Over a million developers have joined DZone.

Semantic Search Allows You to Talk to Books

DZone's Guide to

Semantic Search Allows You to Talk to Books

Google Research has released a semantic search tool to allow you to search across Google Books for inspiration, quotes, or a new book to read.

· AI Zone ·
Free Resource

Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.

Here’s a brief summary of Google Research’s recent announcement of Talk to Books: a search tool with a difference.

When you submit a search query, the AI running the search has scanned more than 100,000 English-language books containing over 600 million sentences. It returns results within Google books rather than a set of links to web pages as for a traditional search. Here are the results from a sample search for, “What is fun about computer programming?”

There is another difference to the search, too: you should make your query a complete sentence rather than use a set of traditional keywords. That’s because the AI behind Talk to Books uses semantic search. It is trained on human conversations so is better able to understand natural human language.

“With Talk to Books, we provide an entirely new way to explore books. You make a statement or ask a question, and the tool finds sentences in books that respond, with no dependence on keyword matching,” says the Google Research Blog.

Talk to Books was developed by a team at Google Research led by the legendary Ray Kurzweil. As he explains in a post on his own blog, “Semantic search is based on searching meaning, rather than on keywords or phrases. Developed with machine learning, it uses ‘natural language understanding’ of words and phrases.”  

AI that understands natural language uses word vectors to learn relationships between words by taking examples of actual language usage. The vector models map semantically similar phrases; that is, those using similar language. Google says their model “extends the idea of representing language in a vector space by creating vectors for larger chunks of language such as full sentences and small paragraphs.”

The semantic search model is explained further on Google’s new Semantic Experiences page, where they have published Universal Sentence Encoder and have also provided a pretrained semantic TensorFlow module so anyone can experiment with sentence encoding and phrase encoding.

It’s fair to say that Talk to Books is not going to replace Google search when you need to find a quick recipe or check the distance between your house and the nearest swimming pool. On the landing page for the tool, Google suggest you use it as a “creativity tool to explore ideas and discover books by getting quotes that respond to your queries.” It’s a nice way to look at what others have written on a subject, for inspiration, or simply to find an apposite quote.

I asked “How can I stop procrastinating?” and spent a happy hour browsing the returned results. Whoops!

Your machine learning project needs enormous amounts of training data to get to a production-ready confidence level. Get a checklist approach to assembling the combination of technology, workforce and project management skills you’ll need to prepare your own training data.

ai ,nlp ,search engine ,semantic search

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}