MT Research Research Articles

Re-thinking Machine Translation Post-Editing Guidelines

Machine Translation Post-Editing (MTPE) is a challenging task. It frequently creates tension between what the industry expects in terms of quality and what translators are willing to deliver as an end product. Conventional approaches to MTPE take as a point of departure the distinction between light and full MPTE, but the division gets blurred when implemented in an actual MTPE project where translators find difficulties in differentiating between essential and preferential changes. At the time MTPE guidelines were designed, the role of the human translator in the MT process was perceived as ancillary, a view inherited from the first days of MT research aiming at the so-called Fully Automatic High Quality Machine Translation (FAHQMT). My proposal challenges the traditional division of MTPE levels and presents a new way of looking at MTPE guidelines. In view of the latest developments in neural machine translation and the higher quality level of its output, it is my contention that the traditional division of MTPE levels is no longer valid. In this contribution I advance a proposal for redefining MTPE guidelines in the framework of an ecosystem specifically designed for this purpose.

Music Therapists’ Experiences of Therapeutic Relationships with Clients with Marginalized Gender Identities and Sexual Orientations

Abstract This phenomenological inquiry investigated the ways in which music therapists perceive and discuss their music therapy relationships (MTRs) with people with marginalized gender identities and sexual orientations. Five participants completed semi-structured interviews that were transcribed and analyzed with reflexive emphasis on member checking and data triangulation. First-level analysis utilized values coding and second-level analysis refined values themes to distill individual participant essences (Saldaña, 2016). The coding manual for qualitative researchers. SAGE). Subsequent horizontalization across participants was then used to discern collective participant essences (Saldaña, 2016). The coding manual for qualitative researchers. SAGE). Critical feminist, queer, disability, anti-oppressive, and humanist theory lenses shaped the foundation and perception of the project with particular attention to intersectionality, intersubjectivity, and power dynamics in relationships. Findings included commentary on MT/client identities and balance in relationships; contextual influences on MT processes; necessity of cultural humility (CH) and cultural sustainability (CS) and respect for identity complexity; and the role of implicit/explicit communication in MTRs. Crucially, participants and the researcher endorse a non-clinical view of people with marginalized gender identities and sexual orientations, and the researcher urges introspection regarding identity, supervision, and the space that MTs and clients occupy within the MTR construct. Discussion, recommendations for future research and applications to MT theory, research, and practice are provided.

Investigating Contextual Influence in Document-Level Translation

Current state-of-the-art neural machine translation (NMT) architectures usually do not take document-level context into account. However, the document-level context of a source sentence to be translated could encode valuable information to guide the MT model to generate a better translation. In recent times, MT researchers have turned their focus to this line of MT research. As an example, hierarchical attention network (HAN) models use document-level context for translation prediction. In this work, we studied translations produced by the HAN-based MT systems. We examined how contextual information improves translation in document-level NMT. More specifically, we investigated why context-aware models such as HAN perform better than vanilla baseline NMT systems that do not take context into account. We considered Hindi-to-English, Spanish-to-English and Chinese-to-English for our investigation. We experimented with the formation of conditional context (i.e., neighbouring sentences) of the source sentences to be translated in HAN to predict their target translations. Interestingly, we observed that the quality of the target translations of specific source sentences highly relates to the context in which the source sentences appear. Based on their sensitivity to context, we classify our test set sentences into three categories, i.e., context-sensitive, context-insensitive and normal. We believe that this categorization may change the way in which context is utilized in document-level translation.

Are Ellipses Important for Machine Translation?

AbstractThis article describes an experiment to evaluate the impact of different types of ellipses discussed in theoretical linguistics on Neural Machine Translation (NMT), using English to Hindi/Telugu as source and target languages. Evaluation with manual methods shows that most of the errors made by Google NMT are located in the clause containing the ellipsis, the frequency of such errors is slightly more in Telugu than Hindi, and the translation adequacy shows improvement when ellipses are reconstructed with their antecedents. These findings not only confirm the importance of ellipses and their resolution for MT, but also hint toward a possible correlation between the translation of discourse devices like ellipses with the morphological incongruity of the source and target. We also observe that not all ellipses are translated poorly and benefit from reconstruction, advocating for a disparate treatment of different ellipses in MT research.

Awareness and adoption of mobile technologies in the delivery of services in academic libraries in Jordan

PurposeMobile technologies (MT) have been attracting the attention of academic library stakeholders for some time, as it is considered a strong communication channel that can offer advanced library services. Furthermore, MT can offer a number of flexible options for library users. Accordingly, the overarching aim of this research is to investigate the potentials of implementing MT within academic libraries in public universities in Jordan from library staff viewpoint. Exploring library staff expectation of MT to enhance library services is another aim. It is also important to investigate the effects of external factors such age, job title, educational level and experience on library staff perception of MT use and role in academic libraries in Jordan.Design/methodology/approachUsing a case study design, a questionnaire was developed and disseminated to all library staff (n = 174) within ten public universities. Some 121 responses were received.FindingsThe study indicates that there is a high level of awareness of the advantages of using MT in academic libraries in Jordan. However, the research also reveals that, although library staff are aware of the role and value of MT within the academic library environment, MT has not yet been fully exploited in these environments in Jordan for various reasons. These include incompatible information technology infrastructure and a lack of training. As a result, these libraries need to invest more in MT and enhance the use of technologies among their users to raise their performance and the quality of services that they offer. Such enhancements could lead to more meaningful interactions with users and a better overall user experience.Research limitations/implicationsThis study investigated awareness levels among library staff of MT to provide library services. Users perspectives regarding receiving library services on mobile devices have not been explored. MT should be assessed from user perspective to understand how MT might enhance academic library services. Library staff could also be interviewed to explore these issues in closer detail. Library and Information Science departments might also upgrade their curricula and include MT application development in library courses.Originality/valueThis study fills a gap in MT research in the context of library services, as it is the first such attempt to conduct research into this topic in the Middle East in general, and in Jordan specifically.

A Case for Mixed Methods Research in Massage Therapy

-

Issues in Machine Translation

AbstractMachine translation (MT) is still a huge challenge for both IT developers and users. From the beginning of machine translation, problems at the syntactic and semantic levels have been faced. Today despite progress in the development of MT, its systems still fail to recognise which synonym, collocation or word meaning should be used. Although mobile apps are very popular among users, errors in their translation output create misunderstandings. The paper deals with the analysis of machine translation of general everyday language in Lithuanian to English and English to Lithuanian language pairs. The results of the analysis show that more than two thirds of all the sentences were translated incorrectly, which means that there is a relatively small possibility that a mobile app will translate sentences correctly. The results are disappointing, because even after almost 70 years of MT research and improvement, researchers still cannot offer a system that would be able to translate with at least 50% correctness.

Visualization, Search and Analysis of Hierarchical Translation Equivalence in Machine Translation Data

Abstract Translation equivalence constitutes the basis of all Machine Translation systems including the recent hierarchical and syntax-based systems. For hierarchical MT research it is important to have a tool that supports the qualitative and quantitative analysis of hierarchical translation equivalence relations extracted from word alignments in data. In this paper we present such a toolkit and exemplify some of its uses. The main challenges taken up in designing this tool are the efficient and compact, yet complete, representation of hierarchical translation equivalence coupled with an intuitive visualization of these hierarchical relations. We exploit a new hierarchical representation, called Hierarchical Alignment Trees (HATs), which is based on an extension of the algorithms used for factorizing n-ary branching SCFG rules into their minimally-branching equivalents. Our toolkit further provides a search capability based on hierarchically relevant properties of word alignments and/or translation equivalence relations. Finally, the tool allows detailed statistical analysis of word alignments, thereby providing a breakdown of alignment statistics according to the complexity of translation equivalence units or reordering phenomena. We illustrate this with an empirical study of the coverage of inversion-transduction grammars for a number of corpora enriched with manual or automatic word alignments, followed by a breakdown of corpus statistics to reordering complexity.

Developments in MT research in the US

The paper argues that the IBM statistical approach to machine translation has done rather better after a few years than many sceptics believed it could. However, it is neither as novel as its proponents suggest nor is it making claims as clear and simple as they would have us believe. The performance of the purely statistical system (and we discuss what that phrase could mean) has not equalled the performance of SYSTRAN. More importantly, the system is now being shifted to a hybrid that incorporates much of the linguistic information that it was initially claimed by IBM would not be needed for MT. Hence, one might infer that its own proponents do not believe ‘pure’ statistics sufficient for MT of a usable quality. In addition to real limits on the statistical method, there are also strong economic limits imposed by their methodology of data gathering. However, the paper concludes that the IBM group have done the field a great service in pushing these methods far further than before, and by reminding everyone of the virtues of empiricism in the field and the need for large scale gathering of data.

Reaction to Somers: Current research in Machine Translation

To begin with, I would like to welcome the idea of having 'forum' style peer group commentary of papers in MT: one of the less appealing features of the field is the general lack of communication between people who work on different projects, or with different assumptions. Apart from being unappealing, it also damaging to the field, since one effect is that people rarely learn about, and so cannot learn from, each other's mistakes (or successes). It may be that this forum will go some way to change things by encouraging debate. In this context, of course, Harry Somers' (HLS) stimulating paper is a good starting point, because, despite having concentrated mainly on work in some way connected with his own institution in the fourth section ('Some Alternative Avenues of Research), he has given, in the earlier sections, what seems to me a very reasonable summary of what is, and has been, going on. However, I have some fundamental disagreements with the paper. The first is with the perspective that HLS takes in his critique of the 'second generation' . I think HLS assumes that MT is about the construction of working, commercial translation systems. It is with respect to this aim that he finds the second generation a failure. Now, while this is an important aim of MT research, it not the only aim. More generally, the main aim of MT research should be to enhance understanding about what is involved in automating the translation process. This involves (i) addressing issues, and (ii) developing techniques. (It is because of a failure to see this that so many conference papers on MT consist of pictures of outline boxes with arrows going between them, and unrevealing captions). I also think MT research may eventually have something to say about the nature of translation. This is an important goal that goes well beyond, and may be in tension with the practical goal of building MT systems. In a sense this perspective is a puzzling one for HLS to take (e.g. he does not continue with it when he talks in section 4 about avenues of research this is all about techniques), but for whatever reason, I think it leads him wrong in his evaluation of a number of things. One example:

Current research in Machine Translation: A reply to Somers

Somers' paper falls short of its declared aim of demonstrating that: 1) G2 principles of modularity are inadequate; 2) current knowledge-based and linguistics-based MT research is wrong in adhering to these modularity principles; and 3) clearcut alternatives to these principles have now emerged. Finally, I would like to point out that Somers' review of current MT research contains some important omissions. For instance, it seems to me that a comprehensive review of the MT scene can no longer ignore some recent results in multilingual text generation (Kittredge 1989): in certain situations, generation clearly has an edge over MT.

What do we want from MT?

As a long-time student of linguistics and a developer and user of machine translation, I feel moved to respond to Harold Somers' call for new architectures by asking the basic question: What is it that we want from MT? The definition of research priorities must necessarily flow from the purpose(s) sought. The needs will differ depending on what we hope to accomplish. We must first decide whether or not it is a meaningful research goal to use MT merely as a test bed for the solution of challenging linguistic problems in other words, to develop models in which the purpose is to come up with specific solutions rather than to build an integral translation system. Certainly most linguists would agree that linguistic problems can only be dealt with in the context of real language. Man-made scenarios with contrived input do not lead to solutions that stand up in a practical world. Systems must test pieces of real language, no matter how narrow the subset. Moreover, since their structural and lexical elements come from the larger language, the study of such pieces is worthwhile and valid only to the extent that the input is natural and unconstrained and has originally been generated for a clearly defined purpose. Personally, I also believe that the knowledge bases of such systems must incorporate necessary links to the rest of the larger language. In my view, research on selectively encoded toy systems that do not take these criteria into account runs two major risks: (1) it is not likely to stand up theoretically on laiger corpora, and, therefore, (2) it is not likely to contribute to overall progress in machine translation. If this position is accepted, and we agree that MT research should envision the development of working systems, then we must specify goals not for linguistic research as such but rather for integral systems. Several possible goals come to mind:

Responses to Current Research in Machine Translation by Harold Somers

"Current Research in Machine Translation" is a sound presentation of issues from the perspective of long experience with the affirmed MT architectures, and the presumed directions of evolution from one architecture to the next. Somers carries this experience, despite its disappointments, into an enthusiastic outlook of a different set of architectures. There are several points, two in particular, made by Somers' paper which I found particularly gratifying. These assertions agree with some conclusions I had come to recently, and held rather closely, if not privately. These have to do with what is or can be non-linguistic about MT. The first of these points is suggested by the accurate observation that in the (now 25) years since ALPAC, MT does not produce radically improved results over the output available then. I think anyone who compares modern output (both from research and from commercial systems) with the ALPAC set, either from a linguistic coverage or correctness perspective, will come to a similar conclusion. Yet, this is not to say that MT is therefore as useless as it was in 1966. On the contrary, MT has an established place in the commercial world, and a relatively long-term commitment from a variety of governments. This is because, in 1966, input was submitted to attendants of a monolithic "computer" and output was provided in the form of (at best) printouts on a green-bar. There was no way that the information consumer of that time could be expected to modify the output to make it legible. The advent of a user environment, and all that has followed from that, has led to word processing tools that make bad output editable into useful output, in ways that the ALPAC writers could not have envisioned. Thus the good news is that MT quality today is usable; the bad news is that the reason that it is usable has much less to do with MT research than we would like to claim. This is I think, consistent with Somers' point. At the same time, though, I think that there is another side to the quality/usefulness discussion that bears some delving into. MT, and natural language in general, is not going to be used for as many things as the ALPAC-era people may have thought. We thought we would use speech as the primary user interface to computers, and, in advance of that, natural language interfaces. Neither of these appear likely, not only because we do not know how, but because the interim, alternate interface modalities turn out to be better solutions.

Current MT research orientation/disorientation

Somers' article brings into focus the state of technology in the subject of MT. In a certain measure, the reflection of Somers “despite nearly 25 years since the ALPAC report...for all the investment...perhaps one could have expected better results” should be applied not only to MT systems, but equally to NLP systems. However, this reflection of Somers should be interpreted and reevaluated in accordance with the distance between the evolution of our knowledge of the mecanisms of translatology and the ambitious objectives of producing FAHQT for more than 80% of sentences in unrestricted texts.

BITS, A Hebrew-English bibliographic translation system

This article provides an overview of the BITS Hebrew-English bibliographic translation system developed at CCL, UMIST. This is an experimental machine translation system for translating bibliographic references in which first generation translation techniques are combined with more recent developments in computer science. The experiment explores whether simple techniques can be used to achieve usable translations in this context. The input material, the pre-editing system and the general architecture of the BITS system are described and evaluated in view of current MT research.

Machine translation of natural languages

The field of machine translation has recently entered a new, third period in its evolution. In its early period, for roughly fifteen years from 1950 MT was an expanding field of study in which both research and development efforts were undertaken. It is well-known and well documented (Bar Hillel, 1960; ALPAC, 1966) that this early MT paradigm could not and did not produce fully automated high quality translation systems. In fact, the practical results were quite negligible for such a high-scale effort. This period also witnessed an engineering-oriented trend which considered MT primarily an engineering problem. In the positive sense, it helped promote the idea of nonnumerical computation. As another consequence, the engineering, nontheoretical attitude to MT has been inherited by the next period in MT history. A more significant and lasting influence was exerted by the less "practical" aspect of early MT research: it helped achieve advances in theoretical and computational linguistics as well as artificial intelligence.

Present Objectives of MT Research in the United Kingdom

Present Objectives of MT Research in the United Kingdom