Automatic-Assessment-of-Speaking-Skills

This Repository contains the implementation of Automatic Assessment of Speaking Skills Using Aural and Textual Information. The details of the models implemented in this repo can be found in this paper: https://aclanthology.org/2021.icnlsp-1.19/

Requirements

• Python3
• PyTorch
• scikit-learn
• numpy
• pyAudioAnalysis
• SpeechRecognition
• librosa
• fasttext
• tqdm

Run pip install -r requirements.txt to install all the required packages.

Folder Structure

Automatic-Assessment-of-Speaking-Skills
├───AudioAnalysis
│   ├───arousal.py
│   ├───emotion.py
│   ├───valence.py
│   ├───TrainANN.py
│   ├───TrainSVM.py
│   └───Saved models
├───HighLevelAnalysis
│   ├───highLevel.py
│   ├───TrainANN.py
│   ├───TrainSVM.py
│   └───Saved models
├───Test
│   ├───Test.py
│   ├───sample.wav
│   └───Models saved
├───TextAnalysis
│   ├───arousal_BiLSTM.py
│   ├───arousal_FastText.py
│   ├───emotion_BiLSTM.py
│   ├───emotion_FastText.py
│   ├───valence_BiLSTM.py
│   ├───valence_FastText.py
│   ├───TrainANN.py
│   ├───TrainSVM.py
│   └───Saved models
├───README.md
├───requirements.txt
└───Presentation.pdf

Datasets

The Datasets used in this paper are:

1. IEMOCAP
2. Savee
3. Emovo
4. Emo-db
5. Ravdess
6. PuSQ

Usage

To test the final model, run the following commands:

cd Test
python3 test.py

Enter the name of the audio file which needs to be tested

Enter Audio File Name: sample.txt

The output will two boolean values based on Enjoyment and Expressiveness.