Explore machine learning and how it benefits you
Machine learning is a type of artificial intelligence (AI), which focuses on the study of algorithms that allow computers to become more accurate at predicting outcomes without being explicitly programmed to do so. In other words, this AI uses algorithms to tell a machine how to organize information, find patterns and make predictions efficiently. The learning part occurs as the algorithm self-improves, adjusting based on the data it is given.
While there are many types of machine learning, we will focus on three.
Deep learning is a subfield of machine learning that uses multiple layers of information to process, understand and learn more in every subsequent layer. This series of layers is often described as a neural network, inspired by the way our brain communicates. There are many advanced applications of deep learning, such as computer vision, and we are just beginning to uncover useful applications in various industries.
Unsupervised learning is a type of machine learning that allows the machine to self-improve based on large data sets that haven't been previously categorized. Unsupervised learning is most often used for organizing information into different groups by identifying similarities among large data sets. For example, banks use unsupervised learning to identify fraudulent charges by categorizing your normal pattern of spending and finding a charge that doesn't match.
Supervised learning is the simplest type of machine learning with the most commercial use cases. This model requires a set of data that is clearly defined and labeled by humans. The machine can learn from the categorized data by recognizing the differences and similarities among the information.
Today, supervised learning models are used across industries, like real estate, legal and finance, to optimize many tasks, including contract analytics, document classification and due diligence — this is how and why DFIN’s contract analytics solution, eBrevia, was designed.
Think about it this way:
You are given 500 contracts and asked to identify every appearance of the term "limitation of liability". Considering "limitation of liability" is a concept that can be expressed in different ways, it may take you weeks to read through every page and accurately find each use.
Using a supervised learning algorithm, you can provide a machine the different ways any concept may occur — for example, "limitation of liability" may also appear as "waiver of consequential damages," "carve-out for gross negligence," "actual and direct losses," etc. — and incorporate natural language processing to learn the patterns to identify the concept you need to identify, regardless of the exact vocabulary used.
By uploading your documents, the machine can use the learned information to process the language on every page, highlighting every appearance of the concept for you to review.
Now, instead of spending weeks reading every page, you only need minutes to review all the phrases identified by the computer with efficiency and confidence.