Sentiment Analysis of Online Reviews
I worked on a project to help the client better understand customer opinions about their products by analyzing online reviews. Using Python with BeautifulSoup, I scraped reviews from a selection of websites to create an initial dataset. I cleaned and preprocessed the texts from the reviews using NLTK and, since the reviews were written in Portuguese, I used an adapted version of the Vader lexicon and rule-based sentiment analysis library for Portuguese text called Leia. With this tool, we can now categorize the reviews as positive, negative, or neutral.
To make these insights accessible to everyone on the client's team, I created an interactive dashboard using Streamlit. The dashboard showed the sentiment breakdown and key themes from the reviews, while also letting users filter the data as they like.
And to keep the dashboard always up to date, I automated the scrapping process with a scheduled task that ran daily. This new data is sent through the preprocessing and sentiment analysis pipeline and then updates the table in the database used by the dashboard.