Development of an NLP algorithm to temporally track patient emotional state using digital footprint
X Start of:
X Week 1: Meetings w/ narges, meetings w/ each other, work schedule all set; Stanford 224n course close to if not entirely completed (Rohan and Kamil); identify professors who could give advice (psychology etc)
X Week 2: Data collected; database and management system in rough stages
Week 3: Data cleaned; database complete and functional; exhaustive list of ideas for implementation of model created and narrowed
^^ very meeting heavy week, can be distributed across week 4 as well if neeeded
Week 4: Papers read to finalize few models to try, tasks delegated as necessary; evaluation metrics set; human accuracy estimated; any data needing labeling labeled; any learning needing to be done identified (torch, deeper research into particular topics or implementation methods)
Spend Week 4-5 learning and implementing/modifying other models on data
Start of Week 6: benchmark results acquired, identify and consolidate where gains and improvements can be made, divide work as necessary
Week 6-9: continue improving model and pivoting as necessary until results are achieved
Week 10: Clean up, comment, and optimize code; throw together react to demonstrate results on real graph (optional)
LIWC Regression Models (Linear, Logistic) SVM Vectorize & Cluster Vectorize w/ Bigrams & Cluster RNN LSTM GRU Transformer 1D CNN Bi-Directional Attention BERT