Skip to main navigation Skip to main content

Farshad Jafari

Photo of Farshad Jafari
Education BS, Amirkabir University of Technology
MS, Amirkabir University of Technology
CCM Lab Role MSMT Student
Pronouns he/him

I am a graduate of Amirkabir University of Technology in Tehran, Iran, with both Bachelor’s and Master’s degrees in Computer Science. My research primarily revolves around leveraging modern computational methods to understand, create, and analyze art and humanities. A significant portion of my work also focuses on developing and evaluating datasets, which are crucial for any computational research. My journey began during my undergraduate studies, where I collected and pre-processed a dataset from movie subtitles for machine translation. To evaluate this dataset, I utilized LSTM models, which were the most advanced at the time for understanding long-term dependencies.

My exploration continued into my Master’s thesis, where I ventured into the realm of music. I managed to create the first digital dataset of Iranian music with lyrics in MusicXML format, annotating musical symbols semi-automatically from printed versions by myself. The aim of my thesis was to automatically generate melodies for given lyrics. Through extensive experimentation with various LSTM model architectures and hyperparameters, I developed a model capable of producing plausible melody progressions. This was my first foray into field research, where I created a survey and collected over hundred responses to assess the model’s ability to produce pleasing outputs.

Currently, my research involves the application of cutting-edge computational models, particularly transformers with GPT architecture, to analyze various musical aspects such as rhythm, harmony, and melody. The initial phase of my work aims to evaluate the predictive prowess of these models. Subsequently, I intend to investigate how different musical features interact and influence these predictions. The overarching goal of my research is to quantify musical aesthetics, a process that necessitates extracting information-theoretic metrics through the predictive analyses provided by these advanced models.

Looking ahead, I anticipate my academic efforts will delve into the intersection of music, film, and narrative forms, seeking to uncover the complex ways these mediums convey meaning and evoke emotions. My goal is to push the boundaries of interdisciplinary and multimodal research in art. I aim to enhance our ability to analyze and interpret complex subjects and to help create new tools for artistic expression and innovative media formats.

Other experience: