Google brings support for custom translations and text categorization to AutoML – TechCrunch

Pre-trained machine learning models are suitable for many use cases, but to get the most out of this technology, you need custom models. Since getting into machine learning isn’t exactly easy, Google (and others) have gone for a hybrid approach that allows users to upload their own data to customize existing models. Google’s version is AutoML, which until now only provided this capability for machine vision tasks under the nickname AutoML Vision.

Starting today, the company adds two new features to AutoML: AutoML Natural Language for predicting text categories and AutoML Translation, which allows users to upload their own language pairs to achieve better translations for texts in highly specialized fields, for example. Additionally, Google is launching AutoML Vision out of preview and in its private beta.

Rajen Sheth, director of product management for Google Cloud AI, said this AutoML extension is a further step towards the company’s vision of democratizing AI. “What we’re trying to do with Cloud AI is enable everyone in the world to use AI and create models for their purposes,” he said. For most of its customers, however, the pre-trained models aren’t good enough, but for most companies, it’s hard to find the machine learning experts who would allow them to create their own custom models. Given this demand, it is perhaps not surprising that to date, approximately 18,000 users have signed up for the AutoML Vision preview.

“Natural language is something that is really the next frontier in this,” noted Sheth when discussing the new natural language API. “It’s something very useful for customers. Because over 90 percent of the information our customers have in their businesses is unstructured and free information. And a lot of it is text documents or emails or whatever. Many clients are trying to find ways to extract meaning and information from these documents.

As for AutoML Translation, the advantages of this type of customization are quite obvious, given that the translation of highly specialized texts remains the domain of experts. For example, Sheth noted that “driver” in a tech document could refer to a device driver for Windows 10, for example, while in other text it could just be a person driving a car (up to ‘that computers take over this task, too).

Comments are closed.