THE CONCEPT
This generative AI framework explores the intersection of classical literary structure and modern linguistic expressiveness to algorithmically reimagine the works of W.B. Yeats and William Shakespeare. By synthesizing Elizabethan aesthetics with contemporary English, the system enables the autonomous creation of poetic verses that bridge the gap between historical elegance and algorithmic innovation.
THE ENGINEERING
I architected a sophisticated natural language generation (NLG) pipeline that moves beyond the limitations of stochastic n-gram models to achieve long-term thematic coherence. The core engine utilizes a Bidirectional Long Short-Term Memory (Bi-LSTM) neural network, which processes text in both forward and backward directions to capture deep semantic dependencies. I curated a specialized corpus of 30,000 tokens, refining the training data to balance the intricate sonnet structures of the 16th century with the accessible imagery of early 20th-century lyricism. To move the output closer to intentional art, I implemented an experimental custom loss function designed to prioritize phonological patterns and rhyming constraints. The resulting system evaluates its own creative output through the lens of computational creativity frameworks, ensuring each stanza maintains high stylistic value and structural integrity.
TECH STACK
Neural Networks: Bidirectional LSTM (Long Short-Term Memory), RNNs.
Languages & Frameworks: Python, TensorFlow/Keras.
Natural Language Processing: NLTK, Custom Tokenization, Markov Chains (Baseline).
Data Science: Pandas, NumPy, Custom Corpus Curation.
Creative Frameworks: Computational Creativity Evaluation Metrics.