The repository contains team SPPU_AKAH's submission for the ICON 2020 TechDOfication Shared Task (Subtask-1f). We propose a hybrid BiLSTM-CNN based Attention Ensemble model for the task of coarse-grained automatic technical domain identification of short texts in the Marathi Language. Our system resulted in the best submission for the subtask-1f.
machine-learning
natural-language-processing
deep-learning
text-classification
word-embeddings
shared-task
domain-identification
marathi-text-classification
nlp-for-indian-languages
-
Updated
Dec 28, 2020 - Jupyter Notebook