The wait is over. Five months ago Amazon has announced the development of a new service, Amazon Kendra. Today the product has officially launched.
What is Amazon Kendra?
Amazon Kendra is an enterprise search service that´s built on Artificial intelligence and Machine Learning. The output of the queries is in powerful natural language and the service can be embedded into your website or application. The service is useful for internal users such as employees at the time of searching company data, or end users such as website users looking for specific data.
The process is fairly simple, you start by connecting your stored data to the application. Stored data may be in any repository such as S3 storage or a Salesforce Rack. Then the user inputs a query in natural language, which Kendra understands, and the service delivers information as an output also in natural language. One might think what is this useful for if someone can manually look for something stored. Well, the answer is simple, the automation of the queries through AI is much more efficient than a human hand looking for something. It is faster and more accurate.
Forget about browsing through files or links that may contain the information you were looking for. It´s obvious that through thousands of files of a company nobody really knows where an exact value of data is stored.
Its use is not limited to one company department. It is currently being used to browse HR, Operations, Support, IT, Financial Services, Insurance, Pharmaceutical, Legal, Media, and Entertainment.
The product´s feature “Autocompletion” reduces the user´s typing by 25%. You start typing and the service completes your queries based on multiple factors. It is based on suggestions and can finish your phrases by the most common types of searches based on the beginning of your phrase.
As being part of the AWS (Amazon Web Services) branch of Amazon, they have decided to price it in a way accessible and reliable for the user. You pay for what you use. You are billed for the time from creating and index to the time of deletion.
The product has two versions. Enterprise Edition, which is targeted for production workloads. And Developer Edition which has been built as a low-cost opportunity for developers who want to try to build a minimum viable product based on the service. 750 hours for 30 days are offered for free, after that the prices vary depending on the selected plan.
The developer edition offers: up to 10,000 documents, up to 4000 queries per day, the user can source data from up to 5 sources, billed per hour $2,50, and billed per month $1800.
The Enterprise edition offers: up to 500,000 documents, up to 40,000 queries per day, the user can source data from up to 50 sources, billed per hour $7, and billed per month $5040.
The service is composed of the following components. Index, Source Repository, Data Source, and document addition API. The Index is created from the selected source documents. The Source Repository contains the documents from Index. The Data Source, Syncs the data from the sources to the index. So that when the sources are updated the index is updated too. Lastly, the document addition API serves to add documents through an API extension instead from the local index.
What kind of questions kind you ask Kendra? You can ask Factoid Questions, Descriptive questions, and Keyword Searches. Factoid Questions are based on 4 commands, “who”, “what”, “when”, “where”. Descriptive Questions come with longer answers, from a short sentence to a paragraph, to an entire document. Keyword searches are based on using a word that will return relevant information related to the word. The question is not specific, but the word returns enough information to make a query.
Talking about the market of the “Search as a service” Tools. We can include Google for web searches for example, but narrowing it down to enterprise searches, we can include a company that does a similar object. The company is called Elastic Search. Started as a startup and scaled so neatly that today it has more than 1000 employees, more than $100M raised in funding, and are actively looking to expand its team and technologies. The company is 8 years old and has 46K GitHub stars. Differently to Amazon Kendra, this company has its technologies available as Opensource, allowing developers to extract code and collaborate.
In Conclusion, a new technology has emerged and allows end users to search through extensive file data. The service is offered in two plans. A low-cost version, and an enterprise version. The service is cost-effective and intelligent to the point of understanding English fully and responding to queries as natural language. It works for multiple fields and we can expect it to be a growing tool for big companies and startups.