AWS-Online-Tech-Talks
Amazon Kendra is an intelligent search tool, detailed in this video, that allows users to ask natural language queries and receive extracted answers from any indexed document or webpage. Kendra offers incremental deployment and easy setup using built-in connectors to popular data sources, making it an efficient enterprise searching application. Three primary use cases for Kendra are enterprise search, customer service, and embedded search. Amazon Kendra's analytics dashboard provides detailed search metrics, including top queries, top documents clicked, search click-through rate, and zero results, to help users understand search trends and areas for improvement. The video provides a demo on how to create an intelligent search application with Kendra and showcases document enrichment features, FAQs, and custom synonyms for better search results, a publisher for customization, and tips on running a quick Proof of Concept.
In this section, the speaker discusses the challenges faced by enterprises while searching for information on their websites and how Amazon Kendra can solve these challenges. With a focus on natural language queries and deep learning models, Kendra provides a unique search experience that feels more like interacting with an expert than clicking through long lists of documents. The service also offers incremental learning and easy setup using built-in connectors to popular data sources, as well as a WYSIWYG UI search application builder for easy customization. Security is also a top priority, with all customer data encrypted in transit and at rest. The goal of intelligent search, according to the speaker, is to provide end users with accurate answers quickly, so they can return to their tasks.
In this section, the benefits of using Amazon Kendra for intelligent search are discussed. Kendra allows users to ask natural language queries and returns extracted answers from any indexed document or webpage, eliminating the need for users to click through documents to find answers. The search results returned by Kendra include suggested answers, relevant documents, FAQ matches, and explicit user feedback options. Kendra's search relevance can also be influenced by incremental learning, custom synonyms, and curated input. Three main use cases for Kendra are enterprise search, customer service, and embedded search. Customer success stories from companies like 3M and Gilead Sciences illustrate the efficiency gains and improved productivity that can be achieved through the use of Kendra.
In this section, the video discusses a case study with Magellan RX Management, a full-service pharmacy benefit manager that needed to manage large volumes of complex and changing healthcare information for their customers and members. They implemented Amazon Kendra to build a secure and scalable agent assist solution that resulted in an average reduction in call times of about 9 to 15 seconds, saving over 4,000 hours on over 2.2 million calls per calendar year. The value proposition for intelligent search includes increased employee productivity and customer satisfaction, time and cost savings, and risk mitigation. Amazon Kendra provides a machine learning-powered search service that is fully managed, requires no machine learning expertise, and easily connects to popular content repositories. The video also covers a simple three-step process for getting Kendra up and running with your content, including data ingestion, search relevance tuning, and deploying Kendra-powered search to your application using APIs, code snippets, or code-free experience builder. Kendra's experience builder allows for fully functional customizable canvas search applications to be deployed in just a few clicks, without any coding or machine learning expertise required.
In this section, we learn about the detailed search metrics that Amazon Kendra provides via its analytics dashboard, which includes top queries, top documents clicked, search click-through rate, and zero results. The purpose of these reports is to provide customers with actionable information to help them understand search trends and areas of improvement. The reports can be viewed in the console or via an API. Following this, we are given a demo on how to create an intelligent search application with Kendra by adding two different sources: a subset of documents from AWS Machine Learning Blog website and a pre-configured data source available on the Kendra console. The demo shows the simple steps of creating an empty index, adding data sources to it, and synchronizing the content to make it searchable.
In this section, we learn how to add an additional data source to an existing index and test a new search on it. After adding a pre-configured data source (sample AWS documentation), we see the same answer returning from the blogs along with results coming from a new internal documentation website. We can filter these results to a particular data source based on author or categories via facets. We can use the tuning panel to increase relevance based on specific criteria like document freshness, popularity or to boost search results from one data source over another. Additionally, we can use Custom Document Enrichment features to tag documents flowing into Kendra based on their data source and test out enrichment rules to pre-process these documents and add metadata that can be used for filtering or boosting.
In this section, the speaker demonstrates how to create rules for document enrichment in Amazon Kendra by extracting entities and assigning tags to documents from different sources. The rules are created by setting conditions based on the data source, such as specific URLs, and assigning tags based on that condition, by storing the metadata in the category field. Once the rules are set, the two data sources are synchronized, and the metadata is enabled as a facet filter. The speaker also shows how FAQs can be added to the index, and the curated answers are displayed for each relevant search query along with the match documents. Additionally, Amazon Kendra provides the option to import custom synonyms for better search results.
In this section of the video, the experience builder for deploying a Kendra search application to end users is demonstrated. The user can create a new experience, choose data sources, create a role, and configure AWS SSO for securing the application. The editor in builder mode allows the user to customize different components of the search page. Customization features, such as font size, background text, and order of fields, can be modified. Changes can be tested in live view mode without affecting the production application until it is published. Additional users and groups can be added to the application who can access it via a secure link. Overall, creating a standalone search application powered by Amazon Kendra is simple and efficient.
In this section, the speaker provides tips on how to run a flash Proof of Concept (POC) on your own data in under 30 minutes to quickly see the power of Kendra's intelligent search capabilities. First, one needs to pick a use case that aligns with their needs and make sure the content is easily accessible. Then, by using the website connector or S3 connector, one can add content into Kendra. Next, write down a few questions that you'd like to ask Kendra about this content. By typing in these questions in the AWS Kendra search console, one can get a rough idea of the search results and compare them to their existing search solution to get an idea of the improvements Kendra can bring to their search experience.
No videos found.
No related videos found.
No music found.