Google programmers have released two new applications that take advantage of the company's latest advances in natural language comprehension by computers. In an effort to develop algorithms that better understand human language, the search giant has focused on designing hierarchical vector models.
These models use vectors to help the program self-learn, understand the relationship between words in phrases and the idea of a statement. In addition, Google software engineers note that they have already begun to use vectors to define relationships between larger clusters of words such as sentences and short paragraphs. The Hierarchical Vector Model is the same machine learning model that powers the Smart Reply service in Gmail.
Google Semantic Experiences
You can see how both applications work on the Google Semantic Experiences website. One of them is called Talk to Books. Its mission is to help users search for literature by answering their questions. The algorithm is able to analyze the contents of books and extract information from them that meets the needs of users. However, Google warns that the technology is far from perfect. For example, there are cases when a program takes information out of context, as a result of which its original meaning is lost. In addition, the algorithm may have difficulty understanding complex questions and statements.
A game of associations for artificial intelligence
On the same page as Talk to Books, you can check out Google's second development, the Semantris game. It is an association game in which machine learning is used to find the relationship between words on the screen and what the user is typing. Semantris is available in two modes - arcade and blocky. In arcade mode, you need to act and think quickly. Blocky has no time limits, in it the player can react not only to individual words, but also to phrases.
Google hopes that in the near future this algorithm will find applications in data classification, semantic clustering, and whitelisting. Developers interested in this technology can jump into experimentation and develop their own applications using an adapted semantic algorithm model from the TensorFlow platform.
The Topic of Article: Google is teaching AI through literature and games.