Video and slides synchronized, mp3 and slide download available at URL http://bit.ly/2juApkk. Daniel Tunkelang talks about what search looks like when viewed through a query understanding mindset. He focuses on query performance prediction, query rewriting, and search suggestions. Filmed at qconsf.com. Daniel Tunkelang is a data science and engineering executive who has built and led some of the strongest teams in the software industry. He's currently working on a stealth startup while advising various companies on search and discovery. Previously, he worked for Endeca as a chief scientist, led a local search team at Google and was director of data science and engineering at LinkedIn.

1. Query Understanding:
A Manifesto
Daniel Tunkelang
queryunderstanding.com

2. InfoQ.com: News & Community Site
• 750,000 unique visitors/month
• Published in 4 languages (English, Chinese, Japanese and Brazilian
Portuguese)
• Post content from our QCon conferences
• News 15-20 / week
• Articles 3-4 / week
• Presentations (videos) 12-15 / week
• Interviews 2-3 / week
• Books 1 / month
Watch the video with slide
synchronization on InfoQ.com!
https://www.infoq.com/presentations/
query-manifesto

3. Purpose of QCon
- to empower software development by facilitating the spread of
knowledge and innovation
Strategy
- practitioner-driven conference designed for YOU: influencers of
change and innovation in your teams
- speakers and topics driving the evolution and innovation
- connecting and catalyzing the influencers and innovators
Highlights
- attended by more than 12,000 delegates since 2007
- held in 9 cities worldwide
Presented at QCon San Francisco
www.qconsf.com

4. Overview
• What is query understanding?
• Query performance prediction.
• Query rewriting.
• Query suggestions.
• Search is a conversation.
tl;dr: Query understanding is about focusing
less on the results and more on the query.

5. Query understanding means
figuring out what the searcher wants.

6. What query understanding is not:
scoring and ranking of results.

7. Query understanding means not relying
on ranking to filter out irrelevant results.

8. Challenge: accurately identify searcher's intent
while minimizing searcher’s effort.

9. Old-school search:
searcher enters query, search engine interprets query.

10. Modern search:
searcher starts to type, search engine suggests queries.

11. And in 2016,
modern search means mobile search.

12. March Hare: “You should say what you mean.”
Alice: “I do. At least I mean what I say.”

13. But what if the search engine doesn’t
know what the searcher means?

14. And how does a search engine achieve the
self-awareness to know what it doesn’t know?

15. Query understanding is about measuring
and optimizing for query performance.

16. History is the best teacher.
For common queries, learn from historical performance.

17. How you define performance depends on why
searchers use your site. But define it and measure it!

18. Can also learn from similar queries:
re-ordered terms, subqueries, similar results, etc.

19. Stuck in the long tail?
Predict performance from query and result set features.

20. Lots of academic work on query performance prediction.

21. So you have a poorly performing query. Now what?
Rewrite it!

22. Rewriting to improve precision:
query segmentation, field restriction, query classification.
tag: skill OR title
related skills:
search, ranking, …
tag: company
id: 1337
industry: internet
verticals:
people, jobs
intent: exploratory

23. Rewriting to improve recall:
query expansion, query relaxation.

24. When query expansion is done right, users barely notice.
Query relaxation is risker and should be messaged.

25. Query rewriting should improve query performance.
If it doesn’t, you’re doing wrong. Measure!

26. Rewriting is great when searchers type whole queries.
But what about autocomplete and search suggestions?

27. Query suggestions have two jobs:
reduce searcher effort and improve query performance.

28. Searchers are people. People are lazy.
All else equal, help them type as little as possible.

29. But don’t let searchers be too lazy!
It’s easy to create miscommunication.

30. Search suggestions offer strong information scent.
But searcher may not recognize query ambiguity.

31. Reduce the likelihood of a bait-and-switch:
only offer high-performance search suggestions.

32. Should you score search suggestions based
on total engagement or query performance?

33. Total engagement is the value delivered by search.
But low query performance means unhappy users.

34. Can combine as single utility function. Or score using total
engagement but require minimum query performance.

35. Finally, let’s remember that search is more than
a single query and response. It’s a conversation.

36. Don’t focus so much on single-query performance
that you forget to optimize for the overall journey.

37. The best journey is a sequence of small successes.
Not a sequence of frustrations followed by success.

38. Summary
• Query understanding means figuring out what the
searcher wants.
• Measure and optimize for query performance.
• Use query rewriting to improve precision and recall,
in order to improve query performance.
• Query suggestions have two jobs: reduce searcher effort
and improve query performance.
• Search is a conversation.
tl;dr: Query understanding is about focusing
less on the results and more on the query.

39. Thank you! Learn more at queryunderstanding.com!

40. Watch the video with slide
synchronization on InfoQ.com!
https://www.infoq.com/presentations/
query-manifesto