We’re pleased to announce that Chapters 4 and 5 are available for early access for Relevant Search! Please read and give us feedback. This is early access for a reason: we want to hear what you think!
What do we have in store?
Chapter 4 teaches you to carefully control how your text and data is analyzed into tokens. This process is key for building a search index that can be used to intuitively find articles, people, ideas, locations, and just about anything your users hope to find. You need to learn to shape the resulting tokens that go into building the inverted index precisely use case! This of course applies to text. But more importantly, the chapter teaches that features of just about anything can be expressed to a search engine as searchable tokens. Heck, we did this recently in a pretty naive image search – turning pixels and colors into tokens! Any entity users might be looking for – places, ideas, even melodies – can all be tokenized and searched! Check out the chapter to learn more.
Chapter 5 teaches how to compose searches over multiple search criteria or fields. Often developers turn to Elasticsearch’s multi_match query or Solr’s edismax, listing every field they think is important to search over. But do you truly know the behavior of these queries? Or does it remain mystical to you? In this chapter, we teach you that in order to combine the influence of multiple fields. You’ll need to understanding two key techniques. First you’ll learn how to manipulate relevance scores into becoming true signals. Instead of numbers, you treat them as measurements of specific user or business ranking criteria like “the user searched for the article’s author” or “the restaurant is close by”. Second, you’ll mastering control over ranking function – how multi field queries like dismax and multi_match combine scores together. You’ll see how signals are combined or selected to combine useful ranking signals.
So check out these chapters. We’re eager to hear your feedback! Get in Touch!