Q: What is Amazon Kendra?
A: Amazon Kendra is a highly accurate and easy to use enterprise search service that’s powered by machine learning. Kendra enables developers to add search capabilities to their applications so their end users can discover information stored within the vast amount of content spread across their company. This includes data from manuals, research reports, FAQs, HR documentation, customer service guides, and is found across various systems such as file systems, web sites, Box, DropBox, Salesforce, SharePoint, relational databases, Amazon S3, and more. When you type a question, the service uses machine learning algorithms to understand the context and return the most relevant results, whether that be a precise answer or an entire document. For example, you can ask a question like "How much is the cash reward on the corporate credit card?” and Kendra will map to the relevant documents and return a specific answer like “2%”. Kendra provides sample code so that you can get started quickly and easily integrate highly accurate search into your new or existing applications.
Q: How does Amazon Kendra work with other AWS services?
A: Amazon Kendra provides ML-powered search capabilities for all unstructured data customers store in AWS. Kendra offers easy integration with popular repository types such as file systems, web sites, Box, DropBox, Salesforce, SharePoint, relational databases, Amazon S3 so that you can index your content with a few clicks, and search your information without writing a single line of code.
Q: What types of questions can I ask Amazon Kendra?
- Factoid questions (who, what, when, where): “Who is Amazon’s CEO?” or “When is Prime Day 2019?”. These require fact-based answers that may be returned in the form of a single word or phrase. The precise answer, however, must be explicitly stated in the ingested text content.
- Descriptive questions: “How do I connect my Echo Plus to my network?” or “How do I obtain tax benefits for lower income families?” where the answer could be a sentence, passage, or an entire document.
- Keyword searches: “Fitbit waterproof” or “Honda warranty”, where the intent and scope are not very clear. Kendra will leverage its deep learning models to return relevant documents in these cases.
Q: What if my data does not contain the precise answer Amazon Kendra is looking for?
A: When your data does not contain a precise answer to a question, Kendra returns a list of the most relevant documents ranked by its deep learning models.
Q: What types of questions will Amazon Kendra be unable to answer?
A: Kendra does not yet support questions for which the answers require cross-document passage aggregation or calculations.
Q: How do developers get up and running with Amazon Kendra?
A: The Kendra console provides the easiest way to get started. You can point Kendra at unstructured and semi-structured documents like FAQs stored in S3. After ingestion, you can start testing Kendra by typing queries directly in the “search” section of the console. You can then deploy Kendra search in your own application with a few lines of code, or by copying short code samples from the reference implementations to replicate specific functionality.
Q: How can I customize Amazon Kendra to better fit my company’s domain or business specialty?
A: Kendra offers domain specific expertise for IT, pharma, insurance, oil and gas, industrial, financial services, legal, media and entertainment, travel and hospitality, health, human resources, news, telecommunications, mining, food and beverage, and automotive. You can extend beyond this by providing your own synonym lists (coming soon). You just upload a file with your specific terminology and Kendra will use these synonyms to enrich user searches.
Q: What file types does Amazon Kendra support?
A: Kendra supports unstructured and semi-structured data in .html, MS Office (.doc, .ppt), PDF, and text formats.
Q: How does Amazon Kendra handle incremental data updates?
A: Kendra provides two methods of keeping your index up to date. First, connectors provide scheduling to automatically sync your data sources on a regular basis. Second, the Kendra API allows you to build your own connector to send data directly to Kendra from your data source via your existing ETL jobs or applications.
Q: What languages does Amazon Kendra support?
A: Kendra supports US English.
Q: What code changes do I need to make to use Amazon Kendra?
A: Ingesting content does not require coding when using the native connectors. You can also write your own custom connectors to integrate with other data sources, using the Kendra SDK. You can easily deploy search in your application with a few lines of code, or copying short code samples from reference implementations to replicate specific functionality. The SDK provides full control and flexibility of the end user experience.
Q: In what regions is Amazon Kendra available?
A: Kendra is in a public preview. It is currently available in the following AWS regions: Northern Virginia, Oregon, and Ireland. See the AWS Region Table for more details.
Q: How does Amazon Kendra improve its machine learning models?
A: Amazon Kendra's machine learning models are regularly retrained and tuned for each customer by analyzing and incorporating end-user search patterns and feedback. Improvements to a customer’s models are not shared with other customers.
Q: Does Amazon Kendra use my content to improve machine learning models for other customers?
A: Content processed by Amazon Kendra is not used to develop or improve the quality of Amazon Kendra for other customers or other Amazon machine-learning/artificial-intelligence technologies.