AlchemyAPI and The State of Deep Learning
AlchemyAPI recently announced a taxonomy and a sentiment analysis API based on deep learning that can help transform digital content into ad inventory. Matching the content of a webpage with relevant ads is the goal of every advertiser as it improves click through rates and drives down cost per click. AlchemyAPI delivers an API that can drill down to 1,000 categories and down to five levels deep. The API also offers the ability to create new categories from any arbitrary phrase. This can be useful for niche content that can’t otherwise be targeted by ads.
Deep learning is a bottom-up approach using unsupervised learning to train a system and then using feedback from other supervised learning systems to refine the results. Coupled with Big Data and several layers, Deep Neural Networks can deliver better accuracy across a broad range of applications.
Throughout the past months there has been high activity in the deep learning space with a series of product launches and acquisitions.
IBM is utilizing Watson, the Jeopardy winner for research on personalized medicine. Watson uses DeepQA to analyze, process and answer questions with confidence.
Yahoo has acquired LookFlow and IQ Engines, both in the image recognition space. LookFlow is using deep learning for image recognition and will join the Flickr team and apply their knowledge in deep learning projects for Yahoo.
Facebook hired NYU deep learning expert Yann LeCun to run its AI lab. Yann LeCun is leading efforts to make Facebook better understand people and pick the correct subset of information to show to them.
Microsoft recently demoed a speech-to-speech English to Mandarin translation engine, claiming to have reduced error rates by 30% the most in speech recognition since hidden Markov modelling was introduced in 1979.
Last but not least, Google acquired DeepMind in January, a startup that is working towards building an AI system that thinks, according to Carnegie Mellon professor Larry Waserman. DeepMind boasts a lineup of great researchers in the AI field, on par with Google’s own researchers. Google is investing heavily in deep learning and not without a reason as search, pattern recognition for security puproses, social and commerce can all benefit from better AI systems.
Ralph Winzinger Nov 25, 2014
John Krewson, Steve Ropa and Matt Badgley Nov 24, 2014