Python program which records audio, converts it into text and perform simple sentiment analysis on it
I have taken much help from this article for the sentiment analyser and it would be a crime to not mention it OwO: https://towardsdatascience.com/a-beginners-guide-to-sentiment-analysis-in-python-95e354ea84f6
First of all, you have to download https://www.kaggle.com/datasets/snap/amazon-fine-food-reviews?select=Reviews.csv dataset and move the Reviews.csv file into the main directory.
Then, you have to run Speech-to-text.py. This will create two files in the directory- recording.wav and speech.txt.
Finally, run the Analyser.py which will add the sentiment in another line just below the transcribed text in speech.txt.
Keep in mind this is a shallow model using just countvectorizer and Logistic Regression. Some of the further improvements and interesting things that can be done are:
- Use a more complex model, maybe a sequential neural network?
- Use a better feature extractor such as tf-idf vectorizer.
- Plot the embededded word vectors and see how close the similar words lie.
Thats all from me. See you next time.
Laters!
