Machine learning (ML) models demand comprehensive and accurate datasets to ensure proper performance. According to DataRobot nearly 96% of companies encounter delays getting ML models into production and 78% of ML projects stall before deployment. Everyone can agree that data must be useful to train models. The offshore model of working through vast amounts of unstructured data to annotate assets is untenable. DCL works with organizational data to automate the creation of training sets by building against your own documentation, or we can use our own training sets to support your ML initiatives.

Extracting and structuring content from text- or image-based tables has long been a challenge. Transforming tabular content into a structured model such as XML or HTML is nearly always a manual or semi-manual process. Tables are tough to structure due to inconsistencies with tabular content, high diversity of layouts, complicated elements such as straddle headings, various alignments of contents, the presence of empty cells, and other intricacies. That said, DCL and Fusemachines created an AI model that finds and extracts information from all tables in a document using a combination of Computer Vision (CV) and Natural Language Processing (NLP).

According to McKinsey, AI-enabled supply-chain management has allowed adopters to reduce logistics costs by 15%, service levels by 65%, and inventory levels by 35%. Some of the biggest companies using AI to manage their supply chain include Amazon, Walmart, Johnson & Johnson, Pepsico, L’Oréal, and more. There are many valid applications of AI in supply chain management—inspection and quality control, intrusion and theft detection, driverless delivery vehicles and more. Of course with the benefits also come limitations and the critical need for specialized skills and knowledge to operate AI.

Accessibility in journalism is important for everyone and The Washington Post demonstrated its commitment by hiring its first accessibility engineer, Holden Foreman. Accessibility in media is more than alt text or any other single issue. As Holden explains, "There are many different types of disabilities, including temporary and situational disabilities — like a broken arm, bad internet, or loud surroundings — that change how people interact with journalism. These folks may rely on a keyboard to navigate content, but many websites and apps are built with the assumption that everyone can hover a mouse over a map, or click and drag on the screen. Assumptions like these hinder accessibility."