores on the news provider's content in February 2012.

Next month, Alexandria will add a new classifier to its assessments that assigns a relevancy score to each company mentioned in an article, says Alexandria chief executive and co-founder Eugene Shirley.

"If there are three companies mentioned in article and the sentiment is primary about one company, but... it also has an effect on the other companies that are mentioned, the relevancy score would say "This is a positive article, and it is primarily positive about this company but it does have some positive effect on these others,'" Shirley says.

The vendor generates the scores in real time and delivers them to trading systems in as little as 20 to 30 milliseconds, or as an end-of day-file to less latency-sensitive customers such as long-term investors. Alexandria is exploring opportunities to integrate the sentiment within various displays, though the vendor's primary focus is serving quantitative investors, Shirley adds.

The algorithm used to calculate the sentiment was developed by Alexandria's chief scientist Ruey-Lung Hsiao, whose previous work in academia involved extracting critical intelligence from genomic data, which the firm is now applying to news and unstructured data to serve financial services firms.

The methodology─which delivers around 90 percent accuracy, according to internal testing─draws on an observation-based approach that "teaches" the engine to create its own rules, rather than traditional word-based or rule-based approaches.

"Word-based sentiment gives a sentiment assignment to a word like ‘upwards' or ‘downwards,' then counts up the number of positive and negative words to assign sentiment─but it's based on the fallacy of having a perfect dictionary that contains all nuances of language," Shirley says. "Other systems have been built based on writing rules, but that's based on the fallacy of a perfect rule set. Rather than create rules, we have experts in the field annotate articles in finance and feed these training examples into the system, and the system creates its own rules, like training a child."

Alexandria's core focus is on English-language news from Dow Jones covering all asset classes, but─given the approach of its methodology─Shirley says the vendor is also able to provide sentiment from feeds in other languages, including Japanese-language real-time newswires such as the Nikkei/Dow Jones Japan Report produced by Nikkei in partnership with Dow Jones.

In addition, the vendor is also looking at the potential to apply its sentiment methodology to social media and other forms of unstructured data in future.

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@waterstechnology.com or view our subscription options here: http://subscriptions.waterstechnology.com/subscribe

You are currently unable to copy this content. Please contact info@waterstechnology.com to find out more.

Most read articles loading...

You need to sign in to use this feature. If you don’t have a WatersTechnology account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here