A machine learning approach to the identification of translational language: an inquiry into translationese learning models
Authors
Ilisei, Iustina-NarcisaAdvisors
Mitkov, R., Corpas, G., Inkpen, D.Issue Date
2012-10
Metadata
Show full item recordAbstract
In the eld of Descriptive Translation Studies, translationese refers to the speci c traits that characterise the language used in translations. While translationese has been often investigated to illustrate that translational language is di erent from non-translational language, scholars have also proposed a set of hypotheses which may characterise such di erences. In the quest for the validation of these hypotheses, embracing corpus-based techniques had a well-known impact in the domain, leading to several advances in the past twenty years. Despite extensive research, however, there are no universally recognised characteristics of translational language, nor universally recognised patterns likely to occur within translational language. This thesis addresses these issues, with a less used approach in the eld of Descriptive Translation Studies, by investigating the nature of translational language from a machine learning perspective. While the main focus is on analysing translationese, this thesis investigates two related sub-hypotheses: simpli cation and explicitation. To this end, a multilingual learning framework is designed and implemented for the identi cation of translational language. The framework is modelled as a categorisation task, the learning techniques having the major goal to automatically learn to distinguish between translated and non-translated texts. The second and third major goals of this research are the retrieval of the recurring patterns that are revealed in the process of solving the task of categorisation, as well as the ranking of the most in uential characteristics used to accomplish the learning task. These aims are ful lled by implementing a system that adopts the machine learning methodology proposed in this research. The learning framework proves to be an adaptable multilingual framework for the investigation of the nature of translational language, its adaptability being illustrated in this thesis by applying it to the investigation of two languages: Spanish and Romanian. In this thesis, di erent research scenarios and learning models are experimented with in order to assess to what extent translated texts can be di erentiated from non-translated texts in certain contexts. The ndings show that machine learning algorithms, aggregating a large set of potentially discriminative characteristics for translational language, are able to di erentiate translated texts from non-translated ones with high scores. The evaluation experiments report performance values such as accuracy, precision, recall, and F-measure on two datasets. The present research is situated at the con uence of three areas, more precisely: Descriptive Translation Studies, Machine Learning and Natural Language Processing, justifying the need to combine these elds for the investigation of translationese and translational hypotheses.Publisher
University of WolverhamptonType
Thesis or dissertationLanguage
enDescription
A thesis submitted in partial ful lment of the requirements of the University of Wolverhampton for the degree of Doctor of PhilosophyCollections
Related items
Showing items related by title, author, creator and subject.
-
Multiword units in machine translation and translation technologyRuslan, Mitkov; Monti, Johanna; Corpas Pastor, Gloria; Seretan, Violeta (John Benjamins, 2018-07-20)The correct interpretation of Multiword Units (MWUs) is crucial to many applications in Natural Language Processing but is a challenging and complex task. In recent years, the computational treatment of MWUs has received considerable attention but we believe that there is much more to be done before we can claim that NLP and Machine Translation (MT) systems process MWUs successfully. In this chapter, we present a survey of the field with particular reference to Machine Translation and Translation Technology.
-
Computational Phraseology light: automatic translation of multiword expressions without translation resourcesMitkov, Ruslan (de Gruyter Mouton, 2016-11-01)This paper describes the first phase of a project whose ultimate goal is the implementation of a practical tool to support the work of language learners and translators by automatically identifying multiword expressions (MWEs) and retrieving their translations for any pair of languages. The task of translating multiword expressions is viewed as a two-stage process. The first stage is the extraction of MWEs in each of the languages; the second stage is a matching procedure for the extracted MWEs in each language which proposes the translation equivalents. This project pursues the development of a knowledge-poor approach for any pair of languages which does not depend on translation resources such as dictionaries, translation memories or parallel corpora which can be time consuming to develop or difficult to acquire, being expensive or proprietary. In line with this philosophy, the methodology developed does not rely on any dictionaries or parallel corpora, nor does it use any (bilingual) grammars. The only information comes from comparable corpora, inexpensively compiled. The first proof-of-concept stage of this project covers English and Spanish and focuses on a particular subclass of MWEs: verb-noun expressions (collocations) such as take advantage, make sense, prestar atención and tener derecho. The choice of genre was determined by the fact that newswire is a widespread genre and available in different languages. An additional motivation was the fact that the methodology was developed as language independent with the objective of applying it to and testing it for different languages. The ACCURAT toolkit (Pinnis et al. 2012; Skadina et al. 2012; Su and Babych 2012a) was employed to compile automatically the comparable corpora and documents only above a specific threshold were considered for inclusion. More specifically, only pairs of English and Spanish documents with comparability score (cosine similarity) higher 0.45 were extracted. Statistical association measures were employed to quantify the strength of the relationship between two words and to propose that a combination of a verb and a noun above a specific threshold would be a (candidate for) multiword expression. This study focused on and compared four popular and established measures along with frequency: Log-likelihood ratio, T-Score, Log Dice and Salience. This project follows the distributional similarity premise which stipulates that translation equivalents share common words in their contexts and this applies also to multiword expressions. The Vector Space Model is traditionally used to represent words with their co-occurrences and to measure similarity. The vector representation for any word is constructed from the statistics of the occurrences of that word with other specific/context words in a corpus of texts. In this study, the word2vec method (Mikolov et al. 2013) was employed. Mikolov et al.’s method utilises patterns of word co-occurrences within a small window to predict similarities among words. Evaluation results are reported for both extracting MWEs and their automatic translation. A finding of the evaluation worth mentioning is that the size of the comparable corpora is more important for the performance of automatic translation of MWEs than the similarity between them as long as the comparable corpora used are of minimal similarity.
-
Comparing post-editing difficulty of different machine translation errors in Spanish and German translations from EnglishZaretskaya, A M; Vela, G; Seghiri, M; Corpas Pastor, Gloria (Center for Promoting Ideas, 2016-08-30)Post-editing (PE) of Machine Translation (MT) is an increasingly popular way to integrate MT in the professional translation workflow, as it increases productivity and income. However, the quality of MT is not always good enough to blindly choose PE over translation from scratch. This article studies the PE of different error types and compares indicators of PE difficulty in English-to-Spanish and English-to-German translations. The results show that the indicators in question 1) do not correlate between each other for all error types, and 2) differ between languages.