What does it really mean to deliver an "AI-powered Search" solution? In this talk, we’ll bring clarity to this topic, showing you how to marry the art of the possible with the real-world challenges involved in understanding your content, your users, and your domain. We'll dive into emerging trends in AI-powered Search, as well as many of the stumbling blocks found in even the most advanced AI and Search applications, showing how to proactively plan for and avoid them. We'll walk through the various uses of reflected intelligence and feedback loops for continuous learning from user behavioral signals and content updates, also covering the increasing importance of virtual assistants and personalized search use cases found within the intersection of traditional search and recommendation engines. Our goal will be to provide a baseline of mainstream AI-powered Search capabilities available today, and to paint a picture of what we can all expect just on the horizon.

1. Trey Grainger
Chief Algorithms Officer
Lucidworks
Next Generation
of
The

2. What is
?

3. AI?
Machine Learning?
Data Science?
Neural Search?
AI-Powered Search?
Deep Learning?

5. http://aiPoweredSearch.com
... is my new book!

6. Keyword
Search
Knowledge Graph
User Intent
Personalized
Search
Semantic
Search
Domain-aware
Matching
The Goal of AI-powered Search
Content
Understanding
Domain
Understanding
Collaborative
Recommendations User
Understanding

7. image credit: SAS Institute, 1998
Definitions, circa 1998

15. AI-powered Search

16. AI-powered Search
Question / Answer
Systems
Virtual Assistants
• Signals Boosting Models
• Learning to Rank
• Semantic Search
• Collaborative Filtering
• Personalized Search
• Content Clustering
• NLP / Entity Resolution
• Semantic Knowledge Graphs
• Document Classification
• etc.
• Neural Search
• Word Embeddings
• Vector Search
• Image / Voice Search
• etc.
• Question / Answer Systems
• Virtual Assistants
• Chatbots
• Rules-based Relevancy
• etc.

17. Proudly built with open-source
tech at its core: Apache Solr &
Apache Spark
Personalizes search
with applied
machine learning
Proven on the
world’s biggest
information systems

18. Query Intent
(query classification, semantic query parsing,
semantic knowledge graphs, concept expansion,
automatic query rewrites, clustering,
personalization, question/answer
systems, virtual assistants)
Search Intelligence Spectrum
Basic Keyword Search
(inverted index, tf-idf, bm25,
multilingual text analysis,
query formulation, etc.)
Taxonomies / Entity Extraction
(entity recognition, taxonomies,
ontologies, business rules,
synonyms, etc.)
Automated Relevancy Tuning
(Signals boosting, AB testing / multi-
armed bandits testing, signals-based
back-testing, genetic algorithms,
Learning to Rank)
Self-learning

19. People to
People
People to
Information
People to
Insights
Search + AI=

21. Watson: “You appeared to [see a good deal] which was quite invisible to me”
Sherlock: “Not invisible but unnoticed, Watson. You did not know
where to look, and so you missed all that was important.”
The Adventures of Sherlock Holmes, ADVENTURE III. A CASE OF IDENTITY, Sir. Oliver Conan Doyle

22. Head?
Pipe?
Coat Collar? Back of Hat?
Hat?
Smoke?
Nose?
Abstract Concept of
Detective with Pipe
Specific hypothesis from Experience (leveraging social cue that this is probably a well-known answer)
Detective (Deerstalker) Hat!
Final Answer + conceptual context

23. Natural Language Understanding

25. Ok, Google…
Is Agave Nectar good for you?

30. How can we help

31. =
=
=
=

32. Significance of Feedback Loops
User
Searches
User
Sees
Results
User
takes an
action
Users’ actions
inform system
improvements

33. Signals
Boosting
Popularized Relevancy
Learning
to Rank
Generalized Relevancy
Collaborative
Filtering
Personalized Relevancy

34. Signals
Boosting
Popularized Relevancy
Learning
to Rank
Generalized Relevancy
Collaborative
Filtering
Personalized Relevancy

35. Signal Collection & Processing
User
Searches
User
Sees
Results
User
takes an
action
Users’ actions
inform system
improvements
User Query Results
Alonzo ipad doc10,
doc22,
doc12, …
Elena printer doc84,
doc2,
doc17, …
Ming ipad doc10,
doc22,
doc12, …
… … …
User Action Document
Alonzo click doc22
Elena click doc17
Ming click doc12
Alonzo purchase doc22
Ming click doc22
Ming purchase doc22
Elena click doc2
… … …
ipad ⌕
Signal Processing
& Machine Learning
Learned Relevance Models

36. ipad

38. ipad

39. Crowdsourced
Relevancy
Crowdsourced
Biases

42. Still Available
through
Query Variations
Manual Override
By Facebook

43. ipad
Bananas

44. ipad
Bananas

49. Risks Potential Solutions
Reinforces Current Biases • Learning to Rank
• Use-case-specific Click Models
• User-group bias detection
• Inject conceptual diversity
Subject To Manipulation • Session Filtering
• Quality vs. Quantity Signal Weighting
• Only Count Explicit Actions
Cold Start Problem • Content-based Relevance
• Learning to Rank
Signals Boosting + Collaborative Filtering

50. Knowledge
Graphs
Domain modeling
Multi-modal
Learning
Merged Content Modalities
Thought
Vectors
Conceptual scoring

51. Learned Knowledge Graphs
Trey Grainger works for Lucidworks.
He is speaking at the Activate 2019
conference.
#Activate19
(Activate) is being held in Washington, DC
September 9-12, 2019.
Trey got his masters degree from
Georgia Tech.
Trey’s Voicemail

52. Learned Knowledge Graphs
Trey Grainger works for Lucidworks.
He is speaking at the Activate 2019
conference.
#Activate19
(Activate) is being held in Washington, DC
September 9-12, 2019.
Trey got his masters degree from
Georgia Tech.
Trey’s Voicemail

53. Multimodal Learning

54. Phrase: Vector:
apple juice: [ 1, 5, 0, 0, 0, 4, 4, 3 ]
cappuccino: [ 0, 5, 3, 0, 4, 1, 2, 3 ]
cheese bread sticks: [ 5, 0, 4, 5, 0, 1, 4, 2 ]
cheese pizza: [ 5, 0, 4, 4, 0, 1, 5, 2 ]
cinnamon bread sticks: [ 5, 0, 4, 5, 0, 1, 4, 2 ]
donut: [ 5, 0, 1, 5, 0, 4, 5, 1 ]
green tea: [ 0, 5, 0, 0, 2, 1, 1, 5 ]
latte: [ 0, 5, 4, 0, 4, 1, 3, 3 ]
soda: [ 0, 5, 0, 0, 3, 5, 5, 0 ]
water: [ 0, 5, 0, 0, 0, 0, 0, 5 ]
Ranked Results: Green Tea
0.94 water
0.85 cappuccino
0.80 latte
0.78 apple juice
0.60 soda
… …
0.19 donut
Vector Similarity Scores:
Vector Similarity (a, b):
a · b = |a| × |b| × cos(θ)
Ranked Results: Cheese Pizza
0.99 cheese bread sticks
0.91 cinnamon bread sticks
0.89 donut
0.47 latte
0.46 apple juice
… …
0.19 water

55. Sentence Embeddings:
[ 2, 3, 2, 4, 2, 1, 5, 3 ]
[ 5, 3, 2, 3, 4, 0, 3, 4 ]
. . .
Document Embedding:
[ 4, 1, 4, 2, 1, 2, 4, 3 ]
Word Embeddings:
[ 5, 1, 3, 4, 2, 1, 5, 3 ]
[ 4, 1, 3, 0, 1, 1, 4, 2 ]
. . .
Paragraph Embeddings:
[ 5, 1, 4, 1, 0, 2, 4, 0 ]
[ 1, 1, 4, 2, 1, 0, 0, 0 ]
. . .
Thought Vectors

56. Keyword
Search
Collaborative
Recommendations
Knowledge Graph
User Intent
Personalized
Search
Semantic
Search
Domain-aware
Matching
The Goal of AI-powered Search
Content
Understanding
User
Understanding
Domain
Understanding

57. The Next
Generation of
AI-powered Search

59. Trey Grainger
trey@lucidworks.com
@treygrainger
Other presentations:
http://www.treygrainger.com http://aiPoweredSearch.com
Books: