Amazon Kendra: Knowledge-Based Management System (KMS)

A Basic Summary

Introduction

Amazon Kendra is a revolutionary new service provided by AWS that allows for enterprises and individual developers alike to have access to building powerful AI search engines based on Machine Reading Comprehension (MRC), Paraphrasing (FAQ), and Keyword search algorithms.

Through this service, users can provide accurate and convenient answers to their questions and avoid scrolling and wasting time searching through various documents and pages to find their answers, as seen below.

At the top of this image, you can see that the search engine was able to recognize natural language inquires, i.e. a question phrased how a person would naturally like to ask it. This is possible through Amazon Kendra’s advanced NLU engine and its Information Retrieval (IR) search. Kendra’s MRC engine allows for quick and efficient filtering of unstructured data to find the exact right answer and save the user’s precious time. Kendra’s Paraphrasing engine allows for the FAQs listed below the boxed answer, as the engine finds similar questions that the user may need. User feedback is also considered, as shown with the thumbs up and thumbs down icons in the image above, to adjust the search results.

Why is this service necessary?

According to Mckinsey, employees spend around 20% of their time at work searching for information and are only successful in finding the required information around 44% of the time according to IDC. This creates a massive loss in efficiency as well as time, and in a large company enterprise, this wasted time can equate to an enormous loss of money and resources, with a company of 20k employees wasting around $114M per year to search time alone. Without proper search engines, information is easily lost or duplicated unnecessarily, overall affecting the efficiency of the company as a whole. Of course, we all can understand the frustration of tirelessly searching for an answer with no fruition, so a poor search engine also brings down customer satisfaction.

Most search engines within company enterprises at the moment focus on keyword-based searches, which are both inaccurate, unable to properly search through unstructured data, and require an incredible amount of manual upkeep.

Amazon Kendra’s Features

Amazon Kendra is able to combine several incredible features, as seen in the image below:

Through this service, the search engine provides better answers to users through advanced AI reading comprehension, FAQ matching, and even document ranking to find the best answer. The search engine’s ability to interpret both natural language queries and keyword queries allow for the user to ask questions freely and consistently get accurate and helpful answers.

Amazon Kendra is optimized for 14 major domains: IT, Finance, Insurance, Pharmaceutical, Industrial, Energy, Legal, Media & Entertainment, Travel & Hospitality, Health, Human Resources, News, Telecommunications, and Automotive. With these diverse background domains, most queries can be met with a helpful answer.

A newer feature of Amazon Kendra is its ability to learn incrementally, which means it automatically improve its results over time through click-through statistics and user feedback.

Users are also able to fine-tune their results based on their relevance. Users can prioritize certain data based on the level of authority/confidence they have in particular data sources or authors, the time the data was created, as well as popularity among users, such as upvotes or views on particular websites.

Some other exciting features offered by Amazon Kendra are depicted in the image below:

Amazon Kendra allows for extensive user manipulation, such as the In-Console Search function that allows users to test their queries and tune the relevancy in the results based on their own preferences, all with guaranteed data security in transit and at rest.

Code samples are also provided to the users in order to replicate Kendra search components in their own applications.

The step to building your very own search engine using Amazon Kendra is simple and can be done in the following simple steps.

Use Cases

Some of the main use cases for Amazon Kendra include:

• Internal Search

Supports business functions, such as general operations, customer support, and Research and Development

• External Search

Helps customers find the exact information they need efficiently

• Content Management (CRM)/eDiscovery

ISVs can build more intelligent and data-driven applications

• FAQ Chatbots

Handles FAQs without any specific bot training or programming required. Support ad hoc queries and automatically extracts answers from unstructured data (a common issue among competitor search engines)

Some Enterprises that currently utilize Amazon Kendra include:

Insight & Summary

With the growing relevancy and necessity of a more efficient search engine, services like Amazon Kendra will be more and more essential to small and large enterprises alike. With improvements in NLU and paraphrasing technology, autonomy will increase exponentially and employees’ time will be used much more efficiently. In the search engine industry, domain is increasingly important, and the fact that Amazon Kendra supports such a large variety of domains means that its services are highly adaptable and capable in various companies with diverse backgrounds and focuses. With a self-improving search engine and user-specific UI/UX filters, the sky is truly the limit on the future of information search efficiency in the workplace.

Reference: ‘Amazon Kendra: Transform the Way You Search and Interact with Enterprise Data Using AI’