Massey Documents by Type

Now showing 1 - 3 of 3

Deep learning for low-resource machine translation : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Computer Science, School of Mathematical and Computational Sciences, Massey University, Albany, Auckland, New Zealand. EMBARGOED until further notice.
(Massey University, 2025-09-01) Gao, Yuan
Machine translation, a key task in natural language processing, aims to automatically translate text from one language to another while preserving semantic integrity. This thesis builds upon existing research and introduces three deep-learning methods to enhance translation performance under low-resource conditions: (i) an effective transfer learning framework that leverages knowledge from high-resource language pairs, (ii) a pre-ordering-aware training method that explicitly utilizes contextualized representations of pre-ordered sentences, and (iii) a data augmentation strategy that expands the training data size. Firstly, we develop a two-step fine-tuning (TSFT) transfer learning framework for low-resource machine translation. Due to the inherent linguistic divergence between languages in parent (high-resource language pairs) and child (low-resource language pairs) translation tasks, the parent model often serves as a suboptimal initialization point for directly fine-tuning the child model. Our TSFT framework addresses this limitation by incorporating a pre-fine-tuning stage that adapts the parent model to the child source language characteristics, improving child model initialization and overall translation quality. Secondly, we propose a training method that enables the model to learn pre-ordering knowledge and encode the word reordering information within the contextualized representation of source sentences. Pre-ordering refers to rearranging source-side words to better align with the target-side word order before translation, which helps mitigate word-order differences between languages. Existing methods typically integrate the information of pre-ordered source sentences at the token level, where each token is assigned a local representation that fails to capture broader contextual dependencies. Moreover, these methods still require pre-ordered sentences during inference, which incur additional inference costs. In contrast, our method enables the model to encode the pre-ordering information in the contextualized representations of source sentences. In addition, our method eliminates the need for pre-ordering sentences at inference time while preserving its benefits in improving translation quality. Thirdly, to address data scarcity in low-resource scenarios, we propose a data augmentation strategy that employs high-quality translation models trained bidirectionally on high-resource language pairs. This strategy generates diverse, high-fidelity pseudo-training data through systematic sentence rephrasing, generating multiple target translations for each source sentence.. The increased diversity on the target side enhances the model's robustness, as demonstrated by significant performance improvements in eight pairs of low-resource languages. Finally, we conduct an empirical study to explore the potential of applying ChatGPT for machine translation. We design a set of translation prompts incorporating various auxiliary information to assist ChatGPT in generating translations. Our findings indicate that, with carefully designed prompts, ChatGPT can achieve results comparable to those of commercial translation systems for high-resource languages. Moreover, this study establishes a foundation for future research, offering insights into prompt engineering strategies for leveraging large language models in machine translation tasks.
An investigation into teaching description and retrieval for constructed languages : a thesis presented in partial fulfilment of the requirements for the degree of Master of Science in Computer Science at Massey University
(Massey University, 2004) Hoang, Son
The research presented in this thesis focuses on an investigation on teaching concepts for constructed languages, and the development of a teaching tool, called VISL, for teaching a specific constructed language. Constructed languages have been developed for integration with computer systems to overcome ambiguities and complexities existing in natural language in information description and retrieval. Understanding and using properly these languages is one of the keys for successful use of these computer systems Unfortunately, current teaching approaches are not suitable for users to learn features of those languages easily. There are different types of constructed languages. Each has specific features adapted for specific uses but they have in common explicitly constructed grammar. In addition, a constructed language commonly embeds a powerful query engine that makes it easy for computer systems to search for correct information from descriptions following the conditions of the queries. This suggests new teaching principles that should be easily adaptable to teach any specific structured language's structures and its specific query engine. In this research, teaching concepts were developed that offer a multi-modal approach to teach constructed languages and their specific query engines. These concepts are developed based on the efficiencies of language structure diagrams over the cumbersome and non-transparent nature of textual explanations, and advantages of active learning strategies in enhancing language understanding. These teaching concepts then were applied successfully for a constructed language, FSCL, as an example The research also explains howr the concepts developed can be adapted for other constructed languages. Based on the developed concepts, a Computer Aided Language Learning (CALL) application called VISL is built to teach FSCL. The application is integrated as an extension module in PAC, the computer system using FSCL for description and retrieval of information in qualitative analysis. In this application, users will learn FSCL through an interconnection of four modes: FSCL structures through the first two modes and its specific query engine through the sccond two modes After going through four modes, users will have developed full understanding for the language. This will help users to construct a consistent vocabulary database, produce descriptive sentences conducive to retrieval, and create appropriate query sentences for obtaining relevant search results.
The multimedia documentation of endangered and minority languages : a thesis presented in partial fulfilment of the requirements for the degree of Master of Philosophy in Linguistics at Massey University
(Massey University, 2002) Petterson, Robert
This thesis examines the impending loss of linguistic diversity in the world and advocates a change in emphasis in linguistic research towards the documentation of minority and endangered languages. Various models for documentation are examined, along with some of the ethical issues involved in linguistic research amongst small groups, and a new model is proposed. The new model is centred around the collection of a wide variety of high-quality data, but includes the collection of other related materials that will be of particular use and interest to the ethnic community. The collected data and other materials are then structured as an internet-ready multimedia documentation designed for use by the ethnic community as primary audience, while still catering for the needs of linguistic researchers worldwide. A pilot project is carried out using the model.

Massey Documents by Type

Browse

Filters

Settings

Sort By

Results per page

Search Results