Deep-learning-as-a-service startup AlchemyAPI has launched a new product aimed at advertisers that want to better classify their content and turn it into targeted ad inventory. Automation is probably the major benefit of deep learning technologies right now because users are able to tag and classify text and images without using many human resources and with fairly high confidence the results will be accurate.
AlchemyAPI claims its new Taxonomy API recognizes 1,000 product categories and is able to drill down five levels below the surface (e.g., from financial services down to automobile loans) with little or no input from users. The company also announced on Wednesday that its sentiment analysis API can now better understand text containing poor grammar, slang and mixed messages. Another new product, which it’s calling a combined call API, lets users start with a high-level classification task and then automatically switch to a lower-level task such as extracting entities.
Several months ago, in September, AlchemyAPI showed off a demo version of an API it’s working on to classify the the objects in images.
Deep learning is an approach to artificial intelligence that has made big advances lately thanks to improvements in computing power, data volumes and algorithms, with much of the credit (at least publicly) going to companies such as Google, Facebook, Microsoft and the researchers they now employ. At a high level, deep learning works by creating models that layer multiple neural networks on top of each other, each one of which learns a particular set of features that comprise the data they’re analyzing. Most current efforts are focused on computer vision and text analysis.
Here is a video of AlchemyAPI Founder and CEO Eliiot Turner, along with Stephen Gold, a vice president in IBM’s Watson Group, discussing artificial intelligence as a service during our Structure Data conference last week.