IBM announced upgrades to IBM Watson Discovery‘s natural language processing (NLP) capabilities. These planned upgrades will benefit corporate users in financial services, insurance, and legal services. Through the discovery of insights, these upgrades are leveraged to enhance customer service and accelerate business operations.
Businesses are increasingly turning to natural language processing (NLP) and machine learning to assist them in sifting through increasing amounts of documents and data sets in various formats. By leveraging AI to extract insights from documents, business users can reduce research time. Moreover, it can enable employees to make more fact-based decisions, especially when performing complex, time-sensitive tasks such as processing insurance claims, conducting financial analyses and so on.
The new tools released today by IBM are intended to make it easier for Watson Discovery users to swiftly tailor the underlying NLP models to their business’s unique language. As a result of IBM Research’s advances in NLP, business users can teach Watson Discovery to assist them in reading, comprehending, and surfacing; in addition, more precise insights from massive quantities of complicated, industry-specific documents.
Pre-trained Model
Watson Discovery’s Smart Document Understanding feature is currently available in the Plus, Enterprise, and Premium plans. This model is designed to automatically comprehend a document’s visual structure and layout without the assistance of a developer or a data scientist. In addition, this model enables users to locate previously hidden easily or difficult-to-find information, such as text embedded in complicated table structures or images.
Automatic Pattern Recognition
IBM has launched a beta version of a new advanced pattern generation function in the Plus, Premium, and Enterprise plans. The feature aims to assist customers in swiftly identifying business-specific language patterns inside their documents. This feature is critical for jobs such as reviewing large volumes of contracts or financial reports, which may present the same information, such as an increase or reduction in income, in a variety of different formats or using a variety of different terms. IBM Research developed it to aid in the effective labelling of data and training of models. It is designed to begin learning text patterns from as little as two examples and then refine the pattern based on user feedback. This research enables users to train a model more quickly by eliminating manual and time-consuming procedures such as creating rules and expressions.
Advanced Natural Language Processing
Training NLP models is a time-consuming process that requires extensive data preparation, labelling, and orchestration. Frequently, models trained on generic data sets fail to obtain the correct data. IBM simplifies this process with a new custom entity extractor feature, which is now available in beta for Watson Discovery Premium users. This feature reduces data preparation effort, simplifies labelling with active learning and bulk annotation capabilities, and enables simple model deployment, accelerating training time.
“The continuous stream of innovation flowing into IBM Watson from IBM Research is why global businesses in financial services, insurance, and legal services rely on IBM to assist them in identifying emerging business trends, improving operational efficiency, and empowering their employees to uncover new insights,” said Daniel Hernandez, IBM’s General Manager of Data and AI. “With the pipeline of natural language processing improvements we’re introducing to Watson Discovery, organisations can continue to differentiate the signal from the noise and provide better service to their customers and workers.”
Along with the new functionalities, IBM demonstrates how firms in the legal, financial, and insurance industries leverage Watson Discovery’s existing capabilities to automate and revolutionise business processes.
For more information on Watson Discovery, read here.
For more information about IBM Watson, read here.
