Amazon Kendra features

Create secure, generative AI-powered conversational experiences for your users on top of your enterprise content. Amazon Kendra provides an optimized Kendra Retriever API that allows you to use Amazon Kendra’s high-accuracy semantic ranker as an enterprise retriever for your Retrieval Augmented Generation (RAG) workflow. The Kendra Retriever API finds and retrieves passages from your enterprise content that are most semantically relevant to the user’s question and have optimized granularity to maximize the quality of your RAG payload - without the need for you to have expertise in accurate semantic retrieval. These optimized passages can then be sent, along with the user’s question, to an LLM to get a generative response. The Kendra Retriever API also includes Kendra features like ACL-based filtering, relevance tuning, metadata-based filtering and more.

Using Amazon Kendra and the new Retriever API provides the following benefits for building your Gen AI experiences:

• Smart chunking of documents: Only send the most relevant passages from your content to the LLM.
• Optimized for RAG: Kendra Retriever API returns the most relevant passages with optimal granularity needed for LLM answer accuracy.
• User ACL filtering: Only return passages that the end-user is entitled to see from your enterprise content.
• Relevance Boosting: Improve LLM answers by boosting specific content based on date, source repository, or any metadata.
• Speed up your Gen AI app development: Get up and running quickly by leveraging the above features instead of building your own enterprise retriever from scratch.

To get started with the Kendra Retriever API, refer to the documentation here and refer to this blog post for tips, best practices and code templates to get started.

Amazon Kendra uses ML to deliver more-relevant answers from unstructured data. Search for general keywords (such as "health benefits") or ask natural language questions ("How long is maternity leave?”) and Amazon Kendra will use reading comprehension to give specific answers ("14 weeks”). For more general questions such as "How do I configure my VPN?" Amazon Kendra gives descriptive answers by extracting the most relevant text passage.

Amazon Kendra also supports FAQ matching and extracts answers from curated FAQs using a specialized model that pinpoints the closest question and returns the corresponding answer. 

Amazon Kendra can even find answers in tables embedded in HTML pages. You can ask questions such as “What’s the credit card with lowest annual fees?” where the answer is found in a credit card comparison table on a marketing webpage.

To complement the Intelligent Search features described above, Amazon Kendra uses a deep learning semantic search model for accurate document ranking. Overall, this provides a richer search experience that presents specific answers, as well as related content to explore if you need more information.

Amazon Kendra uses ML to continuously optimize search results based on end-user search patterns and feedback. For example, when users search for “How do I change my health benefits?" multiple human resources (HR) benefit documents will compete for a top spot. To determine the most relevant document for this question, Amazon Kendra will learn from the user interactions and feedback to promote preferred documents to the top of the list. It applies incremental learning techniques automatically without the need for ML expertise.

You can fine-tune search results and boost specific answers and documents in the results based on specific business objectives. For example, relevance tuning helps you boost results based on more-authoritative data sources, authors, or document freshness. Learn more in our relevance tuning blog post.

To extend Amazon Kendra’s understanding of your specific business vocabulary, you can provide your own custom synonyms. Amazon Kendra uses these to automatically expand queries to include content and answers that match the extended vocabulary. For example, when a user asks the question “What is an HSA?” Amazon Kendra would return documents that reference “Health Savings Account” or “HSA.”

Using connectors with Amazon Kendra is quicker and easier—just add data sources to your Amazon Kendra index and select the connector type. Connectors can be scheduled to automatically sync your index with your data source, so you're always securely searching through the most up-to-date content. Amazon Kendra offers native connectors for popular data sources such as Amazon Simple Storage Service (S3), Microsoft SharePoint, Salesforce, ServiceNow, Google Drive, Confluence, and many more. If a native connector is not available, Amazon Kendra offers a custom data source connector as well as a host of partner-supported connectors. For more information about Amazon Kendra connector availability, visit the Amazon Kendra connector library.

Amazon Kendra uses deep learning models to understand natural language queries and document content and structures for a wide range of internal use cases, including HR, operations, support, and R&D. Amazon Kendra is also optimized to understand complex language from domains such as IT, financial services, insurance, pharmaceuticals, industrial manufacturing, oil and gas, legal, media and entertainment, travel and hospitality, health, news, telecommunications, mining, food and beverage, and automotive. For example, a user searching for HR answers could enter "deadline for filing HSA form" and Amazon Kendra would also search for "deadline for filing health savings account form" for broader coverage to get the most accurate answer.

You can now deploy a fully functional and customizable search experience with Amazon Kendra in a few steps, without any coding or ML experience. Experience Builder delivers an intuitive visual workflow to quickly build, customize, and launch your search application securely on the cloud. You can start with the ready-to-use search experience template in the builder, which can be customized by dragging and dropping the components you want, such as filters or sorting. You can invite others to collaborate or test your search application for feedback, and then share the project with all users when you are ready to deploy the experience. Amazon Kendra Experience Builder integrates with AWS IAM Identity Center (successor to AWS Single Sign-On), supporting popular identity providers such as Azure AD and Okta, delivering secure end-user single sign-on authentication while accessing the search experience. For more information about Amazon Kendra Experience Builder, please visit the documentation.

Amazon Kendra Search Analytics Dashboard helps you to better understand quality and usability metrics across your search applications. The dashboard helps administrators and content creators understand how easily the end users are finding relevant search results, the quality of the search results, and gaps in the content. It provides a snapshot of how your users interact with your search application and how effective your search results are. The analytics data can be viewed in a visual dashboard in the console, or you can build your own dashboards by accessing the data through an API. It can empower you to dive deep into search trends and user behavior to identify insights, and also helps to bring clarity to potential areas of improvement. For more information about Amazon Kendra Search Analytics Dashboard, please visit the documentation.

With Amazon Kendra Custom Document Enrichment capabilities, you can build a custom ingestion pipeline that can pre-process documents before they get indexed into Amazon Kendra. For example, while ingesting content from a repository like SharePoint using our connectors, you can enrich documents with additional metadata, convert scanned documents to text, classify documents, extract entities, and further transform the document using custom ETL processes. The enrichment is performed by rules that can be configured in the console or by invoking functions from AWS Lambda. These functions can optionally call other AWS AI Services such as Amazon Comprehend, Amazon Transcribe, or Amazon Textract. For more information about Amazon Kendra Custom Document Enrichment, please visit the documentation.

Amazon Kendra includes the functionality to autocomplete an end user’s search query. Query autocompletion not only helps you reduce typing by about 25%, but it also helps guide you toward more precise and commonly asked questions. These questions typically result in answers that are more relevant and useful. For example, if you start typing “Where is” in the search box, Amazon Kendra can suggest options to complete the query, such as “Where is the IT desk?” or “Where is the cafeteria?” and other related common questions.