AWS Kendra: Search internal company data with Amazon's cloud service

Amazon has launched a new search engine for companies in general under the name Kendra. It enables AWS users to index their structured and unstructured data. These can come from different back-end systems such as applications, relational databases, directly from the file system or the local company network.

Kendra then categorizes the data. For the public launch of the service, the search engine has mastered 14 areas: automotive, health, human resources, law, media and entertainment, news, telecommunications as well as travel and leisure now complement chemistry, energy, finance, insurance, IT and pharma. The latter six categories already existed during the preview phase.

Local and cloud applications as data sources

AWS also provides cloud connections to the OneDrive, Salesforce and ServiceNow applications, which previously existed for S3, Amazon's Relational Database Service (RDS) and SharePoint Online. The provider had first presented Kendra at its re: Invent 2019, since then the developers want according to their own information have also improved the speed and accuracy of the search engine.

The data captured by Kendra can be searched in the associated console. In addition to regular search parameters, it should also be able to understand questions in natural language. The service should be able to weight the indexed pages according to their relevance or provide direct answers to questions. In the future, the developers want to expand Kendra with an auto-completion and offer results tailored to users with ML functions (machine learning).

Amazon makes no statements as to whether and when Kendra can handle languages ​​other than English. Companies can also integrate the service into their own programs; an API and documentation are available for this.

To start with, Kendra can be sourced from the US East (N. Virginia), US West (Oregon) and Europe (Ireland) regions. With the Enterprise Edition, customers can search up to 500,000 documents up to 40,000 times a day for $ 7 an hour. Each capture of a document costs $ 0.000001, each synchronization with a data source $ 0.35 an hour. Additional capacities can be added for an additional charge. Amazon also offers a Developer Edition for testing purposes.

Competition from established providers

Kendra competes with existing services: For example, Google offers a company-specific variant of its search engine called Cloud Search. The service can already be used with over 100 data sources. For its service, which is also called Search, Microsoft particularly emphasizes the integration with other 365 programs, but it can also be used with local applications from the provider such as SharePoint. IBM, on the other hand, relies on Watson's language skills.

(fo)