So, what does a POC look like? I think something like the following:
1. A selection of end user queries we consider candidates
See #1
2. An understand of the "final" response to the query
What does a response look like (after dialog for disambiguation)? Probably something like the following, with some things being optional, dependent on the query and the data to answer it:
a) A summary response in narrative form
b) A visualization of data (or a small sample of it) related to the response
c) province for any key data points (where do they come from)
d) A list of source datasets (and external sources) for the response
3. The logic for clarifying and disambiguating each query
This is the chat interface, OpenAI API configuration, and prompt engineering part.
4. One, or a handful, of CKAN instance to work against
5. A server for data storage and processing
CKAN instances we can query and metadata about them
So, what does a POC look like? I think something like the following:
1. A selection of end user queries we consider candidates
See #1
2. An understand of the "final" response to the query
What does a response look like (after dialog for disambiguation)? Probably something like the following, with some things being optional, dependent on the query and the data to answer it:
a) A summary response in narrative form b) A visualization of data (or a small sample of it) related to the response c) province for any key data points (where do they come from) d) A list of source datasets (and external sources) for the response
3. The logic for clarifying and disambiguating each query
This is the chat interface, OpenAI API configuration, and prompt engineering part.
4. One, or a handful, of CKAN instance to work against
5. A server for data storage and processing
6. A website with a search > chat interface