Skip to main content
This assignment is due before 11:59AM on Tuesday, November 22, 2022.

Semantic Parsing with Sequence to Sequence Models : Assignment 4

In this project, you will implement a system that translates natural language questions about geography to a formal language that can be used to query a database. Unlike classification, the goal is to generate an entire query string, given an entire language string. Each example will require many predictions. You will need to explore neural sequence to sequence approaches to this problem. To do well, you will need to implement an attention mechanism for your model, although it is not strictly required. After you have built your model, run it on the test set and submit your best results to the leaderboard. There is not autograder for this project, beyond an evalution of final output, but also please submit your code for review.

This project requires signficant code base support, and java to evalatue the queries, which we have provided for you. The colab notebook below clones our git repository, and has organized the code you need to produce. While it is not strictly required that you stay within the structure of this code (we are only evaluating your output) we highly recommend it. The models you need to write are runnable in minutes on either GPU and CPU.

Please see the full description of the project attached below.

Here are the materials that you should download for this assignment:

Deliverables

Here are the deliverables that you will need to submit:

  • Code, as always, in Python 3.
  • Predictions on the test set
  • PDF Report (called report.pdf)