Thomson Reuters and its Tryst With Artificial Intelligence

Thomson Reuters has initiated different research projects based on AI.

Thomson Reuters is a leading multinational conglomerate that provides trusted data and information across different industries. The company mainly serves Legal, Tax and Accounting, and Media. Artificial intelligence is the norm that industries are rapidly adopting today. AI and other disruptive technologies like machine learning are enabling businesses to gain drive efficiency and growth. Companies are highly investing in AI and launching innovations. Recently LG AI research invested US$89 million in a project called "Super Giant AI". Thomson Reuters is also committed to artificial intelligence. The company has certain AI principles in place that revolve around trust, security, privacy, and a human-centric approach. Natural Language, Machine Learning, and information retrieval are some of their research areas. Here are some of their AI research projects.

Simplifying Document Review

Thomson Reuters is working on an AI project that aims at simplifying the process of document analysis and development for knowledge workers. For this to be effective, they are trying to find segmentation algorithms that can extract the structural hints from a document and separate it into the heading, subheadings, etc. Next comes NLP-based entity and relation extraction, which can find the relation between different elements of a document. This is critical especially for financial and legal documents. Developing a taxonomy and creating models can be time-consuming and thus establishing a research area can be easier. This includes active learning, hybrid machine learning models, and more. Deviation analysis and synthesizing the product according to each customer's need is what the company is trying to achieve through this research project.

Machine Learning-Driven Text Mining

As text mining and accurate information extraction are essential for tax and legal systems, Thomson Reuters leveraged machine learning and NLP models to solve this. This gives analytical products with higher value. Motion analysis and party outcome detection are the two significant steps in this project. This machine learning-based technique enables Thomson Reuters to aid more clients with valuable insights.

Machine Reading Comprehension

Thomson is highly investing in this AI project to experiment and develop futuristic approaches towards machine reading comprehension problems through deep learning. These models can then be applied to relevant projects that the company deals with and also deploy in question-answering tasks. The company is experimenting with BERT and it states that till now it has produced exceptional results for Legal and Tax based tasks. The company has a future vision to develop a conversational system that can engage in continuous interaction with the users and provide the information they need.