Seek, and ye shall find …by language: Amazon Kendra meets Alexa for Business (Part 3)

Amazon Web Services recently introduced a new machine learning based search service called Amazon Kendra. It is built on natural voice control and thus offers excellent possibilities to be linked to Alexa for Business. The first two of this three-part article have introduced each of the the services that already offer some added value in a business context. Now Ben Freiberg explains how this can be further enhanced by cleverly combining both of them.

Part 3: „Alexa, help Kendra!”

There are two ways to use Kendra via Alexa: If already available, the private skill of the company can be extended by a new Intent. For example, “Alexa, ask Unbelievable Machine how much time I have to take my remaining vacation days.” As an alternative, a new Private Skill can be developed.

Since it is possible and desirable for the user to ask arbitrary questions, it seems difficult at first to find the right examples when creating the interaction model. However, Alexa already offers a suitable slot type, which is meant for exactly this case: AMAZON.SearchQueryThis allows any user requests to be covered. The only restriction is that with this slot type there may be a maximum of one Intent per skill, which is why the creation of a second skill is recommended in this case.

Alexa and Kendra are good individually…

The beauty of the combination of Alexa and Kendra is the complete elimination of friction for the user and the usage. In order to get information, you don’t have to open a slow loading page or think about which of the different tools might contain the answer. Not even the computer itself needs to be turned on. It is sufficient to simply ask a question in natural language.

Alexa takes care of the highly demanding part of text recognition and translation, while Kendra takes care of understanding and finding the information. The only thing the skill has to do is to pass on the question asked. Beyond that, of course, other features can be integrated, for example the direct collection of user feedback: “Has this answer been helpful?”

If echo devices with a display are included, such as an Echo Show, the skill can provide a multimodal experience. This means that in this case, besides the voice output of the best hit, other possible hits can be displayed in a list or suggestions for related search terms can be shown.

…and together even greater than the sum of their parts

Alexa and Kendra are individually very helpful – also for business purposes – and combine their respective strengths to the fullest. Cleverly linked together, the result is, for example, a company’s own powerful search engine for distributed knowledge and internal documents, controlled by the best language-based interface currently available and a service for the user that is as simple as it is smooth and adds value.

If it comes to the question how companies and users can achieve this best, do not ask Alexa – just ask us. We are here to answer your questions.

The basics in part 1: „Alexa, let’s talk business!
and part 2: „Kendra, fill me in!

Ben Freiberg is a Senior Data Engineer and proven AWS expert at the *um location and AWS Competence Center in Frankfurt.

This post is also available in: German