Italian Language Transcript Generator from YouTube Video URLs (Generatore automatico di trascrizioni in lingua italiana dagli URL dei video di YouTube)
This project streamlines the process of downloading and saving Italian language transcripts, along with their English translations, using the YouTubeTranscriptApi in .docx
format. No major conceptual contributions were made in the writing of this code (as of the June 19, 2021 version @ 6:07 PM CDT). That is, this code implements the API, cleans up the text data and saves it Word documents. The idea behind the project is to create a tool for language learners.
To install all dependencies, simply navigate to the local directory in which this project is saved and run
pip install -r requirements.txt
Simply run python3 gen_transcript_from_URL.py
. The terminal will prompt you for a YouTube URL. Either copy and paste or type in the URL by hand and press enter. If a transcript exists, the program will download both the Italian original language and English translation texts and save them nicely into Word documents, whether or not the text is structured/punctuated, inside folders named with the corresponding video title.
By unstructured/unpunctuated text it is meant that the transcript comes as a single string with no capitalization, commas or periods. Thus, it would be desirable to use a trained NLP model to take unstructured text as input and reconstruct sentence boundaries. This is a very non-trivial instance of sentence tokenization. An obvious approach would be to find a large corpus of punctuated Italian sentences and remove the punctuation to create a training set. Another improvement would to use language-translation NLP to construct a combined Italian+English transcript for unstructured instances.