This project is a music recommendation engine that takes into account the emotions that a particular song evokes. Emotions associated with a song are an abstract concept and can be subjective, that's why our system allows you to update the representation of emotion associated with each song that our model predicted. Over time, as more people use this web app, the model performance will improve and give emotion predictions that are closer to ground truth (or at least what majority of people believe the emotional ratings of songs to be).

GitHub Link: https://github.com/aliforgetti/MoodZen

Application: https://share.streamlit.io/shaswat01/moodzen/main/src/conn.py

Humans are able to identify similarities in stories by identifying the similarities within different elements of a narrative. For example, the classic story about the rabbit and the hare is the classic example of a story where the underdog wins unexpectedly. Previous work explored identifying similar stories in terms of similarities in the plot events and resemblances between characters and their social relationships. This project aims to explore another dimension - the sentiment trajectory. To achieve this, two models were trained on the TweetEval dataset to be able to detect the sentiment of a tweet. The best model (Facebook's RoBERTa) was then used to extract the sentiment trajectory of a given story and then compared to other stories to find stories similar in how their narratives evolved temporally.

Note: Parts of this project are currently in-progress but the report present here should be able to give you an idea of the work that was done.

Image classification application to identify lung disease from X-Ray scans and cancerous lesions from CT Scans.

GitHub Link: https://github.com/aliforgetti/deep_XT

Application creating using flutter and Nashville crime data to detect the frequency of crime along your walking path.