The Secret Life of Queries:
Parsing, Rewriting & SEO
Rebecca Berbel | Oncrawl
Koray Tuǧberk Gübür | Holistic SEO
slideshare.net/Oncrawl
@RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
^(([^:/?#]+):)?(//([^
/?#]*))?([^?#]*)(?([
^#]*))?(#(.*))?
https://www.google.com
/search?q=search+engine+k
eywords#anchor
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
WHAT
THAT
MEANT
FOR SEO
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
NOT
ANY
MORE.
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
LET'S
TAKE A
LOOK.
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
PARSING
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
Parsing a query
#brightonSEO @RebBerbel @KorayGubur
REWRITING
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
Query segments
#brightonSEO @RebBerbel @KorayGubur
Keyword clustering
#brightonSEO @RebBerbel @KorayGubur
Keywords → Questions
#brightonSEO @RebBerbel @KorayGubur
Question Templates
#brightonSEO @RebBerbel @KorayGubur
Query refinements
#brightonSEO @RebBerbel @KorayGubur
CONTEXT
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
User Intent
#brightonSEO @RebBerbel @KorayGubur
User Intent
#brightonSEO @RebBerbel @KorayGubur
Location/Time
#brightonSEO @RebBerbel @KorayGubur
FINDING
ANSWERS
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
BERT
#brightonSEO @RebBerbel @KorayGubur
Candidate Passages
#brightonSEO @RebBerbel @KorayGubur
Candidate Passages
#brightonSEO @RebBerbel @KorayGubur
Candidate Passages
#brightonSEO @RebBerbel @KorayGubur
PHEW.
LET'S
TAKE A
STEP BACK.
#brightonSEO @RebBerbel @KorayGubur
THERE'S MORE
(but we'll stop here)
#brightonSEO @RebBerbel @KorayGubur
GOOGLE DOES NOT USE
THE TEXT YOU TYPED
TO SEARCH FOR
RESULTS.
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
WHAT
THAT
MEANS
FOR SEO
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
BUILD
TOPICAL
RELEVANCE
Oncrawl / Koray
#brightonSEO @RebBerbel @KorayGubur
REMEMBER
NEURAL
MATCHING
Danny Sullivan (Twitter)
#brightonSEO @RebBerbel @KorayGubur
GROUP
TARGETED
QUERIES
Stefan Neefischer
#brightonSEO @RebBerbel @KorayGubur
THINK
SEARCH
INTENT
Johanna Maier / Lidia Infante / Lazarina Stoy
#brightonSEO @RebBerbel @KorayGubur
DOES IT
WORK?
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
#brightonSEO @RebBerbel @KorayGubur
Semantic networks & topical relevance
#brightonSEO @RebBerbel @KorayGubur
Semantic networks & topical relevance
10K to 70K
clicks per month
(5 months)
10K to 200K+
clicks per month
(5 months)
0 to 330K+
clicks per month
(6 months)
+400%
growth
(6 months)
#brightonSEO @RebBerbel @KorayGubur
Thank you!
@RebBerbel
rebecca@oncrawl.com
@KorayGubur
Holistic SEO
ktgubur@holisticseo.digital
#brightonSEO @RebBerbel @KorayGubur

Search Query Processing: The Secret Life of Queries, Parsing, Rewriting & SEO

Editor's Notes

  • #5 http://norvig.com/ - Peter Norvig, Director of Research at GOogle (Levenshtein Distance)
  • #6 2011 / 2013
  • #7 Keyword stuffing
  • #8 Misspellings - 10% of queries ? Pre-2014 tool for "PPC" Digg, August 2008
  • #9 2011 / 2013
  • #11 2011 / 2013
  • #12 2011 / 2013
  • #14 2011 / 2013
  • #20 2011 / 2013
  • #21 Probabilistic Graphical Model for the Keywords to Questions system’s “User Inquiry Intent Model” (K2Q)
  • #22 Probabilistic Graphical Model for the Keywords to Questions system’s “User Inquiry Intent Model” (K2Q)
  • #23 This is by Jakob Uszkoreit, whose work today is almost exclusively on attention-based transformers (NLP)
  • #24 2011 / 2013
  • #28 "Candidate answer passages are generated from both structured content and unstructured content according to corresponding selection criteria. This allows the user to not only receive prose-type explanations but also to receive a combination of prose-type and factual information, which, in turn, may be highly relevant to the user's informational need."
  • #30 MuM "Neural matching" SERP features Information retrieval processes Cost optimization
  • #32 2011 / 2013
  • #33 Putting keywords on a page is not sufficient LSI keywords don't exist Websites are evaluated as a whole, not page-by-page