concept location in category aws

appears as: locations, location
AI as a Service: Serverless machine learning with AWS

This is an excerpt from Manning's book AI as a Service: Serverless machine learning with AWS.

Table B.1. Data requirements for AI services
Application Service Data required Training support

Machine Translation

AWS Translate

Text in the source language

Translate does not support or require custom training. You can however define custom terminology that is specific to your domain.

Document Analysis

AWS Textract

High quality images of documents

No training is required

Key Phrases

AWS Comprehend

Text

No training is required.

Sentiment Analysis

Text

No training is required.

Topic Modelling

Text

No training is required.

Document Classification

Text with classification labels

Training is required. We covered custom classifiers in Chapter 6.

Entity Extraction

Text. For custom entity training, labelled entities are required.

Standard entities (names, dates and locations) can be extracted without training. It is also possible to train AWS Comprehend with custom entities by providing a set of text with entity labels.

Chatbots

AWS Lex

Text utterances

Training is not required. AWS Lex builds models based on sample utterances and configured slots. This was covered in Chapter 4.

Speech-To-Text

AWS Transcribe

Audio files or streaming audio

No training is required but it is possible to add custom vocabularies and pronunciations to refine results.

Text-To-Speech

AWS Polly

Text, optionally annotated using SSML

No training is required. AWS Polly was covered in Chapter 4.

Object, scene, and activity detection

AWS Rekognition

Image or video

No training required.

Facial recognition

Image or video

No training is required but custom faces can be added.

Facial analysis

Image or video

No training is required.

Text in images

Image

No training is required.

Time Series Forecasting

AWS Forecast

Time series data and item metadata

Training is required. AWS Forecast trains models based on historical data and metadata you provide.

Real-time personalization and recommendation

AWS Personalize

Item catalog and user data

Training is required. AWS Personalize can train a model and choose an optimal algorithm based on the data provided.

For our conference event information extraction scenario, we are aiming to extract the names of people, locations, and dates from website data. This will then be stored and made accessible to users.