Ref: https://learn.cantrill.io/courses/1820301/lectures/42157338
DEMO: https://www.udemy.com/course/aws-ai-practitioner-certified/learn/lecture/44977263
Amazon Comprehend - Key Concepts
- 🔧 Natural Language Processing (NLP) service
- Finds insights and relationships in text
- Input = Document
- 💡 conceptually, think of it as text
- Output = offers many common NLP tasks:
- Extract entities (key phrases, places, people, brands, events…)
- e.g. New York, Michael Jackson, Adidas…
- Detect Language of the text
- Detect Personal Identifiable Information (PII)
- Understand sentiments (e.g. how positive or negative the text is)
- Analyze text syntactically using tokenization and parts of speech
- Automatically organize a collection of text files by topic
- Can use pre-trained models or custom models
- Can do
- Real-time analysis for small workloads
- Async jobs for larger workloads
- Example use cases:
- Analyze customer interactions (emails) to find what leads to a positive or negative customer experience
- Create and group articles by topics that Comprehend will uncover
Amazon Comprehend - Custom Models
Ref: https://www.udemy.com/course/aws-ai-practitioner-certified/learn/lecture/44886687 and https://www.udemy.com/course/aws-certified-machine-learning-engineer-associate-mla-c01/learn/lecture/45536581
- Useful for
- Document classification
- Custom entity recognition
- ‼️ Trained on your own data!
- Model versioning managed automatically by Comprehend
- May be copied between AWS accounts
- Attach IAM policy to model version, authorizing other account
- Other account can import model in its same region (requires ARN, region, and KMS key if encrypted)
Custom Document Classification