# DTCC - Brainstorming

## Previously on DTCC
* ST Embeddings + RF
* GradientTreeBoosting: ~0.62
* PyTorch MLP: ~0.62 (best case)
* gbert-base finetuned
* transformer for SeqClassification: 0.597 (very best case)
* into lstm/gru: 0.63 (0.52 with wrong finetuning -> check out)

## Ideas
* custom encoding (instead of BERT)
* feature ablation
* jiant mtl
* wordnet / germanet
* hf classifiert pooling layer