Noisy Text Classification

Approach 1

​

Stacked DeBERT: BERT-based model with token reconstruction.

​

​

NLP Metrics

Repository with notebooks and explanations regarding NLP Metrics: BLEU, GLEU, WER.

Noisy Text Classification

Approach 2

​

EmbraceBERT: BERT-based model with attentive embracement for robust hidden latent representation.

STT Error

Repository to make dataset with Speech-To-Text error by applying TTS and STT to text.

Sentence Correction

Sentence Correction in incomplete data, obtained from POS-Tagging and deletion of irrelevant words, using Temporal Hierarchies in Sequence to Sequence.