Senior Projects - AI Research & Development
Senior Projects in Natural Language Processing & Emotional AI
Welcome to my senior project research on text embedded AI models. These represent my exploration into teaching machines to understand not just what people say, but how they say it and what it means.
Featured Projects
Emotionally Adaptive Chatbots
Building Empathy with AI Through VAD Emotion Detection
This groundbreaking project teaches AI to detect and respond to human emotions in text. Using Valence-Arousal-Dominance (VAD) scoring, I achieved variance scores above 0.70 with over 300,000 labeled entries, enabling chatbots that don’t just understand words, but feelings.
Key Achievements: - Variance scores > 0.70 across all VAD dimensions - Over 300,000 emotion-labeled text entries - Real-time emotional tone adaptation - Customer service empathy modeling
Technologies: Python, NLP, Sentence Transformers, spaCy, VAD Modeling
AI-Powered Resume Scoring System
From Reading Emotions to Reading Resumes
Leveraging the breakthroughs from the VAD emotion project, this system uses semantic similarity and pretrained text embeddings to match resumes to job descriptions, not by counting keywords, but by understanding meaning.
Key Features: - Semantic matching (not just keyword scanning) - Intelligent keyword extraction with NLP - Weighted scoring (65% meaning + 35% keywords) - Real-time resume analysis
Technologies: Python, Sentence Transformers, spaCy, Flask, NLP
The Connection: Transfer Learning in Action
These projects aren’t just separate experiments. They’re connected by a powerful insight:
When you teach AI to understand one type of context (emotions), you unlock the ability to understand all types of context (skills, intent, meaning).
The pretrained text embeddings and semantic similarity techniques developed for emotion detection became the foundation for the resume scoring system. This is transfer learning in action: knowledge from one domain (emotional intelligence) accelerating breakthroughs in another (professional skill matching).
Research Impact
| Metric | Achievement |
|---|---|
| Training Data | 300,000+ labeled text entries |
| Model Variance | >0.70 across all dimensions |
| Resume Scorer Weighting | 65% semantic + 35% keyword coverage |
| Deployment | Production-ready web applications |
Technologies & Skills Demonstrated
- Natural Language Processing (NLP): spaCy, NLTK, text preprocessing
- Machine Learning: Sentence Transformers, embeddings, transfer learning
- Deep Learning: Pretrained models, fine-tuning, semantic similarity
- Data Engineering: Large-scale text corpus management
- Web Development: Flask applications, deployment pipelines
- Statistical Analysis: Variance analysis, scoring algorithms
- Cloud Deployment: Render.com web services
What’s Next
These projects represent Phase 1 of an ongoing research initiative. Future developments include:
- Industry-specific semantic models for specialized domains
- Real-time feedback systems for resume optimization
- Multi-document ranking algorithms for candidate comparison
- Integration with enterprise ATS platforms
- Expanded emotion detection for mental health and education applications
Connect
Interested in collaborating, learning more, or testing these systems?
Email: official@wilkinjones.com
Schedule a Call: calendly.com/official-wilkinjones
LinkedIn: Connect with me
These projects represent the intersection of emotional intelligence, natural language understanding, and practical AI applications, built with real data, real code, and real impact.