Why existing AI techniques are not suitable for Contract Analytics
During this year’s TechCrunch Disrupt conference the subject of Artificial Intelligence (AI) and machine learning came up several times. With more and more companies from Life Sciences, Healthcare, FinTech and Transportation shifting their R&D efforts to understanding language, the topic was predictably popular. What wasn’t as predictable was the general consensus that AI technology for understanding language is still too young to be utilized as usefully as we’d like to hope.
This is particularly true when it comes to contract analytics. Google has a fairly advanced system for understanding and translating natural language – namely – Google Translate. But would you use it to translate your legal contracts in English to Spanish? Absolutely not! As a proof, simply take any of your existing contracts, translate it to Spanish and then translate it back to English and observe the results. Most natural language processing systems fail spectacularly when trying to make sense of contracts and legal documents.
Prior to founding Infinote, my experience with content management and background in artificial intelligence led me to the question of whether there was a better, more agile way of analyzing contracts. I was faced with the choice of whether to create a product based on natural language processing techniques or design algorithms that understand patterns and attempt to model information storage and retrieval mechanism of the human brain. Ultimately, I chose the latter and I’ll explain why.
Natural Language Processing is not yet the right approach to contract analytics
Let’s first begin with why when developing Infinote, I chose to steer clear of natural language processing and similar hyped up AI technology, despite my personal fascination with it. When developing software solutions for discovery and analysis, accuracy is everything. Unfortunately, AI and natural language processing, as they stand today, cannot accommodate the need for 100% accuracy. Contract analytics software solutions that are based on AI mechanisms operate on the basis of natural language algorithms, designed to “guess” and not “know”. It’s an ever-learning and constantly evolving mechanism, which is without doubt a great concept. But our simple demonstration of attempting to translate a contract from English to Spanish and back to English clearly shows major deficiencies of this approach.
In regulatory and legal circles, accuracy is critical
When my team and I first started developing Infinote, we wanted to create a solution which enabled users to search, analyze, update and audit contracts with 100% accuracy and existing AI-based technology couldn’t deliver this. That’s why we developed sophisticated algorithms to organize content and use computers for what they excel at – pure computation. Our algorithms attempt to mimic how the human brain stores and retrieves information.
Applied mathematics has an important role in developing tools for analysis and optimization of complex problems. Designing a solution that will achieve complete accuracy requires the development of mathematical foundations based on an understanding of complex systems. Using computational mathematics, we are able to carefully classify the problem, then develop powerful algorithms to deliver an accurate result. Using this approach, our optimized algorithms and computational horse-power ensure that we deliver 100% accuracy, and not a “best guess”. You know each time you run a query that your search results are displayed accurately and within the correct context. This approach eliminates the “teaching” and “learning” aspect of AI-technology and ensures your results are what you expect, and need, straight from the get-go.
Ultimately, what our team at Infinote strives to achieve is to set the standard for the accurate identification and analysis of enterprise content. With this we can save users days of manual, error prone effort and allow them to focus on doing the things that drive business growth.
If you want to know more about Infinote, or just want to spar around the pros and cons of AI and natural language processing technologies then drop me a note. I’d love to hear from you.