Natural Language Processing – What It Is and Why You Should Care

Whether you’re asking Alexa to tell you a dad joke, wondering how Google predicted the embarrassing question you were asking, or thanking Word for correcting basic grammar errors (thanks again, old friend), we all use Natural Language Processing (NLP) as part of our everyday lives. In fact, the rapid progression of these technologies and the steady release of effective solutions to problems we didn’t even know we had, have made it pretty unimaginable to go a day without using it.

Alongside its partner in crime, Machine Learning, NLP is a field of Artificial Intelligence where we have seen a number of developments in the last few years, particularly where LegalTech is concerned. But before we get into some of its meaningful legal applications, let’s rewind for a minute and take a look at what NLP actually is and where it came from.

NLP combines computer science, linguistics, and artificial intelligence and is rooted in solving problems through the application of technological solutions which process and analyze natural language or speech. NLP dates back to the post-war era of the late 1940s. It is thought that Weaver and Booth kicked off the first machine translation project in 1946 which, understandably, was heavily rooted in ambitions to crack enemy code. Arguably, it was Chomsky, however, who provided a better linguistic understanding of how language can actually be broken down, through his 1957 publication, ‘Syntactic Structures’, which presented the idea of generative grammar. This ultimately allowed natural language research to branch out into other areas of interest, such as speech recognition.

Since 1960, the world has witnessed revelations in the field, both in terms of theoretical development and the actual production of systems. In the 1980s and beyond, those working in the field became increasingly focused on how systems could be applied to real-life issues. This hope was validated with the introduction of the internet and the rise in available data just screaming out to be processed.

If we fast forward to today, I don’t need to sit here typing on my HP ProBook to try and convince you that technology has changed. It is no surprise to anyone that we are living in an era where data is as valuable as gold. And it is indeed a gold mine for data scientists who have been able to accomplish a lot with this.

And yet, aside from the functions we are now comfortable with, there can still be some mist clouding our view of NLP’s promising applications in many sectors and the benefits that come hand in hand with its use. Let’s take a look at some of the biggest applications of NLP in the legal sector today.

Document Automation
Document automation and contract management are two areas where NLP has been widely researched. Tools created to help support these areas of work aim to assist in the creation of electronic documents. This may include using parts of pre-existing text to create a new document. For example, in some instances, users input relevant information, such as client and supplier details, and a contract is then drafted using this information. There are multiple document automation products on the market, such as NextChapter Docs, which works exactly like this. The software allows firms to build templates and generate legal forms using existing text, while also tailoring these for the client in question. This not only speeds up the process for firms and clients but is more efficient and removes the risk of human error. Imagine all the lawyers who are now able to take a lunch break! Like most important developments, once adopted, we can stick our 20/20 hindsight specs on and say things like, “why didn’t we do this 15 years ago?”. Who knows why, but better late than never.

Electronic Discovery
In terms of legal electronic discovery, the most common tools focus on perceiving issues as information retrieval tasks and are therefore used to locate and identify relevant text. For example, tools to aid eDiscovery tasks may do the following:

·         Search for documents which are relevant to the user.

·         Search documents by phrases or keywords, often involving a “bag of words” method which spits out similar words to those searched for by the user.

·         Search via concepts, such as “debt” which encompasses more documents than would be provided by a keyword search.

Let's look at a quick example of how this may be applied in a work setting:

Earlier this year, Epiq, a legal services company, launched a system which offers pre-trained models to clients to help with reviewing litigation cases.

Although the review process differs depending on each case there is some common language and themes which arise in most instances. The system captures these, identifying repeating information and key documents for further review. By using these tools, the litigation review process is sped up significantly for lawyers by acting as a digital highlighter which returns the most important information. This is of particular use to firms dealing with similar matters on a regular basis, such as workplace disputes. The models are then stored for future use, so when a similar matter arises, the corresponding model can be applied.

Overall, this makes the eDiscovery process a much simpler one. Why start from scratch every time when you could be supported by existing data?

Billing Analysis
The idea of using NLP tools to act like a digital highlighter is one that can be applied to the billing process. Billing analysis software seeks to replace email and paper billing, alleviating legal teams from tasks such as reviewing invoices and generating reports.

Let’s consider another recent example of how NLP can help in these instances.

Onit is a US company which has recently developed new features as part of their InvoiceAI tool which now includes invoice review in addition to invoice analysis. To break this down a bit more, the tool has been trained using millions of invoice charges to identify errors, such as areas of overpayment and unnecessary expenses. The tool can then amend the invoice accordingly, therefore reducing the number of invoices that require human review.

Overall, billing analysis technology helps with 3 aspects in particular:

·         Automating tasks in the everyday workflow;

·         Saving money; and

·         Generating data which can then be used for reporting and decision making.

This is just another way that NLP tools can be applied to the traditional legal workflow to support staff.

The above applications are far from exhaustive, and NLP is proving an important asset for law firms to have under their belt to support staff. However, the word ‘support’ is key here. Although tools which apply these techniques are highly complex and thoroughly trained, they are limited to what we allow them to do. In other words, the tools may pick out relevant terms or documents, but that’s not to say they understand the meaning of the words they are trained to identify. There is plenty of room for humans and machines alike when it comes to the everyday workflow.

Machines are best suited to complete time-consuming, repetitive tasks such as those discussed above. Luckily, these just so happen to be the tasks that humans don’t really want to do. We are therefore likely to see an abundance of solutions that will help to revolutionise traditional working practices in the near future.