Optimizing Unstructured Data
25 likes•12,829 views
The document discusses the complexities of semantic search and the importance of understanding meaning and context in unstructured data. It emphasizes techniques such as named entity recognition, coreference resolution, and machine learning to identify salient entities and improve search optimization. The text argues for focusing on user intent and structured data to enhance content visibility to both search engines and users.
Optimizing Unstructured Data
- 3. My name is AJ Kohn
- 4. Blind Five Year Old Since 2007
- 7. We have a problem
- 8. Ugh, as if!
- 10. WHAT?!
- 12. WHAT?!
- 14. Semantic search is about understanding meaning
- 15. OKAY!
- 16. OKAY!
- 17. Context
- 18. Context matters
- 19. Context matters
- 21. Finding all expressions that refer to the same entity in a text Coreference Resolution Part of Speech (POS) Tagging Assign a part of speech to each word in a text
- 22. The word quiet isn’t spelled wrong but Google knew that I probably meant to write quite awesome instead
- 23. Machine learning
- 24. Making predictions based on patterns and rules from prior data
- 26. Google is better at getting meaning from text because of access to more data
- 28. Entities
- 30. Things
- 31. “New York” hasPopulation: 8.046 Million hasPointsofInterest: Empire State Building hasAddress: 350 5th Avenue hasHeight: 1,250 feet
- 33. Connections and relationships between entities and documents
- 34. Named Entity Recognition (NER)
- 35. One size doesn’t fit all
- 36. Context-Dependent Fine-Grained Entity Type Tagging
- 37. Not just any entities but salient entities
- 38. 66 entities on a page and less than 5% are salient http://bit.ly/bigdealentities
- 39. How do you train a machine learning model to identify salient entities?
- 40. Hello McFly!
- 41. Word up
- 43. Words
- 44. “Keywords don’t matter anymore”
- 45. Ice Bear cried, but just inside
- 46. I love structured data but optimizing unstructured data is far more powerful
- 47. Text on the page is more important now
- 48. Words = Entities ^ Context ^ Meaning
- 49. We can turn unstructured content into structured data
- 50. How much do you trust Google? How much do you trust Google?
- 51. Stop writing for people and start writing for search engines http://bit.ly/focusedwriting
- 53. 28%
- 54. Most users don’t read but skim and scan instead http://bit.ly/usersdontread
- 55. First you looked here Then here
- 56. A penny for a paragraph return
- 57. Mirroring
- 58. Not only do we mirror body language we seek it out when searching
- 60. Keyword rich text and subheads allow users to resume reading at any time
- 61. Keyword is not a four letter word
- 62. Better to you query syntax call it
- 63. But what about user delight?
- 64. Could you not
- 65. Task Completion > Aesthetics
- 66. Our job is to reduce friction
- 67. After writing your content go back and find where you can replace pronouns with nouns Remember that readers won’t often ‘see’ these nouns but will use them as visual signposts
- 68. “It’s such a gorgeous work of art” “Lobster and Cat is a beautiful painting” ArtworkType: paintingArtworkTitle: Lobster and Cat hasArtist: Pablo Picasso
- 70. Intent
- 71. Google may better understand the meaning of my query but do they know why I’m searching?
- 72. Why are they really searching?
- 73. Why are they really searching? Common Problems with the Eureka 4870 Eureka 4870 Troubleshooting Tips Local Vacuum Cleaner Repair Shops Eureka 4870 Replacement Parts Guide to Buying a New Vacuum Cleaner
- 74. Why are they really searching? Common Problems with the Eureka 4870 Eureka 4870 Troubleshooting Tips Local Vacuum Cleaner Repair Shops Eureka 4870 Replacement Parts Guide to Buying a New Vacuum Cleaner
- 75. Our job is to decode the intent from the query syntax http://bit.ly/aggregatingintent
- 77. What are we really talking about?
- 78. This is a factbox triggered by entities and the Knowledge Graph
- 79. This answerbox is triggered by semi-structured data
- 80. This answerbox is triggered by specific patterns of text
- 81. Answerbox triggered by patterns of text and specific understanding
- 82. Answerbox triggered by patterns of text and specific understanding
- 83. Answerbox triggered by patterns of text and semi-structured data
- 84. Answerbox triggered by patterns of text and specific understanding
- 85. Game's the same, just got more fierce
- 86. Skate to where the puck is going to be, not to where it has been
- 87. The Link Graph
- 88. The Link Graph + Scored Entities <entity A> <entity B> <entity C> <entity B> <entity C> <entity A> <entity A> <entity D> <entity B> <entity D>
- 89. Entity authority could flow through links similar to anchor text
- 90. TL;DL
- 91. We can help Google to find structure, entities and meaning in our content The easier we make it, the more likely we are to satisfy robots and humans
- 92. AJ Kohn Owner, Blind Five Year Old www.blindfiveyearold.com aj@blindfiveyearold.com @ajkohn