“Children make good snacks.” If you were to read that sentence, you’d likely assume that the author meant something akin to “Kids are skilled at preparing small meals.” Readers less familiar with English might instead interpret this to mean something more outlandish, such as “Small humans are tasty.” Computers once sat in this latter camp. For decades, machines struggled to process and interpret complex human languages. The development of Natural Language Processing (NLP) changed this.
What is Natural Language Processing?
NLP is a form of artificial intelligence that focuses on the interaction between computers and humans using natural, or human, language. It might be helpful to think of NLP in two parts: first, the “processing” piece and, second, the “natural language” piece. Computers are designed to run off of computer languages, otherwise known as code. Any request you give a computer, like printing a document or searching a case on CanLII, gets converted from human language into code. This is called processing, and it allows a computer to understand your request.
Now the natural language bit. Computers have always been able to understand code, but code is starkly different from human language. Code consists of a small set of symbols arranged in a regimented way. There’s no room for error or improvisation. Human languages, by comparison, are full of imperfections. Human languages encompass large, diverse vocabularies; use the same words for multiple meanings; and are often applied improperly but still understood. Humans can slur speech, use slang, omit words altogether, or say something sarcastic and still be understood. NLP allows a computer to understand human language, despite these grammatical errors and lexical ambiguities.
How Does Natural Language Processing Work?
Each NLP program is unique, but the majority of them are built on similar instructions. Just as if you were learning a new language, an NLP program first needs to understand the discrete components of a language. This includes things like words, numbers, and punctuation. Programmers build this into a program through techniques like part-of-speech tagging, which assigns each word or phrase to a category, such as “apple” to noun and “clear” to adjective and verb. Similarly, named-entity recognition assigns each word to a more descriptive category, like “Martin” to person and “Montréal” to location.
Once a program understands the components of a language, it then needs to be instructed about how these components are configured to deliver meaning. A technique like parsing teaches a computer the possible meanings of a string of words based on grammatical rules. Word sense disambiguation gives meaning to a word or sentence based on a broader context. Before the ‘80s, programmers would build these NLP programs manually, inputting words, labelling them, writing grammar rules, and refining the program based on the results it yielded. It’s more common today for developers to rely on machine learning techniques that improve programs automatically through experience.
Applying Natural Language Processing
You likely interact with NLP applications many times a day. It’s what allows Microsoft Word to provide grammar suggestions, Siri to play T-Pain when you ask her to, and your email platform to filter out spam. These applications are not only useful, but illustrative of how NLP can increase the quality of one’s work and decrease the time it takes to complete it.
Martin Luther King Jr. once said that “Justice too long delayed is justice denied.” The efficiencies that NLP offers make justice more accessible. NLP applications allow legal practitioners to work more quickly. Expediency results in more timely legal decisions, as well as cost savings for legal practitioners. Ideally, a portion of these savings is reflected in the cost for clients to attain such legal services in the first place.
To assist legal practitioners in being more efficient, accurate, and cost-effective, a number of companies have harnessed NLP to build valuable legal technologies. For example, Kira systems use NLP to extract and analyze the text in digital contracts to assist lawyers in their due diligence. Litera uses NLP to proofread documents more quickly than a human would, relieving some of these time constraints. Hotdocs, by comparison, uses NLP to produce full standard-form text documents automatically, reducing the time lawyers might spend formatting and preparing early drafts of documents.
What Might the Future Hold for NLP and Access to Justice?
Perhaps NLP’s greatest promise is its ability to make legal concepts accessible to the general public. An old computing adage goes, “Garbage in, garbage out”. The quality of a computer’s function is a direct result of the quality of code upon which it operates. NLP changes this equation. NLP can process complex texts and restructure them into more accessible terms. It allows those without a seasoned understanding of legal fields, jargon, or issues to obtain high-quality legal information by searching using common language. This has obvious implications for the growing body of self-represented litigants and others navigating a complicated legal system without assistance from a legal practitioner.
Ontario Attorney General Doug Downey stated that Ontario’s court system advanced “25 years in 25 days” due to COVID-19. This was only the most recent catalyst to push the legal system towards adopting more time- and cost-saving innovations. NLP continues to be essential in this shift to technology and will have a profound impact on the affordability and accessibility of legal services. It will also continue to ensure that we never mistake children for snacks.
Interested in seeing AI applied to conflict resolution? Check out our free AI-Powered Legal Tools at MyOpenCourt.org