Massive Dataset Mini Project

Importing Necessary Packages

The code begins by importing several Python libraries and packages:

• pandas for data manipulation and analysis.
• numpy for numerical operations.
• os for interacting with the operating system.
• json for working with JSON data.

It also imports specific modules from the googleapiclient and google.oauth2 packages for Google Drive integration.

import pandas as pd
import numpy as np
import os
import json

from googleapiclient.discovery import build
from google.oauth2 import service_account

Python3 Development Environment

The dataset used in the project can be downloaded here (https://amazon-massive-nlu-dataset.s3.amazonaws.com/amazon-massive-dataset-1.1.tar.gz).

It sets a variable pivoit_lang to the path of a JSONL file named 'en-US.jsonl' within a 'data' directory. It lists all the files in the 'data' directory and stores the list in the list_of_files variable. It reads data from 'sw-KE.jsonl' (Swahili data) and 'en-US.jsonl' (English data) JSONL files into Pandas DataFrames. It creates a new DataFrame en_sw by selecting columns from the English and Swahili DataFrames.

Working with Files

This part of the code processes data for three languages: English (en), Swahili (sw), and German (de), creating separate JSONL files for test, train, and dev sets for each language.

For each language:

• It reads data from the respective JSONL file into a Pandas DataFrame.
• It filters data based on the 'partition' column to create separate DataFrames for test, train, and dev sets.
• It writes these filtered DataFrames to separate JSON files in the 'generated_files' directory.

After processing these languages, the code generates one large JSON file (large_json_file) that contains translations from English to other languages for the train sets. Finally, it writes the large_json_file to a JSON file named 'large_json_file.json' in the 'generated_files' directory.

Uploading Files to Google

The code defines a dictionary access_cred containing Google Drive access credentials and creates a list api_url with the Google Drive API URL. It also sets the service_account_file and parent_folder_id variables.

Two functions are defined:

• auth(): This function authenticates the user using Google service account credentials.
• upload(file_path): This function uploads a file specified by file_path to Google Drive.

The code then uses these functions to upload files to Google Drive. It lists all files in the 'generated_files' directory and attempts to upload each file to Google Drive, except for 'de_train.json,' which has an issue causing it to fail to upload. It also prints an error message for any failed uploads.

Important notes

• utt: the raw utterance text without annotations.
• annot_utt: the text from utt with slot annotations formatted as [{label} : {entity}]
• partition: is either train, dev, or test