From Documents to Insights: How Law Students are Advancing Legal AI

Whether studying to become lawyers or paralegals, law students extract information from legal documents in their coursework. Therefore, when Tasq.ai came to us needing subject matter experts to answer questions on Modern Slavery Act documents and extract text to their data labeling platform to train AI, we knew our law students would be perfect for the job. 

‍

The Australian Government requires Modern Slavery documents to report on 6 criteria in a specific way. The 6 questions are:

1. Is the statement approved by the company’s principal governing body?
2. Is the statement signed by a responsible member of the reporting entity?
3. Does the statement clearly identify which entity/s covered by the statement is/are the relevant reporting entity/s required to report under the Act?
4. Does the statement clearly describe the risks of modern slavery practices in the operations and supply chains of the reporting entity?
5. Does the statement clearly describe the actions taken by the reporting entity to assess and address those risks, including due diligence and remediation processes?
6. Does the statement clearly describe how the reporting entity assesses the effectiveness of such actions?

‍

Our team of law students extracted relevant PDF text on Tasq.ai’s data labeling platform according to the specific guidelines provided. The team was asked to extract information on ~5,000 documents for the client to use as ground truth data to train their large language model (LLM). Our team understood the importance of extracting information at 100% accuracy so the AI would learn to extract the same information at high quality. 

‍

How did we guarantee 100% accuracy?

‍

A couple of ways:

1. Test and train the team continually until quality is met before assigning work.
2. Establish a quality assurance process with an internal QC.
3. Assign team lead to manage hours/documents completed to guarantee KPIs are met.

‍

Before coming to Yenda, Tasq.ai first went to their crowd. The crowd was unable to meet the quality standard required by the client. Where the crowd couldn’t deliver, our industry-specific team produced high-quality results. Here’s why:

1. Specialized knowledge and expertise relevant to the work at hand: Our students are familiar with the terminology, jargon, and technical concepts in law.
2. Greater attention to detail and accuracy of work: Our students understand the potential implications of errors and are motivated to ensure the highest quality output. We all know the quality of the AI model is only as good as the data used to train it.
3. Higher engagement: Industry-aligned projects give our freelancers a space to apply knowledge learned in school and the necessary experience to advance their careers.

‍

Within one week, we selected, trained, and on-boarded a team of law students to read, extract information, and answer detailed questions on Tasq.ai’s data labeling platform. The result: accurately annotated legal documents used to train an LLM to detect modern slavery breaches. 

‍

How do we create and train teams quickly?

‍

University partnerships. We partner with top universities throughout Zambia to source industry-specific students to work on client projects. Data labelers with industry knowledge, as seen in this case study, results in high-quality data labeling. It also provides relevant work experience for our students which keeps engagement high. 

‍

We partnered with the Zambia Institute of Advanced Legal Education (ZIALE) on this project.

‍

Our industry-specific talent has proven to perform better at text extraction compared to an outsourced crowd due to their specialized knowledge, attention to detail, and engagement. Tap into industry-specific Yenda teams to get the expertise needed for accurately labeled data. We’d love to connect you to Africa’s emerging talent!