Exploring NLP techniques across Supply Chain Management and Medical/Healthcare domains using deep learning approaches.
Our group project for ADTA 5760 explores the intersection of Natural Language Processing and Artificial Neural Networks, focusing on real-world applications in two critical domains.
Applying NLP techniques to extract insights from supply chain documents, including procurement reports, logistics data, and operational documentation. Our analysis reveals patterns in supplier communications, demand forecasting narratives, and risk assessment language.
Leveraging neural network-based NLP to process medical documents, clinical reports, and healthcare literature. Our research focuses on extracting meaningful information from unstructured medical text, identifying key entities, and understanding document sentiment and themes.
Deep-diving into two high-impact domains where NLP can transform how organizations process and understand large volumes of text data.
The supply chain domain generates vast amounts of unstructured text — from procurement emails and contracts to logistics reports and supplier evaluations. Our research applies NLP to make this data actionable.
Healthcare organizations deal with massive amounts of clinical text, research papers, and patient documentation. NLP can help extract structured insights from this unstructured data, improving decision-making and research efficiency.
Our end-to-end NLP pipeline from raw PDF documents to actionable insights.
Gathering 100+ PDF documents across supply chain and medical domains from academic papers, industry reports, and public datasets.
PDF text extraction, cleaning, tokenization, stop-word removal, lemmatization, and domain-specific preprocessing.
TF-IDF vectorization, word embeddings (Word2Vec, GloVe), and contextual embeddings using pre-trained transformers.
Training RNN, LSTM, and Transformer-based models for text classification, NER, and sentiment analysis tasks.
Model evaluation using accuracy, F1-score, precision, and recall. Results visualized through interactive charts and dashboards.
Key findings from our NLP analysis across both research domains.
Average accuracy across document classification tasks using fine-tuned neural network models.
Named entity recognition performance on domain-specific entities in both supply chain and medical texts.
Speed improvement over manual document review when using our automated NLP pipeline.
Total PDF documents processed and analyzed across both supply chain and medical domains.
Models pre-trained on general text and fine-tuned on domain-specific data consistently outperform generic models by 8-12% on classification tasks.
Medical documents present unique challenges — specialized vocabulary, abbreviations, and complex sentence structures require domain-specific preprocessing.
Transformer-based architectures (BERT, BioBERT) consistently outperform RNN/LSTM baselines on all text classification and NER tasks.
Automated NLP pipelines can significantly reduce manual effort in processing supply chain and medical documents, enabling faster decision-making.
Graduate students in the Advanced Data Analytics program at the University of North Texas.
Focused on supply chain document analysis, model architecture design, and project coordination.
Specializing in medical document preprocessing, data cleaning, and feature engineering for NLP models.
Working on neural network model training, hyperparameter tuning, and evaluation metrics.