Training Textract to Separate Text Blocks Into Separate Components with a Delimiter
0
I'm trying to use Textract to extract the product descriptions form our PDF catalogs in page order. The Textract analysis picks up the descriptions as text blocks, but how do I go about training Textract to split each product description text block into its key components, such as title, author, description, etc?
1 回答
- 最新
- 投票最多
- 评论最多
1
Hello,
To extract the key components like title, author, and description from product descriptions in your PDF catalogs, Textract currently does not have built-in capabilities for that level of customization.
Machine learning models trained on sample catalog pages could help automatically classify the text into different fields. Services like Amazon SageMaker, AWS Glue, etc can help build such models.
已回答 1 个月前
You can develop a post-processing system that applies rules to classify text blocks based on layout patterns, or for a more sophisticated solution, train a custom machine learning model with Amazon SageMaker to recognize and categorize the text appropriately.