Google launches AutoML Natural Language for improved text classification and model training.

2019-12-15

Earlier this year, Google released AutoML Natural Language, an extension of its cloud-based automated machine learning platform to the field of natural language processing. After months of beta testing, AutoML is now launching its general version to global customers, supporting tasks such as classification, sentiment analysis, entity extraction, and a range of file formats. AutoML Natural Language utilizes machine learning to uncover the structure and meaning of text, such as emails, chat logs, and social media posts. It can extract information about people, places, and events from uploaded and pasted text or Google Cloud Storage documents. Users can train their own custom AI models to classify, detect, and analyze emotions, entities, content, and grammar. Additionally, it offers custom entity extraction, allowing identification of domain-specific entities not present in standard language models. AutoML Natural Language has over 5000 classification labels and allows training on up to 1 million documents, with a maximum document size of 10MB. Google states that this makes it suitable for "complex" use cases, such as understanding legal documents or document segmentation for organizations with large content classifications. Furthermore, Google announced that AutoML Natural Language has obtained a moderate authorization from Fedramp, indicating that it has undergone review according to US government specifications to handle data with limited or significant impact. Google states that this feature, along with the newly introduced capabilities to create datasets, train models, and make predictions, allows data and related machine learning processing to remain within a single server region. Hearst Corporation has already been using AutoML Natural Language to assist in organizing the content of its domestic and international magazines. Japanese publisher Nikkei Group is also utilizing AutoML Translate to publish articles in different languages. Chicory is a third early adopter, developing customized digital shopping and marketing solutions for grocery retailers such as Kroger, Amazon, and Instacart. According to a Google Natural Language product manager, the ultimate goal is to provide a simple, unadorned training method for organizations, researchers, and businesses that require customized machine learning models. Natural language processing is a valuable tool for uncovering text structure and meaning, and its model quality can be continuously improved through better fine-tuning techniques and a larger model search space.

RECENT AI NEWS

RECENT AI TOOLS