Deep Learning for Search

Part 2. Throwing neural nets at a search engine

Now that you know something about the fundamentals of search and deep learning, you can start throwing neural networks at a search engine wherever you see fit, right? In theory, yes; in practice, no. Deep neural networks aren’t magic: you need to use extreme care when deciding where and how using such powerful techniques makes sense. Chapters 3–6 look at tasks that every modern search engine commonly performs and highlight their limitations. As we identify them, we’ll explore how to use deep learning to mitigate such issues. You’ll see how to better solve the search engine task, either by looking at example output or by using more rigorous information-retrieval metrics.