In this talk, Shakhina Pulatova gets deep on the art of searching highlighting the importance of machine learning. She takes us through search basics, to what powers search products, and ends with why you should give creating search products a try.
The act of searching means the user has a question that needs answering, so the first key principle in building a search tool is to understand the intent. There are many facets that comprise an intent, and they may shift based on use cases and requirements. One example Shakhina shares is specificity. How specific is your user’s intent? Do they have a vague notion of what they need or do they have a clear demand? When you can better understand your searcher’s intent, you can provide better results.
Finding and Discovering Opportunity
In the second part of her talk, Shakhina shares how she approaches the challenge of finding and discovering opportunity and the machine learning models she uses to create solutions.
Understand or Predict Answers
To help users find answers faster, Shakhina recommends using query logs from past searches to personalize their results. Query tagging helps to deliver more precise results by breaking a query down into logical units (such as title, company, location) and matching against these existing entities.
LinkedIn also helps searchers by expanding queries to include synonyms, abbreviations, and name clusters. So say you’re searching for your former coworker Jeff, you’ll still be able to find his profile under his full name, Jeffery. If you’re looking up Kaitlin, you’ll also see results for Caitlin and Kaitlyn.
Find and Present Answers
In order to present the most relevant answers, Shakhina proposes ranking as a technique to bubble up the freshest results and minimize the spammy ones. She also suggests casting a wider (yet personalized) net by displaying blended result to include people, jobs, and posts, particularly for autocompleted searches.
Inspiring More Questions
Finally, the exploration doesn’t have to stop when results are returned. To give searches opportunities to drill deeper, LinkedIn uses tags to help users navigate through content.
Why you should work on Search and Discovery products (at some point)
Search and discovery products challenge you to think outside of your product and connect the dots between and across product lines. You’ll spend time in the nitty, gritty, micro-experiences and exercise your systems thinking skills. You have metrics galore to guide your product and can make a major impact on your business. Maybe it’s not for everyone, but Shakhina would tell you to explore and see what happens.
The post Building Search Products With Machine Learning – Shakhina Pulatova appeared first on Mind the Product.