Amazon Comprehend uses natural language processing to extract insights from text documents.
Text analysis
Amazon Comprehend reads text and tells you what’s in it: the people, places and organizations mentioned, key phrases, the language and the sentiment. It needs no machine learning expertise.
How it works
- Send text to the API, or start a job over documents in Amazon S3.
- Comprehend returns entities, key phrases, language and sentiment.
- For clinical text, use Amazon Comprehend Medical to detect medical information such as conditions and medications.
Exam cues
- If the question says sentiment, entities or key phrases in text, think Amazon Comprehend.
- If the question says extract medical information from physician notes, think Amazon Comprehend Medical, because it analyzes clinical text.
Remember
- Real-time for small workloads, asynchronous jobs for large sets.
Trap: Comprehend analyzes text; it doesn’t transcribe speech (Amazon Transcribe) or extract text from images (Amazon Textract).
Da documentação da AWS (5)
- Amazon Comprehend uses natural language processing (NLP) to extract insights about documents by recognizing entities, key phrases, language, sentiment and other common elements.
- Amazon Comprehend can search social media feeds for product mentions or scan a document repository for key phrases.
- Amazon Comprehend runs real-time analysis for small workloads or asynchronous jobs for large document sets, with pre-trained models or custom models for classification and entity recognition.
- Amazon Comprehend Medical uses NLP to detect medical information such as conditions, medications and test results in unstructured clinical text like physician notes and discharge summaries.
- Amazon Comprehend detects PII entities in English or Spanish text and can locate them or redact them.
Custom analysis and targeted sentiment
Amazon Comprehend can find entity types and classify documents using categories you define. It can also report sentiment for individual entities, such as products or brands.
How it works
- For business-specific terms or product codes, use custom entity recognition to train a model for your entity types.
- For your own document categories, train a custom classifier, then classify a single document in real time or a set asynchronously.
- For sentiment about each product or brand in a document, use targeted sentiment.
Exam cues
- If the question says extract ingredient names outside the preset entity types, think custom entity recognition, because it finds types you define.
- If the question says sort customer emails into support categories, think custom classification, because it uses your categories.
- If the question says find separate sentiment for a product and a brand, think targeted sentiment, because it reports sentiment for each entity.
Remember
- Custom classification accepts plain text, PDF, Word and images.
- Sentiment analysis gives one dominant sentiment for a document; targeted sentiment gives sentiment for each specific entity.
Trap: A document’s dominant sentiment does not tell you the sentiment for each product or brand. Use targeted sentiment for that detail.
Da documentação da AWS (3)
- Amazon Comprehend custom entity recognition finds entity types of your own, such as product codes or business-specific terms, beyond the preset generic entity types, with automatic annotation and model training.
- Comprehend custom classification trains a classifier on categories you define, such as support request types, then classifies a single document in real time or a document set asynchronously; it accepts plain text, PDF, Word and images.
- Comprehend targeted sentiment gives the sentiment for each specific entity in a document, such as a product or brand, while sentiment analysis gives one dominant sentiment for the whole document.
Flashcards
Nada para revisar aqui agora. Escolha outros tópicos ou outro filtro.
·
Todos os 8 flashcards em lista
A company wants to know whether customer reviews are positive or negative. Which AWS service fits?
Amazon Comprehend, which detects sentiment in text with natural language processing.
Which AWS service extracts entities and key phrases from a large set of text documents?
Amazon Comprehend.
A marketing team wants to find mentions of its products in social media posts. Which AWS service helps?
Amazon Comprehend.
Does Amazon Comprehend read text out of scanned images of documents?
No. Comprehend analyzes text it is given; extracting text from scanned documents is Amazon Textract.
Which AWS service extracts medical entities from physician notes, discharge summaries and test results?
Amazon Comprehend Medical, the clinical-text version of Comprehend.
A food company must extract its own ingredient names from product descriptions, which aren't among Amazon Comprehend's built-in entity types. Which feature fits?
Amazon Comprehend custom entity recognition, trained on the company's own entity types.
A company wants to route incoming customer emails into its own categories, such as billing, support and complaints. Which Comprehend feature fits?
Custom classification: train a classifier on labeled examples of each category, then classify new emails in real time or in batch.
A review says "the tacos were delicious but the service was slow". Which Comprehend feature shows a separate sentiment for the food and for the service?
Targeted sentiment, which scores each entity. Plain sentiment analysis returns one dominant sentiment for the whole review.