Skip to content

Latest commit

 

History

History
60 lines (47 loc) · 2.52 KB

README.md

File metadata and controls

60 lines (47 loc) · 2.52 KB
Raw

CIS668_Final_Project

Sentimental analysis for YouTube comments

Project Proposal (DDL Nov.1st)

Project proposal

  • Find suitable pretrained NLP model that could be applied directly without training (Bert?)
  • Youtube comment download feature (APIs)
  • Model training and fine-tuning

Weekly milestones (starting from Nov.1st)

Youtube public data APIs (Week 11, Nov.2nd to Nov.8th)

  • Implement download comments from specific Youtube videos
  • Preprocessing and data cleaning (nulls check, emoji etc.)

I've finished developing download comments from Youtube. Please see downloader.py for details. Below is the working principle of this youtube comment downloader.py. Architecture

Sample output Architecture

Usage

python downloader.py --output <output-path> --q <query-word> --maxresults <max-num-of-videos> Example: python downloader.py --output res.json --q amazonglacier --maxresults 2

Installation

TL;DR: just simply install requirements.txt pip install requirements.txt

Prerequisites:

  • Python Version >= 3.7
  • pip
  • The Google APIs Client Library for Python:
    pip install --upgrade google-api-python-client
  • The google-auth, google-auth-oauthlib, and google-auth-httplib2 for user authorization
    pip install --upgrade google-auth google-auth-oauthlib google-auth-httplib2
  • Requests
    pip install requests
  • lxml
    pip install lxml
  • cssselect
    pip install cssselect

Resources: Google official Python API client repository:

https://github.com/googleapis/google-api-python-client

YouTube Python API:

https://github.com/youtube/api-samples/tree/master/python

Bert Model Training and fine-tuning (Week 12-13, Nov.9th - Nov.24th)

  • Feed the data into the model, compare the output with groundtruth
  • Fine-tune parameters (Are we doing binary classification here? i.e. 1 for positive emotion and -1 for negative?)
  • Results explanation and investigation

Project report composition and slides (Along with previous milestones)

  • Write the complete report
  • Design slides