No description available.

The notion that consistency is important when translating technical terms is nothing new in itself. Despite this, it is not uncommon to see the very same technical terms being translated into a multitude of words, even when various efforts to unify the process have been proposed. This diversity in the translation of technical terms has previously been shown to occur in the medical field by analyzing equipment manuals. However, there is a lack of research on how technical term translation fares in other fields. To narrow this identified gap, quantitative overviews of the translations from the widely respected operating system Debian are presented. The first overview is a list of the 354 most frequently appearing English technical terms. This was created by looking at the existing textual data of the system quantitatively, together with a judgment-based term selection process with the help of domain experts. The second overview is an analysis of how the 10 most frequently used identified technical terms are translated into Japanese. This is done by Catford’s probabilistic notion of textual equivalence, which shows the different translation alternatives found in the data and the probability of their appearance. The results of this study indicate that there is a varying degree of unification within the translation of technical terms. The synthesized data of the study can be used by the translation community to deepen educated decisions on the translation of technical terms within the field of information technology.

No description available.

Low-resource languages, including sign languages, are a challenge for machine translation research. Given the lack of parallel corpora, current researchers must be content with a small parallel corpus in a narrow domain for training a system. For this thesis, we obtained a small parallel corpus of English text and American Sign Language gloss from The Church of Jesus Christ of Latter-day Saints. We cleaned the corpus by loading it into an open-source translation memory tool, where we removed computer markup language and split the large chunks of text into sentences and phrases, creating a total of 14,247 sentence pairs. We randomly partitioned the corpus into three sections: 70% for a training set, 10% for a development set, and 20% for a test set. After downloading and installing the open-source Moses toolkit, we went through several iterations of training, translating, and evaluating the system. The final evaluation on unseen data yielded a state-of-the-art score for a low-resource language.

Craig Santos Perez's poetic series from Unincorporated Territory describes and decries the U.S. militarization, colonization, and environmental degradation of Guam in the Western Pacific through multilingual, excerpted, and series-long poems. Perez's writing style requires slow, careful reading with translations sometimes appearing on the same page, various pages later, or not at all. I describe this kind of elongated translation as slow translation, recalling translation theorist Michael Cronin's "Slow Language" movement. This thesis invites readers, especially multispecies ethnographers, to slow down the translation of nonhuman species and their stories by paying attention to moments of untranslatability in multispecies literature and interactions. In modeling how to think-with untranslatability, I call upon translation scholars Barbara Cassin and Cronin, who describe untranslatability in temporal and agentic terms, and environmental humanist Donna Haraway, whose tentacular thinking model and multispecies approaches have slowed our tendencies towards linear and assumptive modes of thinking. In conjunction with these thinkers, my multispecies reading of from Unincorporated Territory proposes slow translation as a model for resisting easy or colonizing translations that homogenize the Other. Perez's multilingual, fractured poems create moments of untranslatability, especially when describing nonhuman species or environments, that are difficult to immediately understand due to nontranslations or delayed translations. This thesis pays special attention to such moments as opportunities for slowing down and staying with difference. Thus, moments of untranslatability offer an ethnographic and interactive mode for engaging with difference through slow translation, valuing the process and experience of translation, the agency of the subjects in translation, and the incomprehensibility or unknown nature of the nonhuman and Othered world.

Translation Memory (TM) is a valuable tool that helps human translators in doing their job. TM consists of a collection of previously translated texts, called translation units, that may prove useful in the translation of new text. The main problem faced by translators who wish to take advantage of TM is that, although search tools do exist, there is no standardized way of effectively ranking search results. This thesis proposes a method for ranking TM search results together with a novel approach to efficiently finding common substrings that is used in the ranking process.

Ribosomes translate the genetic information encoded by mRNA into proteins. Defective mRNAs can cause stalling of translating ribosomes. The molecule tmRNA (transfer-messenger RNA) rescues stalled ribosomes in eubacteria. Together with its protein partner SmpB, tmRNA mimics a tRNA by entering the ribosomal A site and linking an alanine residue to the growing polypeptide chain. The ribosome then abandons the defective mRNA template and resumes translation on tmRNA, adding ten more amino acids to the nascent polypeptide. As a result of tmRNA action, stalled ribosomes are released and recycled, the defective mRNA is destroyed, and the aborted protein product is tagged for destruction by proteases. It is unknown how the ribosome correctly chooses the position on tmRNA to resume translation. Previous studies implicate the sequence UAGUC found immediately upstream of the first codon in the tmRNA open reading frame. These nucleotides are highly conserved in natural tmRNA sequences. Mutations in this area cause loss of tmRNA function and improper frame choice. Using a genetic selection that ties the life of E. coli cells to the function of tmRNA, we have identified several SmpB mutants that rescue an inactive tmRNA in which this upstream sequence was altered. This links SmpB to the function of these key tmRNA nucleotides. We show that our SmpB mutants affect frame choice using an in vivo assay for tagging in the various frames. We conclude that SmpB plays a role in setting the reading frame on tmRNA.

Although the ribosome is a very general catalyst, it cannot synthesize all protein sequences equally well. Certain proteins are capable of stalling the ribosome during their own synthesis. Stalling events are used by both prokaryotic and eukaryotic cells to regulate gene expression. Characterization of natural stalling peptides shows that only a few strategically placed amino acids are needed to inactivate the ribosome. These motifs share little sequence similarity suggesting that there are more stalling motifs yet to be discovered. Here we use two genetic selections in E. coli to discover novel stalling peptides and detail their subsequent characterization. Kinetic studies show that some of these nascent peptides dramatically inhibit rates of peptide release by release factors. We find that residues upstream of the minimal stalling motif can either enhance or suppress this effect. In other stalling motifs, such as polyproline sequences, peptidyl transfer to a subset of aminoacyl-tRNAs is inhibited. Translation factor EF-P alleviates pausing of the polyproline motifs, but has little or no effect on other stalling sequences. The EF-P ortholog eIF5A also alleviates pausing of polyproline sequences in yeast. Our studies show that short peptides sequences are capable of stalling the ribosome during elongation and termination through different mechanisms. These sequences are underrepresented in bacterial proteomes and show evidence of stalling on endogenous E. coli proteins.

Traditionally, Machine Translation (MT) systems are developed by targeting fluency (i.e. output grammaticality) and adequacy (i.e. semantic equivalence with the source text) criteria that reflect the needs of human end-users. However, recent advancements in Natural Language Processing (NLP) and the introduction of NLP tools in commercial services have opened new opportunities for MT. A particularly relevant one is related to the application of NLP technologies in low-resource language settings, for which the paucity of training data reduces the possibility to train reliable services. In this specific condition, MT can come into play by enabling the so-called “translation-based” workarounds. The idea is simple: first, input texts in the low-resource language are translated into a resource-rich target language; then, the machine-translated text is processed by well-trained NLP tools in the target language; finally, the output of these downstream components is projected back to the source language. This results in a new scenario, in which the end-user of MT technology is no longer a human but another machine. We hypothesize that current MT training approaches are not the optimal ones for this setting, in which the objective is to maximize the performance of a downstream tool fed with machine-translated text rather than human comprehension. Under this hypothesis, this thesis introduces a new research paradigm, which we named “MT for machines”, addressing a number of questions that raise from this novel view of the MT problem. Are there different quality criteria for humans and machines? What makes a good translation from the machine standpoint? What are the trade-offs between the two notions of quality? How to pursue machine-oriented objectives? How to serve different downstream components with a single MT system? How to exploit knowledge transfer to operate in different language settings with a single MT system? Elaborating on these questions, this thesis: i) introduces a novel and challenging MT paradigm, ii) proposes an effective method based on Reinforcement Learning analysing its possible variants, iii) extends the proposed method to multitask and multilingual settings so as to serve different downstream applications and languages with a single MT system, iv) studies the trade-off between machine-oriented and human-oriented criteria, and v) discusses the successful application of the approach in two real-world scenarios.

Übersetzung als Phänomen hat wahrscheinlich existiert seit Menschen verschiedene Sprachen gesprochen haben. Es ist bekannt, dass sie zumindest seit den Römern existiert, weil sie regelmäßig griechische Werke übersetzten (Munday 2012:19). Trotz ihrer langen Geschichte ist Übersetzungswissenschaft ein neues Fach der Linguistik und es gibt mehr oder weniger keine Regeln oder Normen in Bezug darauf, wie man klar und korrekt übersetzen soll. Es gibt in gewissen Bereichen der Übersetzungswissenschaft einige Regeln und Normen, aber der Übersetzer hat allgemein ziemlich freie Hand. Infolgedessen werden verschiedene Aspekte eines Texts stärker oder schwächer betont, was unten weiter beschreiben wird.