Guided Transformer for Machine Translation: English to Hindi

In the area of machine translation, the objective is to develop a model which can translate a piece of text from one language to another language. Large sized neural networks requires a large amount of training data to achieve a good performance level. The objective of this study is to build an expl...

Full description

Saved in:
Bibliographic Details
Main Authors: Bisht, Akhilesh, Gupta, Deepa, Parida, Shantipriya
Format: Conference Proceeding
Language:English
Subjects:
Blue-score
Data models
Decoder
Dependency Graph
Dependency Parser
Encoder
English-Hindi Translation
Explicit Knowledge
Knowledge engineering
Machine translation
Magnetic heads
Multi head attention
Neural networks
Positional encoding
Self-attention
Training data
Transformers
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the area of machine translation, the objective is to develop a model which can translate a piece of text from one language to another language. Large sized neural networks requires a large amount of training data to achieve a good performance level. The objective of this study is to build an explicit knowledge from the dependency parser, which can be integrated into transformer architecture. Adding an explicit knowledge to a neural machine translation network can enhance its learning process. The baseline transformer model and the proposed guided transformer model were tested on the same test data set of PMIndia Man Ki Baat session. With an addition of explicit knowledge, the proposed guided transformer model surpass the performance of baseline transformer model by +1.0 BLEU (approx.), +1.0 SacreBLEU (approx.), +1.48 Chrf, +1.44 Chrf+ and +1.39 Chrf++.
ISSN:2325-9418
DOI:10.1109/INDICON59947.2023.10440876