Unlocking business value from Named Entity Recognition

Unlocking business value from Named Entity Recognition
FeedStock

FeedStock
20/05/2019

“Data mastery”, just like “the Cloud” and “Artificial Intelligence” has been added to the growing list of buzzwords in the fintech space. Of the small number of experts who do understand these terms, an even smaller number have an idea of how they could be applied in daily corporate life to effect some of the “revolutionary change” which they have been touted for. This Deloitte Insight report entitled “Pivoting to digital maturity” which outlines the seven capabilities central to digital transformation, emphasises the importance of data mastery as a key driver in an organisation’s progress towards digital maturity. Deloitte defines data mastery as “aggregating, activating, and monetizing siloed, underutilised data by embedding it into products, services, and operations to increase efficiency, revenue growth, and customer engagement.” [sic]. In this BrightTALK webinar Principal Analyst at Forrester Research Michele Goetz addresses the potential for growth and added value through data mastery. Goetz talks about the need to transition from traditional Master Data Management (MDM) systems in which enterprises focus on solidifying a logical data model, toward viewing an MDM system as an “intelligent and contextual engine” to drive better business decisions. Both Deloitte Insights and Forrester point to the potential of high quality MDM systems in an enterprise’s ability to optimise revenues. At FeedStock we believe that one key way in which a successful MDM can be used to deliver greater business intelligence is through Named Entity Recognition (NER). NER, also known as Entity Extraction, is when an algorithm is trained to locate named entity mentions in a body of plain text and classify them into pre-defined categories such as person names, organisations, locations, percentages or monetary values. Below is an article which has been labelled by an NER algorithm trained to classify countries (GPE), organisations (ORG), cardinal numbers (CARDINAL) and ordinal numbers (ORDINAL).

This example demonstrates the potential to extract key information from large bodies of text to deliver succinct and actionable business intelligence.

The use of NER in business today

NER adds a wealth of semantic understanding to any large body of text and has multiple business use-cases such as classifying and prioritising news content for newspapers or generating candidate short-lists from a large number of CVs for recruiters. The use of NER technology when integrated into email and chat systems could enable you to extract and collate information from large amounts of documentation across multiple communication channels in a much more streamlined, efficient manner. NER would enable you to instantly view trending topics, companies or stock tickers and provide you with a full overview of all your information channels containing relevant content such as meeting notes shared via email or daily discussions over chat systems. In a world where business managers can send and receive thousands of emails per day, removing the noise and discovering true value in relevant content will be the difference between success and failure. There are a number of popular NER libraries available today including Explosion AI’s SpaCey, Stanford NER and Natural Language Toolkit (NLTK). The different specifications of those three libraries means that they will each be suited to slightly different use cases. NLTK offers broad functionality and ease of use, SpaCey is optimised for ‘blazing fast’ performance but for state-of-the-art accuracy levels, Stanford NER is deemed one of the best. However, regardless of that fact, the precision, accuracy and recall of an NER algorithm will rely largely on whether it has been trained using pre-labelled texts which are similar in context to its end use-case. Since an NER algorithm forms an understanding of entities through grammar, word positioning and context, this final element is of crucial importance and if omitted, can result in poor accuracy scores.

Hidden business intelligence through NER

Like Deloitte and Forrester, we believe that the potential for workflow efficiencies and revenue growth is limitless for enterprises with the correct approach to data mastery. FeedStock is committed to revealing the hidden business intelligence in our clients’ interaction data through providing world-class MDM systems. Through the deployment of highly trained NER algorithms we can automatically and effortlessly distil and synthesize the information from across all your communication channels. In an increasingly competitive business environment with ever decreasing margins, the ability to uncover actionable insights from all your existing data streams is vital for enterprise success. Position your business to win in this era of rapid technological advancement with FeedStock.

Back