Amazon Kendra

• Fully managed document search service powered by ML
• It’s primary aim is to be designed to mimic interacting with a human expert
• Can extract answers from documents (text, pdf, HTML, PowerPoint, Word, etc.)
• Natural language search capabilities
• Learn from user interactions/feedback to promote preferred results (Incremental Learning)
• Use cases:
  • Hotel FAQ chatbot
  • Hotel can have a lot of FAQ’s
  • Kendra can be used to index the FAQ’s and provide a chatbot to answer questions
  • Example: “What is the check-in time? or do you have a parking lot?”

Key Concepts

• Index: searchable data organized in an efficient way
• Data Source: where the data lives, Kendra connects and indexes from this location. Example of data sources are: S3, Confluence, Google Workspaces, RDS, OneDrive, Salesforce, Kendra Web Crawler, Workdocs, FSX, etc.
• We configure Kendra to synchronize a data source with an index based on a schedule. This should keep the index current
• Documents: can be structured (FAQs) and unstructured (HTML, PDF, etc.)
• Index Document → Kendra(knowledge index powered by ML) → User Question → Kendra → Answer

Example Workflow:

Source