Amazon Kendra – AWS Service that Focuses on Curated Enterprise Search.
1 ) Create an Index
2 ) Add your data source (like Word document)
3 ) Test out your queries and deploy search within your application
4 ) Upload an FAQ to give context to common search queries and commonly asked questions
Kendra provides more meaningful results based on document context. It uses Machine Learning, specifically a deep-learning based semantic search model, to return the best results in common speech.
Sure, happy to make this more understandable. That’s what Kendra would do! First, I want to make sure I’m clear about what kind of data Kendra analyses to provide results based on ‘natural language queries’. Kendra connects to a data source like S3, Salesforce, an RDS database, etc. It also supports semi-structured data like HTML, Word documents, PowerPoints, PDFs, and other text-format based documents. Through leveraging deep learning models to better understand natural language requests, Kendra helps customers find information out like ‘when the deadline is for electing healthcare benefits’ when starting your new job. Or you could ask questions like, ‘which holidays does my company observe?’ or ‘does my company have a 401k matching plan?’
And as of today, Amazon Kendra also provides confidence scores, which show customers the estimated accuracy of a search result. These confidence scores are labeled as one of four categories: Very High, High, Medium, and Low.
Some exciting features coming in the near future include:
- Query Auto Completion: used for auto-completing a user’s question
- Analytics and Ongoing Improvement: helpful for tracking user interaction, screen behavior during searches, and quality common metrics
Resources mentioned in this article:
