Venture-capital backed Prattle will begin offering a new service Monday that quantifies and scores the language used in earnings calls and reports filed by 3,000 U.S. public companies.
The company, which is backed by venture-capital firms including New Enterprise Associates and GCM Grosvenor, uses natural language processing and machine learning techniques to improve on current methods of analyzing text and sentiment. Traditional text analysis often uses a simple calculation of the number of positive and negative words in a report or transcript.
Analysts and fundamental portfolio managers are trying to shift from decision-making that relies on human subjectivity to a quantitative approach or a mix of both quant and fundamental analysis, according to Evan Schnidman, Prattle's chief executive officer.
More from Institutional Investor:
Younger, cheaper, smaller, smarter: Meet the new breed of quants
Booze, drugs & fistfights: Another summer in the Hamptons
A hacker's guide to destroying the global economy
"We're quantifying what wasn't quantifiable before," said Schnidman, a game theorist who taught at Brown University and Harvard University before founding Prattle.
After an earnings call, for example, Prattle provides a score, as well as the sentiment of every speaker. Schnidman expects its methodology to be highly accurate based on its experience distilling language used by central banks into quantitative sentiment data that projects policy outcomes. So far Prattle has predicted outcomes from the nuanced language of central bankers with 98 percent accuracy in the year and a half that it has offered the service.
Schnidman says the company is not looking for new patterns or investment insights in big data. Instead, Prattle is making it more efficient to mine proven sources of market-moving information by using advanced computing techniques.