Hi, I'm Ryan đź‘‹
Solutions Architect at Amazon Web Services | Full Stack Software Engineer | Humor & NLP Researcher | Musician
RD
About
Full Stack Software Engineer with close to four years of professional experience, possessing robust foundational skills and a history of collaboration with both startups and large corporations. Highly self-motivated and committed to continuous learning, I bring a diverse skill set aimed at making a significant impact through my work.
Work Experience
Skills
Python
FasAPI
Amazon Web Services
AI Agents
Large Language Models
React
Next.js
Typescript
Node.js
Dart
Flutter
GraphQL
NoSQL
SQL
Elasticsearch
Git
CI/CD
Docker
Kubernetes
Research
Selected Publications
- A
Augmenting Large Language Models with Humor Theory To Understand Puns
Master's Thesis, Purdue University, USA
TThis research applies large language models (LLMs) to pun comprehension, testing two humor theories—the Computational Model of Humor and the Benign Violation theory. By altering theory-specific conditions in a curated English pun dataset, the study evaluates how well LLMs classify puns under each framework. The findings reveal how different theoretical components affect LLM performance, offering deeper insights into humor mechanics and the practical application of humor theories to computational pun analysis. - F
From Sentence Embeddings to Large Language Models to Detect and Understand Wordplay
Lecture Notes in Computer Science, Springer Nature Switzerland
The study presents work on pun detection, location, and interpretation, showing how methods evolved from sentence embeddings to BERT-based models and LLMs. Results reveal strengths and limitations of each approach, as well as challenges in handling nuance, multilingual data, and contextual meanings, offering insights for advancing computational humor analysis. - P
PunDerstand @ CLEF JOKER 2024: Who's Laughing Now? Humor Classification by Genre & Technique
CLEF 2024, Grenoble, France
Humor is subject to individual interpretation, with each person perceiving it differently. Given that humor itself is subjective, this work explores classification of humor by genre and technique through three approaches: manual guided annotation, multi-class classification using BERT-based models with and without sampling, and prompting with large language models. Our experiments revealed insights into the performance of different models and approaches on the humor classification task and opens up further discussions on using guidelines from the annotation to aid large language models. - A
AKRaNLU @ CLEF JOKER 2023: Using Sentence Embeddings and Multilingual Models to Detect and Interpret Wordplay
CLEF 2023, Thessaloniki, Greece
The paper presents work from the Automatic Wordplay Analysis (JOKER) Lab at CLEF 2023, focusing on pun detection and pun location with interpretation. For pun detection, sentence embeddings were used to classify puns. For pun location, the task was modeled as token classification with XLM-RoBERTa. To interpret puns, sentence embeddings combined with WordNet helped identify the intended senses of pun words.