Scalable machine translation in memory constrained environments

Machine translation is the discipline concerned with developing automated tools for translating from one human language to another. Statistical machine translation (SMT) is the dominant paradigm in this field. In SMT, translations are generated by means of statistical models whose parameters are learned from bilingual data. Scalability is a key concern in SMT, as one would like to make use of as much data as possible to train better translation systems. In recent years, mobile devices with adequate computing power have become widely available. Despite being very successful, mobile applications relying on NLP systems continue to follow a client-server architecture, which is of limited use because access to internet is often limited and expensive. The goal of this dissertation is to show how to construct a scalable machine translation system that can operate with the limited resources available on a mobile device. The main challenge for porting translation systems on mobile devices is memory usage. The amount of memory available on a mobile device is far less than what is typically available on the server side of a client-server application. In this thesis, we investigate alternatives for the two components which prevent standard translation systems from working on mobile devices due to high memory usage. We show that once these standard components are replaced with our proposed alternatives, we obtain a scalable translation system that can work on a device with limited memory. The first two chapters of this thesis are introductory. Chapter 1 discusses the task we undertake in greater detail and highlights our contributions. Chapter 2 provides a brief introduction to statistical machine translation. In Chapter 3, we explore online grammar extractors as a memory efficient alternative to phrase tables. We propose a faster and simpler extraction algorithm for translation rules containing gaps, thereby improving the extraction time for hierarchical phase-based translation systems. In Chapter 4, we conduct a thorough investigation on how neural language models should be integrated in translation systems. We settle on a novel combination of noise contrastive estimation and factoring the output layer using Brown clusters. We obtain a high quality translation system that is fast both when training and decoding and we use it to show that neural language models outperform traditional n-gram models in memory constrained environments. Chapter 5 concludes our work showing that online grammar extractors and neural language models allow us to build scalable, high quality systems that can translate text with the limited resources available on a mobile device.