Smarter taxonomy = better insight

Taxonomy is the science of the classification of things and when implemented effectively it helps us to retrieve relevant information in a timely manner.

The Global Industry Classification Standard (GICS) and Industry Classification Benchmark (ICB) are an industry taxonomy developed for use by the financial community, when classifying companies. These Sector classifications are probably the best known taxonomy in the financial field.

However as trading is all about relevancy and contextual information, Heckyl decided that a new level of taxonomy was needed to analyse data, be it news, events or companies on a more granular level.

As highlighted in our last blog data explosion creates problems with regard to the gathering and processing of information, but at the same time enormous opportunities too. But when talking about these data sets how do you take into account the idiosyncrasies of the English language?

An event driven Hedge Fund might be keenly interested in M&A, unfortunately the journalists of the world do not stick to one agreed generic term when writing about it. Therefore our taxonomy is based on machine learning and not just keyword matching. In this instance our system has its own dictionary and parts of speech to consider for news to be tagged as M&A. These include: Merger, Acquisition, Take Over, Buy Out, Hostile, T/O (commonly used on Twitter) and many more.

A combination of 32 different words and business rules are used to classify the underlying event. These keywords are also Sector specific; ‘well’ in the Oil and Gas sector will relate to the discovery of a well, a very different meaning to ‘well’ in the Pharmaceutical sector, which relates to the efficacy of a drug.

Every event brings a potential opportunity and based on five years of backtesting Heckyl has mapped every industry, share movement and associated news. Being able to recall relevant data, from product launches to tax inversion plans quickly, is only possible due to a taxonomy that effectively classifies and filters these events to a very granular level.

A view of relevant news based on a users portfolio segmented by type, using Heckyl's proprietary taxonomy.