Artificial Intelligence with Microsoft Power BI

Chapter 8. Text Analytics

What would you do if you had to review over one million technical documents? Well, the professional services company Accenture had that many legal contracts and was adding thousands of new ones each month. This represented a considerable burden on the company, requiring an incredible amount of manual work to individually review and process each document.

That all changed when Accenture developed an AI solution to extract meaning from this ocean of text. The project used text analytics to identify which keywords related to which legal clauses. This automated matching allowed a user to search for a specific legal clause, like “limited liability,” even if those specific words never showed up within a contract.¹ Accenture decided to build this intelligent system in-house to have greater control over the outcome. This made sense for the company given the scale of the problem, available resources, and the potential upside of creating a customized system. Other companies facing the same scenario, however, may find it better to outsource production of such an AI model or choose an off-the-shelf solution.

Custom Models Versus Pretrained Models

Many organizations face the question of whether to buy or build an AI solution or enlist the help of a partner, and that decision requires understanding the distinction between custom and pretrained models. A custom machine learning model is one that is trained from scratch by an individual or organization with their own ...