Sveva Pepe Portfolio

About Me

NLP & Artificial Intelligence Engineer

Birthday: 23 Dec 1997
City: Rome, Italy
Age: 26

Bachelor's degree: Computer Science and Automation Engineer
Master's degree: Artificial Intelligence and Robotics
Email: sveva.pepe@gmail.com

Curriculum Vitae

Coding Skills

Python 95%

C/C++ 70%

HTML & CSS 90%

PHP 80%

Matlab 80%

Java 75%

Resume

Education

Master of Science in Artificial Intelligence & Robotics

September 2019 - October 2021

Sapienza University of Rome, Rome, IT

Following my bachelor's degree, I pursued a master's degree in artificial intelligence and robotics, concentrating in the machine learning sector, with a particular interest in natural language processing.

Bachelor of Science in Computer Science & Automation Engineering

September 2016 - July 2019

Sapienza University of Rome, Rome, IT

I began my academic journey in 2016 by enrolling in the Computer Science and Automatic Engineering course, which I completed in 2019 with a mark of 110 cum laude.

High School diploma

September 2011 - July 2016

Amedeo Avogadro High School, Rome, IT

In 2016, I graduated from the Amedeo Avogadro High School, which teaches only scientific topics.

Experience

NLP Engineer

March 2022 - Present

Almawave, Rome, IT

Research and development applied in the NLP field produce models and algorithms.
Creation and industrialization of the models within the Iride® platform.
Training models with consolidated best practices, methodologies, and tools of AI / Ops.

Performance Engineer

November 2021 - March 2022

Moviri, Milan, IT

Performance Testing and Tuning with LoadRunner, AppDynamics and Instana.

Scholarship for Research Activities

November 2020 - Jun 2021

Sapienza University of Rome, Rome, IT

The L3DAS project (Learning 3D Audio Sources) aims at encouraging and fostering research on the afore-mentioned topics.

We build L3DAS dataset that contains multiple-source and multiple-perspective B-format Ambisonics audio recordings. The acoustic field is sampled of a large office room, placing two first-order Ambisonics microphones in the center of the room and moving a speaker reproducing the analytic signal in 252 fixed spatial positions.
Relying on the collected Ambisonics impulse responses (IRs), existing clean monophonic datasets is augmented to obtain synthetic tridimensional sound sources by convolving the original sounds with our IRs. Clean files have been extracted from the Librispeech and FSD50K datasets.
Then in particular, we analize two different tasks: i) 3D Speech Enhancement and ii) 3D Sound Source Localization and Detection. In the first, the objective is the enhancement of speech signals immersed in a noisy 3D environment, instead, in the second, the aim is to detect the temporal activities of a known set of sound event classes and to further locate them in the space, in particular, focusing on the sound event localization and detection (SELD).

Work Student

July 2016 - August 2016

Realtime Technologies LTD, Dublin, IRL

Testing electonic hardware in the form of Class 2DSEP printed circuit bords and electro-mechanical systems to ensure their functional compliance for biomedical industry.
Testing involves validation, verification and regression testing of the electronic hardware prior to the initial field trail of prototypes and subsequent production release of both new and existing biomedical products.
Assembling the electronic hardware into biomedical sensing devices for use in galvanic response, electromyography, electrocardiogram ad strain gauge amplifying.
Programming test software to instruct and control the running of the test procedure for the comparison of actual to predicted results using code-drievn testing.
Monitoring test data and corresponding two dimensional electrical signal profiles of the electronic hardware operating under test using graphical user interface testing.

Projects

All
Robotics
HRI
Machine Learning
NLP
Other

STEPS: Semantic Typing of Event Processes with a Sequence-to-Sequence Approach

A Machine Learning Approach to assess the Interlocutor Attention

Optimal Trajectory generation in the Duckietown environment

MARIO: Pepper assistant for patients with senile dementia

Duck Hunt 3D: a remake of the light gun shooter video game

GP Motorcycle Racing Ontology: an ontology on the domain of gran prix motorcycle racing

Share the music: chat with others who share your musical preferences

Named Entity Recognition: identifying and recognizing entities through text

Regular Decision Processes: non-Markovian extension of MDPs

DQNN: Deep Quaternion Neural Network for 3D Sound Source Localization and Detection

Semantic Role Labeling: finding semantic roles for each predicate in a sentence

ASL: Classification of American Sign Language through Neural Networks

Reinforcement Learning: DQN algorithm in Assault-v0 Env

Publications

AAAI 2022

STEPS: Semantic Typing of Event Processes with a Sequence-to-Sequence Approach

Pepe, Sveva, Barba, Edoardo, Blloshmi, Rexhina and Navigli, Roberto

Association for the Advancement of Artificial Intelligence, 2022.

Abtract Paper Code

Enabling computers to comprehend the intent of human actions by processing language is one of the fundamental goals of Natural Language Understanding. An emerging task in this context is that of free-form event process typing, which aims at understanding the overall goal of a protagonist in terms of an action and an object, given a sequence of events. This task was initially treated as a learning-to-rank problem by exploiting the similarity between processes and action/object textual definitions. However, this approach appears to be overly complex, binds the output types to a fixed inventory for possible word definitions and, moreover, leaves space for further enhancements as regards performance. In this paper, we advance the field by reformulating the free-form event process typing task as a sequence generation problem and put forward STEPS, an end-to-end approach for producing user intent in terms of actions and objects only, dispensing with the need for their definitions. In addition to this, we eliminate several dataset constraints set by previous works, while at the same time significantly outperforming them. We release the data and software at https://github.com/SapienzaNLP/steps.

MLSP 2021

L3DAS21 Challenge: Machine Learning for 3D Audio Signal Processing

Guizzo, Eric, F. Gramaccioni, Riccardo, Jamili, Saeid, Marinoni, Christian, Massaro, Edoardo, Medaglia, Claudia, Nachira, Giuseppe, Nucciarelli, Leonardo, Paglialunga, Ludovica, Pennese, Marco, Pepe, Sveva, Rocchi, Enrico, Uncini, Aurelio and Comminiello, Danilo

IEEE International Workshop on MACHINE LEARNING FOR SIGNAL PROCESSING, 2021.

Abtract Paper Code

The L3DAS21 Challenge is aimed at encouraging and fostering collaborative research on machine learning for 3D audio signal processing, with particular focus on 3D speech enhancement (SE) and 3D sound localization and detection (SELD). Alongside with the challenge, we release the L3DAS21 dataset, a 65 hours 3D audio corpus, accompanied with a Python API that facilitates the data usage and results submission stage. Usually, machine learning approaches to 3D audio tasks are based on single-perspective Ambisonics recordings or on arrays of single-capsule microphones. We propose, instead, a novel multichannel audio configuration based multiple-source and multiple-perspective Ambisonics recordings, performed with an array of two first-order Ambisonics microphones. To the best of our knowledge, it is the first time that a dual-mic Ambisonics configuration is used for these tasks. We provide baseline models and results for both tasks, obtained with state-of-the-art architectures: FaSNet for SE and SELDNet for SELD. This report is aimed at providing all needed information to participate in the L3DAS21 Challenge, illustrating the details of the L3DAS21 dataset, the challenge tasks and the baseline models.

Contact Me

Email:

sveva.pepe@gmail.com

Telegram:

@svevapepe

Twitter:

@svevapepe

Instagram:

@sveva_pepe

Github:

pepes97