My presentation is split up into 4 parts, summary below:
1. Semantic search
a. What is it? (definition from my latest SEW article)
b. 4 types of semantic search
c. First engines to use it
2. Mobile
a. Why does Google Now exist?
b. What happens when we are bored on a mobile and its impact on search?
3. Keywords
a. How voice search impacts exact phrase match
b. Stating that although informational, navigational, transactional and connectivity queries have strength and merit we need to think about cognition and search, and thus, identify new keyword categories
4. Future of search with Hummingbird and AI in mind
a. Will engines soon be able to detect the searcher’s sweat glands so they can gain more insight into their emotional state (e.g. anxious) and adjust results accordingly?
3. Sections of this presentation
The real reason why
Google Hummingbird exists
Semantic search
Mobile
Future of search
Keywords
4. Why does Google Hummingbird exist?
Usability principles
User's
language
Consistency
Minization of
user memory
load
Flexibility
and efficiency
of user
Aesthetic and
minimalist
design
Chunking
Progressive
levels of
detail
Navigational
feedback
6. What is a search engine?Definitionone
Utility tool
used to locate
web sites on
the web Definitiontwo
A searcher
driven
program
offering
unique
features to
build and find
information
Definitionthree
Most popular
method of
finding
information
7. One of the most demanding
challenges for modern search
engines is understanding
search intent
8. Copyright free image
Truncation example
Swimming
Swimwear
swimmers
• It is Google’s algorithm
“the whole shebang”
• Came out in August 2013
• Upgrade of: index
stemming (truncation by
the user) and synonyms
• Tries to understand
keyword intent
Swim* =
15. Types of semantic search engines
• Resource Description Framework Schema -- RDF(S)
• This is what powers Knowledge Graph and Bases
› Proper term for Google Knowledge Graph is Conceptual Graphs
• Uses command to enhance results, such as:
› isInterestedBy
› hasForWorkInterest
› hasForPersonalInterest
RDF-based querying languages
Closely linked
with location
Length
Breadth
Height
16. Why did the major engines sign up to Schema?
To get a head start on semantic…voluntarily tell the engine
17. Types of semantic search engines
• Looks at string distance metric, generic lexical resources, such as WordNet
plus structure of the input ontology
• Heavily replies on Machine Learning (which promoted Google RankBrain)
Question-answering tools
18. Types of semantic search engines
• Joining the dots, then learning from them
• Distributed extensibility is a very important part of semantic search
Semantic-based keyword search
19. Knowledge Graph and Bases
• Presenting media objects to the user using an additional SERP snippet or
two
› Encouraging the searcher to stay on the SERP
• Knowledge Graphs and Bases link up media objects:
› E.g. webpages, images, audio clips, social media profiles
• AI is helping entities go in the right direction
20. Google Hummingbird…
• Google Hummingbird is about enhancing synonyms
› It is not semantic search
• It is a new engine of a car or new wheels of a bike, if you like, in order to
improve performance (speed and precision)
• It is helping to improve Google Knowledge Graph
…as it stands
21. Concepts familiar to the user
How do I get to the nearest restaurant?
Semantic search
Keyword filtering1
Required
& optional keyword
2
Subject keyword is
semantically searched
3
User profile
22. Geo
Concepts familiar to the user
How do I get to the nearest restaurant?
User profile
Real semantic search
By means
of
By what
method
By whose
help
From what
source
Synonyms = what Google Hummingbird is currently working on
25. Keywords
Use of words and phrases
iPhone
Mobile technology
explosion
1 or 2 words?
What the hell
happened?
26. Voice search is going to diminish the
searcher’s use of incorrect keywords
and therefore impact exact phrase
keywords
27. Categorising search queries
• When we search we often think of things we do not type
Samsung Galaxy Edge 7
Cognition
subjective reviews on
comprehensive information
Exhaustive Comprehensible Objective
Subjective Concrete Abstract
We often think
with these
keywords but
we do not type
them
28. Semantically, renaming search queries
Direct
[subjective 5 star
hotel in Miami]
Transformed
[review latest
blackberry handset]
None
[blackberry]
No more informational, navigational, transactional, connectivity
0
0.5
1
1.5
2
2.5
3
3.5
Nouns (unique) Nouns
(overlapping)
Verbs Adjectives
Difference between typed and voice
search queries and their component
words
Typed queries
Voice search
Transactional Informational
30. Mobiles
• In 1991, Mark Weiser, a US scientist, coined the terms:
› Inch-sized computer
› Mobile
› Foot-sized computer
› Tablet
› Yard-sized computer
› Web-enabled TV
Origin
Organisation for Economic Co-operation and Development
said in June 2010 mobile had:
1.5 times fixed broadband
31. Mobiles
• Most personal and confidential piece of technological device
• Used in active or personal contexts and activities in a natural and dynamic
way
• Used in a variety of situations:
Today
Rush
While
commuting
To fill idle
time
While
queuing
At home
Comfortably
sitting on the
sofa
32. What we know about mobile
Lots of things, or is it?
Visually the same
as desktop
Size of screen
does not matter
Social activity
Reduces button
tapping accuracy
by 30%
Clutter free
High-end mobiles
are similar to
desktop searchers
Good
abandonment is
higher
~70% of searches
in work or at
home
Mostly static
searches
Under-
researched
area
33. Mobile has come a long way!
• Blood pressure monitors
• Skin conduction
• Respiration sensors
In 2001
81% accuracy
34. Boredom and mobile interaction
.
More indicative
of boredom
Switching phone on
Changing screen orientation
Not so
indicative
Social network notifications
Frequency of open notification centre
Change screen status
App launches
Charging time
Amount of transmitted data
Activesearch
forstimuli
Analysing if we are bored
35. Google Now
• There:
› to support task continuation
› help bored searchers
Purpose for mobile search
Boredom state
I don’t like,
I’m bored
Boredom trait
I used to
like, but
now I’m
bored
36. Boredom and search
• Mobile has an ephemeral nature
.
Evoked by an urgent
need
When is the next bus home?
Location-based filtering
Semantic tools
37. Boredom and search
• Mobile has an ephemeral nature
.
Triggered by desire to
fill idle time
Funniest cinema movies
Mobile-tailored content
Social tagging
38. Boredom and search
• Mobile has an ephemeral nature
.
Prompted by an
event, situation, no
need to fill
Which cinema shows the film I just
seen the announcement about?
Ephemeral need, response is not yet
needed
39. The future
Linking technology with the searcher
Link-less
world
Mobile
Conversational
search
Artificial
Intelligence
Sweat glands
Technology
Predictive
search
Semantics
Internet of
Things
Semantics,
technology and
mobile
Snippet length
Query type,
mobile
40. References
• Kato, M.P., Yamaoto, T., Ohshima, H., and Tanaka, K. Cognitive search intents hidden behind queries: A
user study on query formulations. WWW ‘14 Companion
• Schilit, B. N. Mobile computing: Looking to the future. Computer.
• Gomez-Barroso, J. L. Factors required for mobile search going mainstream. Online Information Review.
36(6)
• Xu, Z., Luo, X., Yu, J., and Xu, W. Measuring semantic similarity between words by removing noise and
redundancy in web snippets. John Wiley & Sons Ltd.
• Matic, A., Pielot, M. and Oliver, N. Boredom-computer interaction: Boredom proneness and the use of
smartphone.
• Lane, N., Lymberopoulos, D., Zhao, F., Campbell, A. T. Hapori: Context-based local search for mobile
phones using community behavioural modeling and similarity. Ubicomp.
• Goldsten, J., Kantrowitz, M., Mittal, V., and Carbonell, J. Summarising text documents: Sentence and
evaluation metrics.
• Crossan, A., Murray-Smith, R., Brewser, S., Kelly, J., Musizza, B. Kelly, S.Gait phase effects in mobile
interaction. CHI 2005.
• Vang, K. J. Ethics of Google’s Knowledge Graph: Some considerations. Journal of Information,
Communication and Ethics in Society.
• Pielot, M. Dingler, T. Pedro, J. S., and Oliver, N. When attention is not scarce – detecting boredom from
mobile phone usage. Ubicomp.
• Kamvar, M. and Baluja, S. A large scale study of wireless search behavior: Google mobile search. CHI.
• Lei, Y., Uren, V., Motta, E. SemSearch: A search engine for the semantic web.
Key readings
41. Summary
Search
originated in
1945
Hummingbird
is the actual
algorithm
Form-based
RDF-based
querying
languages
Question-
answering
Semantic-
based
Keywords and
voice search
Direct Transformed
None
Mobile and
boredom
Google Now Future
Hello, thanks for staying. I’m dying for another drink too…
My name is Gerald Murphy. I’m a paid Associate Lecturer of information retrieval in one of the UK’s oldest search departments.
I also write for the long-standing search engine marketing website in the world, Search Engine Watch
And I work with the coolest brands in BrightEdge, singlehandedly the best integrated SEO platform since we have partnerships with MajesticSEO, Facebook and Twitter
As you can tell from the background picture, and my beautiful accent, I’m a Belfast boy. The city that built the unsinkable ship that ended up sinking
Rather than do a boring presentation on just semantic search, I’ve split today’s presentation into 4 parts:
Semantic search
Keywords
Mobile
Future of search
We’ll do this by weaving in solid understanding from the 8 usability principles
V. Bush
Automated bots
Web inception
Engines to organize mess
Loads of definitions of web search…
But, one thing we can agree on is…
…[read]
Google Hummingbird is here to help to understand KW intent
It’s not a ranking factor, it’s the actual algorithm
It came out over 2 and a half years ago
Babysitting anology
There are 4 types, let’s take a look
The first is SHOE
Think forms and databases
And think Schema
Google filed this Patent before they announced Hummingbird.
It splits the KW up into forms, but the index needs to be split and semantically linked. Something Google is unlikely to have
Second, RDF
This is what powers Google Knowledge Graph
Proper term is conception graphs
Unlike SHOE, it uses commands via categories
Closely linked to location and personalization
Both SHOE and RDF rely on additional code, basically Schema, it is no wonder why all the major engines have agreed on Schema code
Question-answering tools, are essentially several dictionaries and encyclopedia
Think about to your time at school. You were given a reading list of various books. Each book had additional and enhancement information in it. This is what question-answering semantics is like.
Since we had to go and verify if our knowledge was correct, just as a uni test would do, AI is used to replace the exam/test
Fourthly, and finally, there is an array of information on the existing web and semantic-based keyword search. Engines scrape this info and learn from references or citations.
Once a KW is searched, it is matched off these citations to form a semantic search.
KG is not semantic search.
It is there to link media objects, which web search engines have always been able to do.
It is also there to keep us on the engine for longer. It’s relevant and sub consciously reinforces quality information.
So, Hummingbird is all about enhancing synonyms.
It’s not semantic search. Ask and engine for a babysitter or get a native speaker to play around with Translate, it’s a mess
Search will slowly move towards semantic search filtering.
Here’s a simplified version:
Long tail keyword is entered
Engine splits it into 2, required and optional
Subject KW is semantically searched
7 factors influence our choice of KWs
For semantics, 2 are particularly important…linguistics and voice search
[explain]
[read]
We think different things than we type.
Let’s say we want to buy a new phone. We often think like ‘subjective’ but we don’t write/type it.
Semantic engines should factor this in.
Today I present a newer version of ‘informational, navigational, transactional and connectivity’
They are direction, explicitly typing what we think
Transformed – some use of thinking
None, staying with old-school search patterns
Voice search uses significantly less nouns compared to typed KWs
First thing’s first, the term mobile is misleading. Most searches occur when we are at home or at work, we’re general static when we carry out a search
Mobile are utilities, much like: gas, water and electric. This is why Google said that 91% of people have their mobile in a 3 meter radius 100% of the time
Although we are starting to understand the human factors of mobile, mobile, as an area of research, is still in its infancy
Boredom is not someone who has nothing to do, it is someone who is actively looking for stimulation but unable to find something stimulating
http://metro.co.uk/2015/04/28/how-many-trees-would-it-take-to-print-every-single-page-of-the-internet-5171497/
http://www.physics.le.ac.uk/jist/index.php/JIST/article/view/100/57
This year the web is only 26 years of age
In 2014, 40% of people in the world are using the web
68 billion