I ran all my code in the given ipynb file on Google Colab. To run the code, please upload the given data file in which I renamed to "src.csv". Write the natural language description of the desired query into variable "input_query". Run all the code blocks in order, the result query would be generated at the end.

With more time, I would improve the model with few shot learning examples and try the second approach of using langChain.

I experimented with different models of GPT, 4 is the best.

My work flow is included in the PDF file.

Some generayed samples are listed below:

Find total contributions by each contributor:

SELECT contrib, SUM(CAST(REPLACE(REPLACE(amount, '$', ''), ',', '') AS INT)) as total_contrib FROM contributions
GROUP BY contrib;

List all contributions made in the year 2022:

SELECT * FROM contributions
WHERE strftime('%Y', date) = '2022';

Find the top 5 recipients who received the highest total contributions:

SELECT recipient, SUM(CAST(REPLACE(REPLACE(amount, '$', ''), ',', '') AS INT)) as total_contrib FROM contributions
GROUP BY recipient
ORDER BY total_contrib DESC
LIMIT 5;

Count the number of contributions made by contributors from the state of California (CA):

SELECT COUNT(*) FROM contributions
WHERE state = 'CA';

Find the average contribution amount:

SELECT AVG(CAST(REPLACE(REPLACE(amount, '$', ''), ',', '') AS INT)) as Avg_contrib FROM contributions;

Maximum donation done by SBF:

SELECT MAX(CAST(REPLACE(REPLACE(amount, '$', ''), ',', '') AS INT)) AS Max_donation FROM contributions
WHERE contribid = 'SBF';