Implementing a neural machine translation engine for mobile devices: the Lingvanex use case

Implementing a neural machine translation engine for mobile devices: the Lingvanex use case Zuzanna Parcheta1 Germ´an Sanchis-Trilles1 Aliaksei Rudak2 Siarhei Bratchenia2 1Sciling S.L., Carrer del Riu 321,Pinedo, 46012, Spain 2Nordicwise LLC, 1 Apriliou, 52, Athienou, Larnaca, 7600, Cyprus {zparcheta, gsanchis}@sciling.com {alrudak, s.bratchenya}@lingvanex.com Abstract In this paper, we present the challenge entailed by implementing a mobile version of a neural machine translation system, where the goal is to maximise translation quality while minimising model size. We explain the whole process of implementing the translation engine on an English–Spanish example and we describe all the difficulties found and the solutions implemented. The main techniques used in this work are data selection by means of Infrequent n-gram Recovery, appending a special word at the end of each sentence, and generating additional samples without the final punctuation marks. The last two techniques were devised with the purpose of achieving a translation model that generates sentences without the final full stop, or other punctuation marks. Also, in this work, the Infrequent n-gram Recovery was used for the first time to create a new corpus, and not enlarge the in-domain dataset. Finally, we get a small size model with quality good enough to serve for daily use. 1 Introduction Lingvanex1 is a trademark for linguistic products made by Nordicwise LLC company. The main focus of the company are translator and dictionary applications that work without internet connection on mobile and desktop platforms. c⃝ 2018 The authors. This article is licensed under a Creative Commons 3.0 licence, no derivative works, attribution, CC- BY-ND. 1https://lingvanex.com/en/ In collaboration with Sciling2, an agency specialised in providing end-to-end machine learning solutions, a small-sized translation model from English to Spanish was implemented. When implementing a mobile translator, it is crucial to understand its purpose. In our case, the purpose was to be able to generate translations on a daily usage scenario, without requiring a Internet connection. This is the typical use case in a travel scenario, where travellers often do not have an internet connection, either because they do not want to assume the cost of a roaming connection, because they do not want to purchase a local SIM card, or even because there is no good connection in the places they are travelling to, such as some countries of Africa. In this scenario, the main purpose of the translation engine is to be able to translate correctly short sentences, composed of common words in a traveller domain, but where other words belonging to e.g. a parliamentary or a medical domain are less frequent. In addition, the model requires to be contained in terms of size, since large models would perform poorly in a mobile device. In this work, we focus on reducing model size mainly through data selection techniques, until a size of 150MB per model. However, there are other techniques which bring promising results as compressing the NMT model via pruning (See et al., 2016). Along this article we determine what is the main influence to model size. For that, …

Content truncated to 500 words