# Paradigm Shift in Language Modeling: Revisiting CNN for Modeling Sanskrit Originated Bengali and Hindi Language

arXiv: https://arxiv.org/abs/2110.13032 · HTML: https://mlsanity.com/paper/2110.13032/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 49 references that are in the map.
- First posted on arXiv 25 Oct 2021.

## Builds on (most cited first, 19 of 19)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Neural Machine Translation by Jointly Learning to Align and Translate](https://mlsanity.com/paper/1409.0473/) (2014)
- [Enriching Word Vectors with Subword Information](https://mlsanity.com/paper/1607.04606/) (2016)
- [Effective Approaches to Attention-based Neural Machine Translation](https://mlsanity.com/paper/1508.04025/) (2015)
- [Regularizing and Optimizing LSTM Language Models](https://mlsanity.com/paper/1708.02182/) (2017)
- [Character-Level Language Modeling with Deeper Self-Attention](https://mlsanity.com/paper/1808.04444/) (2018)
- [Character-Aware Neural Language Models](https://mlsanity.com/paper/1508.06615/) (2015)
- [Language Modeling with Deep Transformers](https://mlsanity.com/paper/1905.04226/) (2019)
- [A Tensorized Transformer for Language Modeling](https://mlsanity.com/paper/1906.09777/) (2019)
- [Towards Understanding Regularization in Batch Normalization](https://mlsanity.com/paper/1809.00846/) (2018)
- [word2vec Parameter Learning Explained](https://mlsanity.com/paper/1411.2738/) (2014)
- [Probabilistic FastText for Multi-Sense Word Embeddings](https://mlsanity.com/paper/1806.02901/) (2018)
- [Do Convolutional Networks need to be Deep for Text Classification ?](https://mlsanity.com/paper/1707.04108/) (2017)
- [Neural Word Segmentation Learning for Chinese](https://mlsanity.com/paper/1606.04300/) (2016)
- [Character n-gram Embeddings to Improve RNN Language Models](https://mlsanity.com/paper/1906.05506/) (2019)
- [Synthetic Error Dataset Generation Mimicking Bengali Writing Pattern](https://mlsanity.com/paper/2003.03484/) (2020)
- [Syllable-aware Neural Language Models: A Failure to Beat Character-aware Ones](https://mlsanity.com/paper/1707.06480/) (2017)

## Abstract (arXiv)

Though there has been a large body of recent works in language modeling (LM) for high resource languages such as English and Chinese, the area is still unexplored for low resource languages like Bengali and Hindi. We propose an end to end trainable memory efficient CNN architecture named CoCNN to handle specific characteristics such as high inflection, morphological richness, flexible word order and phonetical spelling errors of Bengali and Hindi. In particular, we introduce two learnable convolutional sub-models at word and at sentence level that are end to end trainable. We show that state-of-the-art (SOTA) Transformer models including pretrained BERT do not necessarily yield the best performance for Bengali and Hindi. CoCNN outperforms pretrained BERT with 16X less parameters, and it achieves much better performance than SOTA LSTM models on multiple real-world datasets. This is the first study on the effectiveness of different architectures drawn from three deep learning paradigms - Convolution, Recurrent, and Transformer neural nets for modeling two widely used languages, Bengali and Hindi.