9 February 2021
In this article, I’d like to discuss an idea based on some spurious philosophy, originally attributed to the Greek poet Archilochus and then popularised by Isaiah Berlin in his essay “The Hedgehog and the Fox”. “The main principle being that the difference between foxes and hedgehogs, is that foxes know many small things, while hedgehogs know only one big thing.” He probably wasn’t expecting the idea to be evoked as testimony in support of a technology debate raging across the banking community. In fact Berlin said “I never meant it very seriously. I meant it as a kind of enjoyable intellectual game”.
In 2011 I was a self-confessed fox, running a technology project to decipher complex contracts with a myriad of rules (knowing many small things). The Likezero team spent months writing and refining rules to gain a strong understanding of a single type of contract and we did in fact get pretty good at it. By 2013, we were still polishing these rules and were improved but still some way from perfection. The truth is that however many rules you write there will always be bad results (false positives) in your data sets, so checking everything is essential. And when you check results and complete missing data you must do this one document at a time, both handicaps challenge all “foxes”.
Sometime in 2017, the team and I had a revelation and now we’ve proudly become hedgehogs in our thinking and processes – hedgehogs know one big thing, which infuriates foxes. Our single big idea was that rather than trying to spot differences in contracts with rules, to capture data (which is how machine learning technologies work) we would stand the problem on its head and match things that are the same. In our initial prototyping we had some startling results, which gave us the confidence to invest. As we started to acquire clients, we found that data mining and matching can solve the important and unsolved challenges in contract analysis: complex table structures, linking content in and across documents, handling multiple counterparties and amendment processing.
We believe this is the way the market will now solve the problem of understanding unstructured data and make the journey to smart content and contacts. It remains to be seen whether we will maintain our market lead in this approach – but we are busy hedgehogs and we are confident that we will frustrate those hungry foxes!