Grading assignments is a time-consuming part of teaching translation. Automatic tools that facilitate this task would allow teachers of professional translation to focus more on other aspects of their job. Within Natural Language Processing, error recognitionhas not been studied for human translation in particular. This thesis is a first attempt at both error recognition and classification with both mono- and bilingual models. BERT– a pre-trained monolingual language model – and NuQE – a model adapted from the field of Quality Estimation for Machine Translation – are trained on a relatively small hand annotated corpus of student translations. Due to the nature of the task, errors are quite rare in relation to correctly translated tokens in the corpus. To account for this,we train the models with both under- and oversampled data. While both models detect errors with moderate success, the NuQE model adapts very poorly to the classification setting. Overall, scores are quite low, which can be attributed to class imbalance and the small amount of training data, as well as some general concerns about the corpus annotations. However, we show that powerful monolingual language models can detect formal, lexical and translational errors with some success and that, depending on the model, simple under- and oversampling approaches can already help a great deal to avoid pure majority class prediction.

L’objectif des études théoriques et expérimentales présentées dans ce travail était de cerner à l’aide des critères objectifs fiables un niveau optimum de découpage textuel pour la traduction spécialisée assistée par un système de mémoire de traduction (SMT) pour les langues française et polonaise. Afin de réaliser cet objectif, nous avons élaboré notre propre approche : une nouvelle combinaison des méthodes de recherche et des outils d’analyse proposés surtout dans les travaux de Simard (2003), Langlais et Simard (2001, 2003) et Dragsted (2004) visant l’amélioration de la viabilité des SMT à travers des modifications apportées à la segmentation phrastique considérée comme limitant cette viabilité. A la base des observations de quelques réalisations effectives du processus de découpage textuel dans la traduction spécialisée effectuée par l’homme sans aide informatique à la traduction, nous avons déterminé trois niveaux de segmentation potentiellement applicables dans les SMT tels que phrase, proposition, groupes verbal et nominal. Nous avons ensuite réalisé une analyse comparative des taux de réutilisabilité des MT du système WORDFAST et de l’utilité des traductions proposées par le système pour chacun de ces trois niveaux de découpage textuel sur un corpus de douze textes de spécialité. Cette analyse a permis de constater qu’il n’est pas possible de déterminer un seul niveau de segmentation textuelle dont l’application améliorerait la viabilité des SMT de façon incontestable. Deux niveaux de segmentation textuelle, notamment en phrases et en propositions, permettent en effet d’assurer une viabilité comparable des SMT. / The aim of the theoretical and experimental study presented in this work was to define with objective and reliable criterion an optimal level of textual segmentation for specialized translation from French into Polish assisted by a translation memory system (TMS). In this aim, we created our own approach based on research methods and analysis tools proposed particularly by Simard (2003), Langlais and Simard (2001, 2003) and Dragsted (2004). In order to increase the SMT performances, they proposed to eliminate a sentence segmentation level from SMT which is considered an obstacle to achieve satisfying SMT performances. On the basis of the observations of text segmentation process realized during a specialized translation made by a group of students without any computer aid, we defined three segmentation levels which can be potentially used in SMT such as sentences, clauses and noun and verb phrases. We realized then a comparative study of the influence of each of these levels on the reusability of WORDFAST translation memories and on the utility of translations proposed by the system for a group of twelve specialized texts. This study showed that it is not possible to define a unique text segmentation level which would unquestionably increase the SMT performances. Sentences and clauses are in fact two text segmentation levels which ensure the comparable SMT performances.

Text in English / Abstracts in English and Venda / This study has investigated the effectiveness of available tools used for translating from English into Tshivenḓa and vice versa with the aim to investigate and determine the effectiveness of these tools. This study dealt with the problem of lack of effective translation tools used to translate between English and Tshivenḓa. Tshivenḓa is one of South Africa’s minority languages. Its (Tshivenḓa) lack of effective translation tools negatively affects language practitioners’ work. This situation is perilous for translation quality assurance. Translation tools, both computer technology and non-computer technology tools abound for developed languages such as English, French and others. Based on the results of this research project, the researcher did make recommendations that could remedy the situation. South Africa is a democratic country that has a number of language-related policies. This then creates a conducive context for stakeholders with language passion to fully develop Tshivenḓa language in all dimensions. The fact is that all languages have evolved and they were all underdeveloped. This vividly shows that Tshivenḓa language development is also possible just like Afrikaans, which never existed on earth before 1652. It (Afrikaans) has evolved and overtaken all indigenous South African languages. This study did review the literature regarding translation and translation tools. The literature was obtained from both published and unpublished sources. The study has used mixed methods research, i.e. quantitative and qualitative research methods. These methods successfully complemented each other throughout the entire research. Data were gathered through questionnaires and interviews wherein both open and closed-ended questions were employed. Both purposive/judgemental and snowball (chain) sampling have been applied in this study. Data analysis was addressed through a combination of methods owing to the nature of mixed methods research. Guided by analytic comparison approach when grouping together related data during data analysis and presentation, both statistical and textual analyses have been vital in this study. Themes were constructed to lucidly present the gathered data. At the last chapters, the researcher discussed the findings and evaluated the entire research before making recommendations and conclusion. / Iyi ṱhoḓisiso yo ita tsedzuluso nga ha kushumele kwa zwishumiswa zwi re hone zwine zwa shumiswa u pindulela u bva kha luambo lwa English u ya kha Tshivenḓa na u bva kha Tshivenḓa u ya kha English ndivho I ya u sedzulusa na u lavhelesa kushumele kwa izwi zwishumiswa uri zwi a thusa naa. Ino ṱhoḓisiso yo shumana na thaidzo ya ṱhahelelo ya zwishumiswa zwa u pindulela zwine zwa shumiswa musi hu tshi pindulelwa vhukati ha English na Tshivenḓa. Tshivenḓa ndi luṅwe lwa nyambo dza Afrika Tshipembe dzine dza ambiwa nga vhathu vha si vhanzhi. U shaea ha zwishumiswa zwa u pindulela zwine zwa shuma nga nḓila I thusaho zwi kwama mushumo wa vhashumi vha zwa nyambo nga nḓila I si yavhuḓi. Iyi nyimele I na mulingo u kwamaho khwaḽithi ya zwo pindulelwaho. Zwishumiswa zwa u pindulela, zwa thekhnoḽodzhi ya khomphiyutha na zwi sa shumisi thekhnoḽodzhi ya khomphiyutha zwo ḓalesa kha nyambo dzo bvelelaho u tou fana na kha English, French na dziṅwe. Zwo sendeka kha mvelelo dza ino thandela ya ṱhoḓisiso, muṱoḓisisi o ita themendelo dzine dza nga fhelisa thaidzo ya nyimele. Afrika Tshipembe ndi shango ḽa demokirasi ḽine ḽa vha na mbekanyamaitele dzo vhalaho nga ha dzinyambo. Izwi zwi ita uri hu vhe na nyimele ine vhafaramikovhe vhane vha funesa nyambo vha kone u bveledza Tshivenḓa kha masia oṱhe. Zwavhukuma ndi zwa uri nyambo dzoṱhe dzi na mathomo nahone dzoṱhe dzo vha dzi songo bvelela. Izwi zwi ita uri zwi vhe khagala uri luambo lwa Tshivenḓa na lwone lu nga bveledzwa u tou fana na luambo lwa Afrikaans lwe lwa vha lu si ho ḽifhasini phanḓa ha ṅwaha wa 1652. Ulu luambo (Afrikaans) lwo vha hone shangoni lwa mbo bveledzwa lwa fhira nyambo dzoṱhe dza fhano hayani Afrika Tshipembe. Kha ino ṱhoḓisiso ho vhaliwa maṅwalwa ane a amba nga ha u pindulela na nga ha zwishumiswa zwa u pindulela. Maṅwalwa e a vhalwa o wanala kha zwiko zwo kanḓiswaho na zwiko zwi songo kanḓiswaho. Ino ṱhoḓisiso yo shumisa ngona dza ṱhoḓisiso dzo ṱanganyiswaho, idzo ngona ndi khwanthithethivi na khwaḽithethivi. Idzi ngona dzo shumisana zwavhuḓisa kha ṱhoḓisiso yoṱhe. Data yo kuvhanganywa hu tshi khou shumiswa dzimbudziso na u tou vhudzisa hune afho ho shumiswa mbudziso dzo vuleaho na dzo valeaho. Ngona dza u nanga sambula muṱoḓisisi o shumisa khaṱulo yawe uri ndi nnyi ane a nga vha a na data yo teaho na u humbela vhavhudziswa uri vha bule vhaṅwe vhathu vha re na data yo teaho ino ṱhoḓisiso. viii Tsenguluso ya data ho ṱanganyiswa ngona dza u sengulusa zwo itiswa ngauri ṱhoḓisiso ino yo ṱanganyisa ngona dza u ita ṱhoḓisiso. Sumbanḓila ho shumiswa tsenguluso ya mbambedzo kha u sengulusa data. Data ine ya fana yo vhewa fhethu huthihi musi hu tshi khou senguluswa na u vhiga. Tsenguluso I shumisaho mbalo/tshivhalo (khwanthithethivi) na I shumisaho maipfi kha ino ngudo dzo shumiswa. Ho vhumbiwa dziṱhoho u itela u ṱana data ye ya kuvhanganywa. Ngei kha ndima dza u fhedza, muṱodisisi o rera nga ha mawanwa, o ṱhaṱhuvha ṱhoḓisiso yoṱhe phanḓa ha u ita themendelo na u vhina. / African Languages / M.A. (African Languages)