• Digital Humanities and Cultural Heritage
• Publications close
• Research data close
• Other research products close
• Conference object close
• Report close
• CH close
• OpenAIRE close

For machine translation to tackle discourse phenomena, models must have access to extra-sentential linguistic context. There has been recent interest in modelling context in neural machine translation (NMT), but models have been principally evaluated with standard automatic metrics, poorly adapted to evaluating discourse phenomena. In this article, we present hand-crafted, discourse test sets, designed to test the models' ability to exploit previous source and target sentences. We investigate the performance of recently proposed multi-encoder NMT models trained on subtitles for English to French. We also explore a novel way of exploiting context from the previous sentence. Despite gains using BLEU, multi-encoder models give limited improvement in the handling of discourse phenomena: 50% accuracy on our coreference test set and 53.5% for coherence/cohesion (compared to a non-contextual baseline of 50%). A simple strategy of decoding the concatenation of the previous and current sentence leads to good performance, and our novel strategy of multi-encoding and decoding of two sentences leads to the best performance (72.5% for coreference and 57% for coherence/cohesion), highlighting the importance of target-side context. Comment: Final version of paper to appear in Proceedings of NAACL 2018

We present the first prototype of the SUMMA Platform: an integrated platform for multilingual media monitoring. The platform contains a rich suite of low-level and high-level natural language processing technologies: automatic speech recognition of broadcast media, machine translation, automated tagging and classification of named entities, semantic parsing to detect relationships between entities, and automatic construction / augmentation of factual knowledge bases. Implemented on the Docker platform, it can easily be deployed, customised, and scaled to large volumes of incoming media streams.

We propose a method to decide whether two occurrences of the same noun in a source text should be translated consistently, i.e. using the same noun in the target text as well. We train and test classifiers that predict consistent translations based on lexical, syntactic, and semantic features. We first evaluate the accuracy of our classifiers intrinsically, in terms of the accuracy of consistency predictions, over a subset of the UN Corpus. Then, we also evaluate them in combination with phrase-based statistical MT systems for Chinese-to-English and German-to-English. We compare the automatic post-editing of noun translations with the re-ranking of the translation hypotheses based on the classifiers’ output, and also use these methods in combination. This improves over the baseline and closes up to 50% of the gap in BLEU scores between the baseline and an oracle classifier.

The Idiap NLP Group has participated in both DiscoMT 2015 sub-tasks: pronoun-focused translation and pronoun prediction. The system for the first sub-task combines two knowledge sources: gram matical constraints from the hypothesized coreference links, and candidate translations from an SMT decoder. The system for the second sub-task avoids hypothesizing a coreference link, and uses instead a large set of source-side and target-side features from the noun phrases surrounding the pronoun to train a pronoun predictor.

Of the three most widely used implementations of the WLCG Storage Element specification, Disk Pool Manager[1, 2] (DPM) has the simplest implementation of file placement balancing (StoRM doesn't attempt this, leaving it up to the underlying filesystem, which can be very sophisticated in itself). DPM uses a round-robin algorithm (with optional filesystem weighting), for placing files across filesystems and servers. This does a reasonable job of evenly distributing files across the storage array provided to it. However, it does not offer any guarantees of the evenness of distribution of that subset of files associated with a given "dataset" (which often maps onto a "directory" in the DPM namespace (DPNS)). It is useful to consider a concept of "balance", where an optimally balanced set of files indicates that the files are distributed evenly across all of the pool nodes. The best case performance of the round robin algorithm is to maintain balance, it has no mechanism to improve balance.<p></p>\ud In the past year or more, larger DPM sites have noticed load spikes on individual disk servers, and suspected that these were exacerbated by excesses of files from popular datasets on those servers. We present here a software tool which analyses file distribution for all datasets in a DPM SE, providing a measure of the poorness of file location in this context. Further, the tool provides a list of file movement actions which will improve dataset-level file distribution, and can action those file movements itself. We present results of such an analysis on the UKI-SCOTGRID-GLASGOW Production DPM.

12 p.; International audience; The DIRAC system was developed in order to provide a complete solution for using distributed computing resources of the LHCb experiment at CERN for data production and analysis. It allows a concurrent use of over 10K CPUs and 10M file replicas distributed over many tens of sites. The sites can be part of a Computing Grid such as WLCG or standalone computing clusters all integrated in a single management structure. DIRAC is a generic system with the LHCb specific functionality incorporated through a number of plug-in modules. It can be easily adapted to the needs of other communities. A special attention is paid to the resilience of the DIRAC components to allow an efficient use of non-reliable resources. The DIRAC production management components provide a framework for building highly automated data production systems including data distribution and data driven workload scheduling. In this paper we give an overview of the DIRAC system architecture and design choices. We show how different components are put together to compose an integrated data processing system including all the aspects of the LHCb experiment - from the MC production and raw data reconstruction to the final user analysis.