Personal Projects
Here, you will find a variety of curated projects, ranging from university assignments to personal pursuits.
Multi-Head Neural Network for Emotion Classification and Trigger Detection in Multi-Speaker Dialogues
In this project, we tackle the Emotion Discovery and Reasoning its Flip in Conversation (EDiReF) task, which focuses on predicting the emotion of each sentence in a dialogue and identifying any emotional shifts of speakers, or “triggers”. Our methodology began with creating a BERT-based multi-head neural network, which we then aimed to refine by incorporating a concatenation method, a transformer, and a large language model. The outcomes demonstrated strong capabilities in emotion prediction, yet the performance in identifying emotional triggers was less satisfactory. Through our efforts to enhance the model with various improvements, we observed a marginal but significant improvement in the large language model ability to detect triggers. Still, more fine-tuning is required to reliably recognize emotional changes within dialogues.
nlp
emotion-detection
pytorch
multi-task-learning
python
Sensitive Text Clustering
Neural text clustering solution to automatically group and label sensitive, confidential, and personal information types, making it easier to manage and secure. Given a dataset where each sentence (by human judgment) provides some sensitive personal information about a person, write an ML algorithm that clusters data into different types of sensitive personal information and automatically assigns a label to each cluster.
nlp
text-clustering
python
Probabilistic reasoning over credit cards default
This project work - part of the exam related to the Fundamentals of AI and KR course - aims to determine whether differently structured Bayesian Networks infer queries with different accuracy. In particular, using a unique credit card default dataset, we investigate how features such as age, education, and sex might affect an individual’s risk of credit card default using differently designed Bayesian Networks to compare the results.
probabilistic-inference
bayesian-network
python
pgmpy
Research
I am focused on the Natural Language Understanding field, with a particular interest in its applications within the biomedical domain.
Text-to-Text Extraction and Verbalization of Biomedical Event Graphs
Biomedical events represent complex, graphical, and semantically rich interactions expressed in the scientific literature. Almost all contributions in the event realm orbit around semantic parsing, usually employing discriminative architectures and cumbersome multi-step pipelines limited to a small number of target interaction types. We present the first lightweight framework to solve both event extraction and event verbalization with a unified text-to-text approach, allowing us to fuse all the resources so far designed for different tasks. To this end, we present a new event graph linearization technique and release highly comprehensive event-text paired datasets, covering more than 150 event types from multiple biology subareas (English language). By streamlining parsing and generation to translations, we propose baseline transformer model results according to multiple biomedical text mining benchmarks and NLG metrics.
python
nlp
biomedical-events
text2text