Textract with Adapter- Correlating Query Results based on their location

Question

I am performing PDF analysis using AWS Textract. When using the table feature, I did not get sufficiently accurate responses, so I created an adapter. I started making requests using the adapter I created (using the aws-sdk/client-textract package), and I can see data close to accurate for the queries I created for training. However, I notice that in the API response, not all query results contain position information. I need the position in all the responses to my queries, as this is necessary to correlate these results based on their position. Can I have position information in all the query results, or do you have any alternative suggestions?
Thank you.

Answer

Thanks for using Textract service. Unfortunately there is no guarantee that QUERY result will always have position information. You can try search QUERY result text in all word BLOCK and see if can associate with word BLOCK, then use word BLOCK position information instead.

Textract with Adapter- Correlating Query Results based on their location

Relevant content