Creating an aspect based sentiment analysis system. Trained on manually labelled hotel review data.
This project utlizes deep learning models to create a system which can classify hotel reviews into their aspects (staff, amenities, condition, and cleanliness) and their sentiments (positive, neutral and negative). This project was done in collaboration with Intelligent Systems Laboratories and Assumption University as the final project. The team consisted of William Sivutha Poch, Navin Singh, and Jiadong Jiao.
To demonstrate the deep learning model and showcase the review classifications, a simplistic web UI was created and uses charts and graphs to show the distribution of hotel reviews based on their sentiments and aspects. This main page displays the overall sentiment distributions as well as the distribution per aspect.
Some more details are displayed in this page. The aspect and sentiment of each review is shown here with colors denoting the sentiment (green = positive, yellow = neutral, red = negative).
In aspect-based sentiment analysis system there are 2 main parts, the aspect detector and the sentiment classifier.
More details of the design and implementation can be found in this report link.