Asset managers seeking an edge in the uncertainty of the pandemic might do well to turn to natural language processing the way firms including American Century Investments do — and avoid the task of digesting massive volumes of research, according to text analytics company FinText.
In a recent paper looking at the ways finance firms uses the machine learning application, FinText said American Century tries to detect deception in management language during companies’ quarterly-earnings calls. Its sentiment model checks for omission of important disclosures, spin, obfuscation, and blame.
Firms such as Barings Asset Management, State Street Corp., and Deutsche Bank are also using natural language processing, according to the paper. The technology removes “text-related grunt work, allowing employees to focus on higher-value tasks,” FinText said in the paper.
Barings researchers have developed a system that helps digest media content, including financial news, blogging, and tweeting, according to FinText. The system identifies companies tracked by the asset manager, producing a “sentiment score” for the coverage they’ve received.
“Especially during earnings season, when news flow becomes particularly voluminous, ingesting all this information becomes a challenge,” the paper said. “Barings sought to develop an internal solution to help consume information at scale and augment the investment research process.”
The sentiment model American Century developed considers “both the unique style of a given management team and the collective language of its industry peers,” according to FinText. “To develop the model, the researchers first turned to financial journals and psychology texts to identify linguistic patterns associated with deception.”
Spin may be detected through management’s exaggeration and “overly scripted language,” while “complicated storytelling” may signal obfuscation during company earnings calls, the report said.
FinText found that State Street has estimated the vast amount of material produced by top financial research teams can consume as many as 12,000 sheets of paper a day, overwhelming analysts. “With limited time, investors may overlook or miss important research insights,” the paper said. “In this environment, making sense of information quickly can become a competitive edge.”
State Street’s Quantextual Ideas Lab created research aggregation software to review lengthy research reports and classify documents. The algorithms “condense the research into shorter snippets, while preserving their nuanced academic or economic tone.”
Deutsche Bank uses natural language processing to predict companies’ sustainability performance with respect to carbon emissions, according to the paper.
The bank found that companies using “highly active and numeric language” have an average 74 percent chance of lowering them, FinText said, while those that “frequently discuss mitigating or adapting to climate change have a 65 percent higher chance to achieve reductions.”