Amazon Comprehend is used to analyse text to reveal insights and relationships in unstructured data. The data can be any type of free form text such as emails or text messages. For sentiment analysis Amazon Comprehend can tell you the overall sentiment of the text identifying if it was favourable to the subject, or did it contain negative sentiments and how positive or negative the text was.
- AWS docs: https://aws.amazon.com/comprehend/
- AWS FAQs: https://aws.amazon.com/comprehend/faqs/
Video: What is Amazon Comprehend?
Key features of Amazon Comprehend
Out of the box Amazon Comprehend comes with six search capabilities to analyse documents:
- Entities – these can be dates, locations, people, events and other text phrases that identify something or event
- Key phrases – a group of words that identify something
- PII identification and redaction – data that could identify an individule such as name, date of birth, address
- Language – identify the dominant language
- Sentiment analysis – understands how positive or negative the text is
- Syntax – extract the part of speech for each word
Pre-trained or custom models can be used. The custom models have to be provided with labelled training data. There are three types of custom models:
- Custom document classification
- Custom entity detection
- Document topic modeling
Comprehend Medical is a Amazon Comprehend service pre-trained with medical terminology.
Video: Amazon Comprehend Video Snacks
Amazon Comprehend use cases
- Voice of customer analytics, sentiment analysis to find out what your customers think
- More accurate search, based on key words and phrases
- Knowledge management and discovery leading to recommendations of related articles
- Classify support tickets for better issue handling
- Perform Medical Cohort Analysis
Similar SageMaker built in algorithms
Topic modeling can be achieved with the LDA Sagemaker built in alogorithm.
- Photo by the blowup on Unsplash