Brain over brawn: Semantic technology and machine learning take new role in managing finance data

Pairing semantic technology with machine-learning-with-human-intervention enables the company to improve accuracy of data with minimal human impact. The result is 99.9 percent accuracy rate, said Drew Warren, Recognos Financial CEO in an interview, published on Finance Technology Weekly.

 
The finance industry has more than its share of data-intensive processes. But at least one technology firm is making the case that the answer to these challenges is not the traditional approach of ramping up data processing power – it’s smarter data processing.

With a semantic technology-based platform, Recognos Financial is taking a different approach to data processing and applying it to financial processes as wide-ranging as securities master management, onboarding and KYC compliance.

In mutual funds, for example, the company has assembled a securities master file for the mutual fund industry using a variety of tools from semantic technology to machine learning. As an Edgar distributor, Recognos has all mutual fund prospectuses and amendments, and it uses a combination of technologies to extract data from those unstructured data sources.

“Our platform uses about six different methods of data extraction. At the top of the list and the most complex is semantics, but we also use NLP, regular expression, machine learning and machine learning with human intervention,” said Drew Warren, Recognos CEO.

Warren estimates there are approximately 250 data points the company pulls from prospectuses and amendments to create its mutual fund securities master files. Semantic data aids the process because of three main benefits: its capabilities to structure unstructured data, integrate data very quickly, and deeply mine data for relationships. That last aspect is aided by the way semantic technology stores data – in triples, meaning data is stored in context, making it easier to align with other data points of similar contexts.

Pairing semantic technology with machine-learning-with-human-intervention enables the company to improve accuracy of data with minimal human impact. If a piece of data were incorrect, a user to the system could drag-and-drop the correct data in the appropriate field instead. (Data is only lassoed, or dragged-and-dropped, rather than rekeyed, as Recognos maintains that typing data introduces potential errors.)

Furthermore, the process of correcting a field by dragging and dropping a new value into it triggers a new script, ensuring that the new type of data will always be collected for the corrected field moving forward. The result is 99.9 percent accuracy of its mutual fund securities master data file, Warren said, with significantly less manpower.

“We have 12 data analyst that work on this, and that covers all the mutual funds issues in the U.S.,” Warren said. By contrast, a competitor has approximately 300 people working on mutual fund securities master, with lower accuracy results.

In know-your-customer (KYC) compliance, information that customers supply as part of the compliance process is verified against unstructured data pulled from supporting documents.

“Our technology is being used to process those supporting documents to verify information about an entity,” Warren said, adding that a wide range of documents are used to support this process, from articles of incorporation to tax documents, contracts and more.

With onboarding, a key challenge, particularly for a firm that is onboarding a group of clients on behalf of another provider, is that the old provider may have used different data formats than the new one.

“A lot of the data in onboarding is already structured, but it’s going to be coming off the system of one company, and going to the system of another,” Warren said. “You have to normalize the data from the output of one to the input of another.”

The ability of semantic data to store in context and mine more deeply can be used beyond the onboarding process, through the lifecycle of the client as well, Warren said.

“Using this technology, you can read all of the notes and documents for a given client across the entire platform, and learn where problems lie, what kinds of problems clients are having and delve much deeper into satisfaction levels to determine where you need to focus your efforts over time,” he said.

 
Brain over brawn: Semantic technology and machine learning take new role in managing finance data
Published in Finance Technology Weekly
December 3, 2015
By Renee Caruthers

 
http://www.fiercefinanceit.com/story/brain-over-brawn-semantic-technology-and-machine-learning-take-new-role-man/2015-12-03