Part 1.2 Understanding the relevance of your search with Elasticsearch and Kibana - Beginner's Crash Course to the Elastic Stack Series - .pdf
- 1. Lisa Jung Developer Advocate @Elastic Beginner’s Crash Course to Elastic Stack Series Part 1.2: Understanding the Relevance of your search using Elasticsearch and Kibana
- 2. Beginner’s crash course to the Elastic Stack Series ● Part 1.1: Intro to Elasticsearch and Kibana ○ use case of Elasticsearch and Kibana ○ the basic architecture of Elasticsearch ○ perform CRUD(Create, Read, Update, and Delete) operations with Elasticsearch and Kibana
- 3. Missed the first workshop? No worries! ● Part 1.1: Intro to Elasticsearch and Kibana ○ Repo: https://ela.st/workshop-1-repo
- 4. The Elastic Stack Reliably and securely take data from any source, in any format, then search, analyze, and visualize it in real time.
- 5. Elasticsearch Store | Search | Analyze
- 7. Elastic is a search company. We focus on value to users by producing fast results that operate at scale and are relevant. This is our DNA. We believe search is an experience. It is what defines us, and makes us unique. Scale, Relevance
- 8. Relevance
- 9. How do we measure relevance? ● Precision ● Recall
- 10. Store | Search | Analyze I store data as documents! Documents with similar traits are grouped into an index! Index Document Document Document Document Document Document
- 11. When search query is sent, Elasticsearch retrieves relevant documents and presents the documents as search results. Index Document Document Document Document Document Document Search Results
- 12. These two diagrams depict the same thing! Index Index Document Document Document Document Document Document
- 13. Index True positives are relevant documents that are returned to the user. T T T T T True positives
- 14. Index False positives are irrelevant documents that are returned to the user. T T T T T True positives F F False positives
- 15. Index True negatives are irrelevant documents that are not returned to the user. T T T T T T T F F True negatives T T T T
- 16. Index False negatives are relevant documents that were not returned to the user. T T T T T T T F F True negatives False negatives T T T T
- 17. What is precision? Precision = True positives True positives + False positives What portion of the retrieved data is actually relevant to the search query?
- 18. What is recall? Recall = True positives True positives + False negatives What portion of relevant data is being returned as search results? T T T
- 19. Precision and Recall are inversely related Precision I want all the retrieved results to be a perfect match to the query, even if it means returning less documents. I want to retrieve more results even if documents may not be a perfect match to the query. Precision Recall Recall Precision Recall
- 20. Precision and recall determine which documents are included in the search results. Precision and recall do not determine which of the returned documents are more relevant compared to the other!
- 21. Ranking refers to ordering of the results (from most relevant results at the top, to least relevant at the bottom). Most Relevant … … … Less Relevant … … … Least Relevant How to form good habits (Highest Score) (Lowest Score)
- 22. What is score? ● The score is a value that represents how relevant a document is to that specific query ● A score is computed for each document that is a hit
- 23. What is score? ● Term Frequency(TF) ● Inverse Document Frequency(IDF)
- 24. What is score? How to form good habits } Search Query Hits Most Relevant … … … Less Relevant … … … Least Relevant (Highest Score) (Lowest Score)
- 25. Term Frequency(TF) determines how many times each search term appears in a document. How to form good habits TF= 4 TF= 1 If search terms are found in high frequency in a document, the document is considered more relevant to the search query. Search Terms } Search Query
- 26. What is Inverse Document Frequency(IDF)? How to form good habits How to form a meetup group Hits Good chicken recipe How to form a band Good times rolling Good habits 101 Good habits are easy to master! We may contain some of the search terms but we have nothing to do with forming good habits! IDF diminishes the weight of terms that occur very frequently in the document set and increases the weight of terms that occur rarely!
- 27. Fine tuning precision or recall using Elasticsearch and Kibana
- 28. Click on the link to the workshop repo. https://ela.st/workshop-2-repo
- 29. Scroll down to the Resources section & open Free Elastic Cloud Trial link in a new tab.
- 30. Scroll down to the Resources section & open Free Elastic Cloud Trial link in a new tab.
- 31. Go to the email account you signed up with and verify your email.
- 32. Open email from Elastic. Click on verify and accept button.
- 34. Click on start your free trial.
- 35. Select the Elastic Stack.
- 37. Name your deployment then create deployment. Beginner’s Crash Course
- 38. Save your deployment credentials.
- 39. Open Kibana. Beginner’s Crash Course Beginner’s Crash Course
- 40. Click on Explore on my own option.
- 41. Click on Upload a file option
- 42. Download and Unzip News Category Dataset from Kaggle
- 43. Drag and drop a file you want to upload.
- 44. Kibana will give you an analysis of the first 1000 lines of your data and give you a summary of your dataset.
- 45. Field section displays fields identified, high level statistics, and top occuring values
- 46. Click on import button
- 47. Name your index and click on import.
- 48. Then Elasticsearch will do the rest! news_headlines
- 49. Click on menu icon, and open Dev Tools.
- 50. Click on Explore on my own option.
- 51. Fine tuning precision or recall using Elasticsearch and Kibana
- 52. Questions?
- 53. Join the Elastic Austin User Group for updates on future workshops!