AI Terms for Legal Professionals: Understanding What Powers Legal Tech

The legal technology landscape is evolving rapidly in front of our eyes, creating a whole new vocabulary of AI-related terms and phrases. For those of us who aren’t product developers or software engineers, it’s helpful to define what these terms mean. This user guide is designed to equip you with a better understanding of key AI concepts. 

• The Basics  
• Extractive vs. Generative
• Large Language Models (LLM)
• Legal Tech Examples

The Basics

• Algorithm: a coded set of instructions for software that solves a problem or performs a computation.  

• Artificial Intelligence (AI): computer software and systems that learn, plan, reason or process natural language as they go rather than only relying on pre-programmed tasks, i.e., speech recognition, computer vision, translation between (natural) languages, as well as other mappings of inputs. 

• Natural Language Processing (NLP): a type of AI used to analyze, understand, and generate human language. For legal documents, NLP analyzes legal documents, contracts, and other legal texts to identify key provisions, clauses, and risks. 

• Semantic Search: a search that weighs not only the keywords but also the context and intent behind the search query, which is critical in legal analysis, i.e., Lexis+.  

• Machine Learning (ML): a subfield of artificial intelligence, which is broadly defined as the capability of a machine to learn without being explicitly programmed.  Machine learning algorithms build a model based on sample data, i.e., training data, to make predictions or decisions without being explicitly programmed to do so.  

• Data Analytics: the process of automating data analysis through machine learning to target critical insights and actionable intelligence faster and with higher accuracy.  

• Deep Learning: a type of machine learning leveraging neural networks to learn by example, much like a human.  

• Neural Network: a deep learning process that mimics the human brain by using interconnected nodes, or neurons, in a layered structure. There are three types of layers: the input layer, which takes in the data, a hidden or processing layer or layers, which perform the functions of the model, and an output layer, which produces the output. Each layer consists of neurons. Each neuron receives input from the other neurons, processes it, and produces output.  

• Constitutional AI: a method for ensuring that AI systems are in sync with human values. This is a term used by researchers to describe the controls they’ve built to ensure that their AI systems behave ethically.  

Extractive vs Generative AI

• Extractive AI: an AI algorithm that extracts relevant data points from the data it has been trained on, i.e., Lexis Answers, Brief Analysis, Fact & Issue Finder, and many other LexisNexis products.  

• Generative AI: an algorithm that generates new outputs based on the data it has been trained on. Unlike extractive AI systems designed to recognize patterns, extract pre-existing data, and make predictions, generative AI creates new content in the form of images, text, audio, and more.

Large Language Models

• Large Language Model (LLM): a machine learning model that can recognize, summarize, translate, predict, analyze sentiment, and generate text based on the patterns and relationships (probabilities) it has learned from massive datasets. LLMs work by predicting the next term, or word, in a sentence, given the words that came before it. LexisNexis leverages many highly trained LLMs to help users surface critical insights faster in Lexis Answers, Brief Analysis, Fact & Issue Finder, and many others.  

• Bidirectional Encoder Representations (BERT): a ML and NLP framework that analyzes nuanced, niche language and context with a high degree of accuracy. LexisNexis has trained BERT technology on all things legal and leverages it across Lexis+.  

• Conversational AI: a collection of technologies that power a conversational assistant or chatbot. Together they enable efficient, automated communication via text and speech by understanding intent, deciphering language and context, and responding in a human-like manner. 

• Chatbot: a software program that conducts conversational interactions with humans through text or voice, allowing humans to interact with digital devices as if communicating with a real person. It can be as simple as a rudimentary program answering basic questions with a brief response or as sophisticated as a conversational assistant holding a long, nuanced conversation, discerning complex intentions of its human users, and learning and evolving to deliver increased levels of personalization.  

• GPT: Generative Pretrained Transformer is a family of AI language models introduced by OpenAI.  These models can be fine-tuned for various natural language processing tasks, such as text generation, language translation, and text classification. The "pre-training" in its name refers to the initial training process on a large text collection where the model learns to predict the next word in a passage, which provides a solid foundation for the model to perform well on downstream tasks with limited amounts of task-specific data.  

• ChatGPT: an artificial intelligence chatbot launched by OpenAI in November 2022. It is built on top of OpenAI's GPT-3.5 using supervised learning as well as reinforcement learning. Although the core function of a chatbot is to mimic a human conversation, ChatGPT is versatile and acts as Generative AI. For example, it can write and debug computer programs, compose music, teleplays, fairy tales, and student essays; answer test questions (sometimes, depending on the test, at a level above the average human test-taker) and more.   

• GPT-3: the third-generation language prediction model in the GPT series introduced in May 2020 by OpenAI.  It is an autoregressive language model, i.e., it predicts future outcomes using what it has previously observed and produces human-like text. LexisNexis leverages secure GPT-3 technology today. 

• GPT-4: the fourth-generation language prediction model in the GPT series introduced in March 2023 by OpenAI is multimodal and accepts both image and text inputs to ultimately produce text. 

LegalTech Examples

• Legal Analytics: AI that analyzes legal data pertaining to judges, attorneys, courts, and more and provides insights into actionable trends and patterns, i.e., Litigation Analytics and Lex Machina.  

• Legal Research: AI-powered legal research tools to help lawyers and legal professionals efficiently find relevant case law, statutes, and regulations, i.e., Lexis+.  

• Contract Analysis and Management: AI tools can help automate contract review, identify clauses that require attention, and manage the lifecycle of a contract, i.e., CounselLink with Parley Pro.  

• Intellectual Property: AI can help automate the process of patent searches, prior art searches, and trademark searches, i.e., LexisNexis IP solutions.  

• Risk Assessment: AI tools can help assess risk in contracts, transactions, and legal decisions, i.e., Lexis Create, which is launching in the US soon.  

• Sentiment Analysis: This involves using NLP to analyze legal documents, social media, and news sources to identify sentiment around a particular legal issue or case, i.e., Newsdesk.  

The legal technology landscape is shifting quickly, and LexisNexis is poised to support the legal industry with state-of-the-art AI-enabled technology that helps users find actionable insights faster where and when they need them. Learn more about our latest survey that shows 39% of lawyers, 46% of law students and 45% of consumers agree that generative AI tools will significantly transform the practice of law.