Improving Enterprise Search With Amazon Kendra
Over the past few years, open-source technology has radically gone mainstream: from streaming technology (Spark) to data storage and management technology (Hadoop) to languages for machine learning, data science and advanced analytics (Python, R).
Open-source technology can be seen as a “DIY assemble-your-own toolkit”. It means, enterprises need to build their own enterprise-grade platform from scratch using these jumble of open-source components — just like the way they may have piloted or experimented with other AI or ML projects.
Now, the tech giants may have always done it this way, but it’s not an easy task for the traditional enterprise organizations. According to Gartner’s annual survey of 2019, the number of traditional companies deploying machine learning in production has significantly dropped since 2018.
These companies need to put together an enterprise tool that can easily let them manage and scale their scattered data and models. But sadly, most of them got stuck at their pilot stages — not being able to get value-generating production implementations.
Nevertheless, help is on the way. Recently, Gartner’s VP Analyst Rita Sallam forecasted that the enterprises that may have, in the past, conducted experiments with open source technologies in their pilot efforts, are now likely to resort to the commercial AI & ML platforms to pull together their open-source enterprise deployment efforts.
In this time of need, at AWS re:Invent, the biggest cloud event of 2019 that took place in Las Vegas this December, the eCommerce giant introduced a number of AI & ML tools and services. One of which is the enterprise search engine called “Amazon Kendra” and it aims to take down both Microsoft and Google.
How does it work?
Amazon Kendra lets you find your files, documents, and content stored across your company in a much easier and quicker manner. It is achieved through an intuitive search that is powered by natural language queries and machine learning. As a result, it offers a higher degree of accuracy in search results.
Not only can it surface the most relevant links and documents, but it can also give you direct answers to your queries when it’s possible. To give you an example, let’s suppose you throw a question at the system: “how long is maternity leave?” Your question is met with a short and simple response such as “98 days”. Or if you ask “Where is the IT support desk?” Kendra could reply, “On the third floor” in addition to a bunch of relevant links. So, instead of using only simple keywords, it lets you ask easy natural language questions to get the information you are looking for. Thus, you no longer have to sift through the huge lists of search results to find what you need.
Andy Jassy, the CEO of AWS said that the technology will “totally change the value of the data” that enterprises have. Enterprises can get started with Amazon Kendra just by linking their storage accounts and answering some frequent employee questions. It helps Kendra to index those answers and information and understand its context, intent, and relationships using ML and NLP.
How does it benefit enterprises?
Data within enterprise organizations are usually siloed, unstructured, massive in amount and scattered across numerous data sources. Therefore, it’s hard to search and find a particular piece of information you’re looking for. On top of that, files, folders, and documents are usually found in different formats and jargon and are stored across different places such as Dropbox, SharePoint, etc.
This AI service by AWS is an intelligent enterprise search engine that aims to help organizations that don’t have machine learning expertise, to access unstructured data in different formats scattered across different data sources.
Kendra builds a search index from your enterprise documents using machine learning. Then it uses natural language understanding in order to learn the context and intent of the information in those documents and the relationships amongst them. You can test your queries, refine them further and give feedback to make the system better over time.
Through this ML-powered search tool, enterprises are getting better visibility and improved focus on the interactions happening between the organization and its clients minute by minute, which helps to make the most detailed inspections ever possible.
Enterprises are looking to leverage Amazon’s comprehensive AI & ML on a continual basis to address the common organizational challenges. These challenges include the identification of fraudulent transactions, internal search, helping developers write more optimized code as well as improving the overall quality of their enterprise system.