Collect-Audio-Data-and-Text

Video Demo:

Description:

About Me:

• My Name: Eric Cheng
• City and Country: Tainan, Taiwan
• Contact Info: ratherman5796@gmail.com
• Github Repo
• Lastest Updated Date: 2021/10/30

About this Project:

• Purpose:
  • The purpose is to help AI Teams to collect audio-and-text pairs of data. With tons of pairs of data collected, it's likely to teach a Text-To-Speech AI model to speak!
  • 

File Structure:

• Data/
  • Gen_Audio/ # This folder is used to store the generated recordings.
• Src/
  • static/
    • css/
      • main.css # Somehow beautify the user interfacec.
    • img/
      • tts.jpg # This picture will be showed once index page is loaded.
  • templates/
    • base.html # Base template of webpage.
    • index.html # Homepage of my project.
  • app.py # Core src code, mostly follow the framework of Flask
  • Insert Data To Db.ipynb # Use this notebook to (1) get posts , (2) split them into words and sentences, and (3) insert texts into DB.
• .gitignore
• README.md

Project Structure:

• Database: Microsoft SQL
  • For each row of data, it contains (1) text/sentence, (2) path of audio file, (3) duration.
  • (1) text/sentence
    • Source: https://www.theguardian.com/uk
    • After crawl the posts from the above mentioned link, split them into words and sentences and insert into DB.
  • (2) path of audio file
    • The path will be decided after we finished recording the word/sentence.
    • After the recording generates, the corresponding path will be inserted into DB.
  • (3) duration
    • "Duration" means duration of a certain recordings. It's usually no more than 15 sec.
    • This attribute will be decided once the recoding is generated.
• Framework: Flask
  • Use flask_sqlalchemy to TALK to DB.
  • After setting up the environment, use flask run to trigger this project.

Setup Setting up Env

• Step 1: Make sure you've installed anaconde.
• Step 2: conda create -n "CADAT" python=3.8
• Step 3: conda activate CADAT
• Step 4: pip install jupyter notebook
• Step 5: pip install beautifulsoup4
• Step 6: pip install requests
• Step 7: pip install numpy
• Step 8: pip install pandas
• Step 9: pip install pyodbc
• Step 10: pip install flask-sqlalchemy
• Step 11: pip install flask
• Step 12: pip install pydub
• Step 13: pip install librosa
• Step 14: pip install ffmpeg
• Step 15: Download ffmpeg.exe for your OS, and then append this .exe file your into your path.
• (Optional) Step 16: ffmpeg -version
• (Optional) Step 17: git clone https://github.com/Ratherman/Collect-Audio-Data-and-Text.git
• (Optional) Step 18: cd Collect-Audio-Data-and-Text/Src/
• (Optional) Step 19: flask run

About How to ... please check:

• How to Connect Flask to MSSQL?
• How to install MSSQL on Windows 10?

How to Connect Flask to MSSQL?

• Watch this YouTube Video for basic Flask implementation.
• Read this StackoverFlow Forum for connecting Flask to MSSQL.
• Step 1: Import packages

from flask import Flask
from flask_sqlalchemy import SQLAlchemy

• Step 2: Configure app and Create db instance

app = Flask(__name__)

params = urllib.parse.quote_plus('DRIVER={SQL Server};SERVER=HARRISONS-THINK;DATABASE=LendApp;Trusted_Connection=yes;')
app.config['SQLALCHEMY_DATABASE_URI'] = "mssql+pyodbc:///?odbc_connect=%s" % params
app.config["SQLALCHEMY_TRACK_MODIFICATIONS"] = False
db = SQLAlchemy(app)

• Step 3: Setup the table (Demo)

class audio_data(db.Model):
    id = db.Column(db.Integer, primary_key=True)
    content = db.Column(db.String(500), nullable=False)
    audio_path = db.Column(db.String(500), nullable=False)
    duration = db.Column(db.Integer, default=0)
    
    def __repr__(self):
        return "<Content %r>" % self.content

• Step 4: Create Database
• Go to command line (anaconda prompt) and do the following:
• python
• from app import db
• db.create_all()

How to install MSSQL on Windows 10?

• If needed, you're encouraged to follow this Tutorial.
• Step 1: Go to this website and download the express version, i.e., SQL Server 2019 Express.
• Step 2: Open the .exe file.
• Step 3: Choose the Basic option.
• Step 4: Click Accept and Choose the installed location.
• Step 5: Create a database named DB, which will be the one we used in this project.
• (Optional) Step 5: Click Install SSMS and a webpage would be popped up.
• (Optional) Step 6: Click the hypytertext "下載 SQL Server Management Studio (SSMS) 18.10".
• (Optional) Step 7: Open the .exe file.