Amazon Kendra Deprecated2019年〜
An intelligent enterprise search service powered by machine learning
What It Does
Amazon Kendra is an intelligent search service that returns relevant documents and answers when you ask questions in natural language. It indexes documents from diverse data sources including S3, SharePoint, Salesforce, and databases, and uses machine learning to improve search accuracy.
Use Cases
Used for internal knowledge base search, customer support FAQ search, legal document search, cross-repository technical documentation search, and compliance document search.
Everyday Analogy
Think of it like an excellent librarian. When you ask 'Where's last year's sales report?', they search through a vast collection (data sources), find the relevant document, and present it with the key passages highlighted.
What Is Kendra?
Amazon Kendra is an intelligent search service for enterprises. Instead of keyword-based search, it provides search based on natural language understanding, extracting direct answers from documents for questions like 'What were last month's sales?'
Data Sources and Indexes
Kendra provides over 40 data source connectors. It automatically crawls and indexes documents from S3, SharePoint, Confluence, Salesforce, RDS, ServiceNow, and more. Index updates can be automated with scheduling, so changes in data sources are reflected in search results. To deepen your understanding of data sources and indexes, related books (Amazon) are also helpful.
Getting Started
Create an index in the Kendra console and add data sources. Once data source synchronization is complete, you can try natural language queries in the search console. Integrate the Search API into your application to build a custom search UI. You can also create a no-code search page with Kendra Experience Builder.
Things to Watch Out For
- Two editions are available - Developer Edition and Enterprise Edition - with different pricing and features. Enterprise Edition supports high availability and large-scale data
- Charges are based on index hourly billing and query count. May be excessive for small-scale search use cases