# QUESTION:
How to generate structured datasets ?
Lots of AI developers will encounter this issue.
# CONTEXT:
A chatbot can easily be created using Vertex AI Search and Conversation / DailogFlow CX agent. The agent in turn relies on having a good dataset. Uploading unstructured data to create a dataset is very problematic.
# OBJECTIVE:
To have a well-structured dataset the agent finds easy to get the correct responses from.
# ISSUE:
This Vertex AI Search and Conversation RAG embedding and indexing process works well on the Google sample data with respect to pixel phones, but badly or not at all when other unstructured data is used. The RAG process initially turns all the unstructured data into a soup which often turns the raw data into incomprehensible junk.
# NEEDED:
A low-code streamlined method of using Generative AI to convert unstructured data (e.g. PDF or HTML files) into well-structured files (e.g. BigQuery). A sample code that can be run in Colab Enterprise would be ideal ?
LinkedIn
Twitter
