Skip to content

Latest commit

 

History

History
62 lines (55 loc) · 3.21 KB

README.md

File metadata and controls

62 lines (55 loc) · 3.21 KB
Raw

Welcome - Project Overview

  • This is our final project for Interactive Dialogue Systems
  • We created an Question/Answer system using an encoder-decoder model to answer questions about past Notre Dame football games
  • The model is heavily based from this paper: https://arxiv.org/abs/2101.03999

General

  • For the best experience, clone the repository and open it in an IDE like VSCode that supports .png's
  • Each subdirectory has a README.txt which details what is contained in it
  • The root directory contains the files to:
    • Collect the college football data (dataCollection.py)
    • Create the context, answer, and question files (txtCreation.py)
    • Create the model (model.py)
    • Train the model (train.py)
    • Make a single prediction (live_predictor.py)
    • Test the models (test.py)
  • The heatmap directory provides insight into where the model is looking to predict the next word
    • Please look there for analysis on the heatmap images when answering 3 questions
    • The images were generated for the simple model and the sentence context model

TO USE:

  • python3 live_predictor.py -n 1 -q "who was the coach ?"
  • python3 test.py

Setup

Virtual Environment Setup (or use student machine):

  • Upgrade pip: python3 -m pip install --user --upgrade pip
  • Install virtualenv: python3 -m pip install --user virtualenv
  • Create the venv: python3 -m venv env
  • Activate the venv: source env/bin/activate
  • Install requirements: pip install -r requirements.txt
  • Exit: deactivate

This is what the output of the teacher force function is (c=context, q=question, n=next work a=answer):

c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['bob'] a: <s>
c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['davie'] a: <s> bob
c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['was'] a: <s> bob davie
c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['the'] a: <s> bob davie was
c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['coach'] a: <s> bob davie was the
c: <s> the game occurred in 2001 notre dame won the game bob davie was the coach </s> q: <s> who was the coach </s> n: ['</s>'] a: <s> bob davie was the coach

To get this add this code:

for c, a, q, n in zip(train_context, train_answer, train_question, next_word):  
    print("c: " + c + " q: " + q + " n: " + n + " a: " + a)  
quit()  

To see vectors going into the model:  
for i in range(50):  
    print("c: " + str(train_context[i]))  
    print(" q: " + str(train_question[i]))  
    print(" n: " + str(train_next_word[i]))  
    print(" a: " + str(train_answer[i]))  
    print()  
quit()

Useful links

https://github.com/aakashba/paqs2021/blob/paper/paqs_dev/models/ast_attendgru_xtra.py https://github.com/aakashba/paqs2021/blob/paper/model.md https://github.com/mcmillco/funcom/blob/master/models/attendgru.py