Welcome to our third and final post which we hope you’ve been looking forward to reading. In this article, we’ll highlight engaging points behind Amazon comprehend amongst other interesting AWS features. We hope you enjoy the read…
Comprehend is Natural Language Processing used for tagging large pieces of text with organisations, languages, people and other metadata. It can also provide sentiment analysis and the language used.
Rekognition for Video
If you’re familiar with the image processing version of this it’s very similar. Facial recognition, facial characteristic, and key features can all be detected. The real nifty part is it can detect people moving in the frames, and ever track them out of shot — like behind a shelf in a supermarket – and carry on tracking them when they come back into the frame.
This is a huge announcement and reduces the barrier to entry for AI quite dramatically. Not only can you use it to build models with the latest frameworks, it can also then deploy and serve prediction requests. It allows you to convert between the popular machine learning frameworks, and even use ML to detect which ML algo is the best. Inception.
Use cases are detecting industrial machine failure, fraud detection, detecting customer churn to send them offers. The list is endless.
Integrates tightly with Redshift, EMR, Glue and S3.
This service takes audio and, supposedly accurately, transcribes it into text. It also uses Machine Learning to smartly add punctuation.
I guess they got sick of using Google’s API. I wonder how well this stacks up. This was announced alongside Transcribe. It uses an entirely neural based model to translate, which is quite a bit different from how many other services work.
This is an Intel-based small form factor PC, with a GPU. You can easily create ML models in SageMaker, and ship them to the device where it can do local processing on live video. The metadata, or even the metadata and video itself, can be sent up to the cloud.
“I’m still confused as to what makes an ‘AWS’ product and what makes it an Amazon one…”
AWS WAF Managed Rules
This is a Web Application Firewall. They’ve partnered up with some big players in the industry to create lists of rules. These can be paid for on top of the service itself to protect against certain threats.
Amazon S3 + Glacier Select
Instead of pulling all the data down from S3 or Glacier so that you can filter it, you can do it in place. I assume this is what some of the previously announced services that let you run SQL against S3 buckets use. It also works with Glacier!
This is managed GraphQL. It’s not specific to apps, but it does lend itself to those, and SPAs.
This is an encrypted and anonymised channel for services to talk to each other across VPCs and even on-prem via Direct Connect.
We hope you enjoyed the article.