Martin Volk

Phillip Benjamin Ströbel, Simon Clematide, Martin Volk et al.

We apply the TrOCR framework to real-world, historical manuscripts and show that TrOCR per se is a strong model, ideal for transfer learning. TrOCR has been trained on English only, but it can adapt to other languages that use the Latin alphabet fairly easily and with little training material. We compare TrOCR against a SOTA HTR framework (Transkribus) and show that it can beat such systems. This finding is essential since Transkribus performs best when it has access to baseline information, which is not needed at all to fine-tune TrOCR.

Phillip Benjamin Ströbel, Simon Clematide, Martin Volk et al.

The evaluation of Handwritten Text Recognition (HTR) models during their development is straightforward: because HTR is a supervised problem, the usual data split into training, validation, and test data sets allows the evaluation of models in terms of accuracy or error rates. However, the evaluation process becomes tricky as soon as we switch from development to application. A compilation of a new (and forcibly smaller) ground truth (GT) from a sample of the data that we want to apply the model on and the subsequent evaluation of models thereon only provides hints about the quality of the recognised text, as do confidence scores (if available) the models return. Moreover, if we have several models at hand, we face a model selection problem since we want to obtain the best possible result during the application phase. This calls for GT-free metrics to select the best model, which is why we (re-)introduce and compare different metrics, from simple, lexicon-based to more elaborate ones using standard language models and masked language models (MLM). We show that MLM-based evaluation can compete with lexicon-based methods, with the advantage that large and multilingual transformers are readily available, thus making compiling lexical resources for other metrics superfluous.

Samuel Läubli, Chantal Amrhein, Patrick Düggelin et al.

Neural machine translation (NMT) has set new quality standards in automatic translation, yet its effect on post-editing productivity is still pending thorough investigation. We empirically test how the inclusion of NMT, in addition to domain-specific translation memories and termbases, impacts speed and quality in professional translation of financial texts. We find that even with language pairs that have received little attention in research settings and small amounts of in-domain data for system adaptation, NMT post-editing allows for substantial time savings and leads to equal or slightly better quality.

Samuel Läubli, Rico Sennrich, Martin Volk

Recent research suggests that neural machine translation achieves parity with professional human translation on the WMT Chinese--English news translation task. We empirically test this claim with alternative evaluation protocols, contrasting the evaluation of single sentences and entire documents. In a pairwise ranking experiment, human raters assessing adequacy and fluency show a stronger preference for human over machine translation when evaluating documents as compared to isolated sentences. Our findings emphasise the need to shift towards document-level evaluation as machine translation improves to the degree that errors which are hard or impossible to spot at the sentence-level become decisive in discriminating quality of different translation outputs.

Kurt Winkler, Tobias Kuhn, Martin Volk

The Swiss avalanche bulletin is produced twice a day in four languages. Due to the lack of time available for manual translation, a fully automated translation system is employed, based on a catalogue of predefined phrases and predetermined rules of how these phrases can be combined to produce sentences. The system is able to automatically translate such sentences from German into the target languages French, Italian and English without subsequent proofreading or correction. Our catalogue of phrases is limited to a small sublanguage. The reduction of daily translation costs is expected to offset the initial development costs within a few years. After being operational for two winter seasons, we assess here the quality of the produced texts based on an evaluation where participants rate real danger descriptions from both origins, the catalogue of phrases versus the manually written and translated texts. With a mean recognition rate of 55%, users can hardly distinguish between the two types of texts, and give similar ratings with respect to their language quality. Overall, the output from the catalogue system can be considered virtually equivalent to a text written by avalanche forecasters and then manually translated by professional translators. Furthermore, forecasters declared that all relevant situations were captured by the system with sufficient accuracy and within the limited time available.