Google Introducing Semantic Experiences with Talk to Books and Semantris

Google really does significant research into natural language processing and synthesis, however not every project needs to be an innovative Assistant feature or voice improvement. The company has a little fun occasionally, as the master AI allows it, and now it has published a couple of web experiments that allow you to engage with its word-association systems in a playful way.

Initially is an attractive way of exploring by using Google Books, that wonderful database so occasionally mentioned nowadays. Rather than simply searching for text or even title verbatim, you could possibly ask queries, such as “Why was Napoleon exiled ?” or “What is the nature of consciousness ?”

I came across, however it’s not a truly intuitive way to deal with a body of knowledge, at least for me. Every Time I raise a question, I usually expect to get a remedy, not a competing number of quotes which may or may not bear on the inquiry. So when I can’t really picture by using this on a regular basis, it’s an interesting technique to demonstrate the versatility of the semantic engine at work here. And it could perfectly expose you to some new authors, however the 100, 000 books contained in the database are something of a mixed bag.


Google SemantrisAnother project Google emphasizes is a game which refers to Semantrisalthough I have to admit it’s quite so easy that deserve the “-tris” moniker. You’re provided a bunch of words and one particularly is highlighted. You type the word you most associate with that one, and the words will reorder with, as Google’s AI recognizes it, the closest matches to your word on the bottom. When you moved the target word to the bottom, it blows up a few words and adds a few more.

It’s a pleasant little time waster, however I couldn’t help out but really feel I was actually just a guinea pig offering testing and training for Google’s word association adviser. It has been also pretty easy — I didn’t really feel much of an accomplishment for affiliating “water” with “boat” — although maybe it becomes more difficult as it proceeds. I’ve questionedGoogle in case our answers are feeding into the AI’s training database.

For the coders as well as machine learning enthusiasts among you, Google has also provided certain pre-trained TensorFlow modules, and of course, documented their work in some papers associated with the blog post.



Leave a Reply