Enhanced Transformer Architecture for Natural Language Processing

TL;DR

This paper introduces an Enhanced Transformer architecture for NLP that improves performance using novel structural modifications, achieving significantly higher BLEU scores without increasing model size.

Contribution

A new Transformer structure with full layer normalization, weighted residuals, reinforcement learning-based positional encoding, and zero masked self-attention is proposed.

Findings

Achieves 202.96% higher BLEU score than original Transformer.

Validated on Multi30k translation dataset.

Demonstrates improved efficiency and performance.

Abstract

Transformer is a state-of-the-art model in the field of natural language processing (NLP). Current NLP models primarily increase the number of transformers to improve processing performance. However, this technique requires a lot of training resources such as computing capacity. In this paper, a novel structure of Transformer is proposed. It is featured by full layer normalization, weighted residual connection, positional encoding exploiting reinforcement learning, and zero masked self-attention. The proposed Transformer model, which is called Enhanced Transformer, is validated by the bilingual evaluation understudy (BLEU) score obtained with the Multi30k translation dataset. As a result, the Enhanced Transformer achieves 202.96% higher BLEU score as compared to the original transformer with the translation dataset.