• Publicationsclose
• Part of book or chapter of book close
• CH close
• English close
• Archive ouverte UNIGE close
• Hyper Article en Ligne close
• Hyper Article en Ligne - Sciences de l'Homme et de la Société close
• Digital Humanities and Cultural Heritage close

International audience; The High Level Trigger (HLT) of the ATLAS experiment at the Large Hadron Collider receives events which pass the LVL1 trigger at ~75 kHz and has to reduce the rate to ~200 Hz while retaining the most interesting physics. It is a software trigger and performs the reduction in two stages: the LVL2 trigger and the Event Filter (EF). At the heart of the HLT is the Steering software. To minimise processing time and data transfers it implements the novel event selection strategies of seeded, step-wise reconstruction and early rejection. The HLT is seeded by regions of interest identified at LVL1. These and the static configuration determine which algorithms are run to reconstruct event data and test the validity of trigger signatures. The decision to reject the event or continue is based on the valid signatures, taking into account pre-scale and pass-through. After the EF, event classification tags are assigned for streaming purposes. Several powerful new features for commissioning and operation have been added: comprehensive monitoring is now built in to the framework; for validation and debugging, reconstructed data can be written out; the steering is integrated with the new configuration (presented separately), and topological and global triggers have been added. This paper will present details of the final design and its implementation, the principles behind it, and the requirements and constraints it is subject to. The experience gained from technical runs with realistic trigger menus will be described.

Almost entirely overlooked throughout the 20th century, neglected by contemporary medical manuals, the clitoris has gradually returned centre stage thanks to Western feminism.

This chapter is concerned with the organization of categorical sequence data. We first build a typology of sequences distinguishing for example between chronological sequences and sequences without time content. This permits to identify the kind of information that the data organization should preserve. Focusing then mainly on chronological sequences, we discuss the advantages and limits of different ways of representing time stamped event and state sequence data and present solutions for automatically converting between various formats, e.g., between horizontal and vertical presentations but also from state sequences into event sequences and reciprocally. Special attention is also drawn to the handling of missing values in these conversion processes.

The most salient geological features observed during a recent reconnaissance trip to Crooked Island, SE Bahamas, include: (1) altered bioclastic calcarenites of probable Early Pleistocene age; and (2) an elevated intertidal notch carved in last interglacial deposits, indicating that sea level peaked at a higher elevation than previously estimated during that time period. Four main lithostratigraphic units were identified on Crooked Island: (1) highly weathered bioclastic calcarenites that yielded unreliable alloisoleucine/isoleucine (A/I) ratios, and two valid 87Sr/86Sr ratios averaging 0.709147; (2) well-lithified bioclastic/peloidal eolianites, forming low sea cliffs, that gave one A/I ratio of 0.523; (3) a complex and extensive unit including scarce coral framestone, exposed up to +1.2 m above sea level, and oolitic-peloidal calcarenites deposited in subtidal, beach, and eolian environments that yielded A/I ratios averaging 0.411 (n=5); and (4) poorly lithified bioclastic beach ridges congruent with modern sea level. Moreover, a prominent ridge along the north coast of the island shows, at +11 m above sea level, an intertidal notch carved in Unit 3 eolianite and filled by Unit 3 beach facies. Units 4, 3 and 2 can be compared, respectively, to the Rice Bay (Holocene), the Grotto Beach (Late Pleistocene) and the Owl's Hole (Middle Pleistocene) formations, previously identified on many other Bahamian islands. Of probable Early Pleistocene age (between 0.6 and 1 Ma), Unit 1 could represent the lowermost part of the Owl's Hole Formation and the top of the underlying, mostly marine Misery Point Formation recently discovered on Mayaguana. The unequivocal occurrence of an intertidal notch carved in, and sealed by, last-interglacial deposits at +11 m shows that the peak elevation reached by sea level during that time interval was much higher than previously assessed. Finally, stratigraphic units decrease in age from N to S, suggesting that the island grew differently than other Bahamian islands or, alternatively, that the northern margin of the Crooked-Acklins bank collapsed in a recent past.

Preliminary biostratigraphic and geochronological results obtained from a 44 m-long core drilled on the northern coast of Mayaguana Island (SE Bahamas) show that the topmost layers of the core date from the Burdigalian (Early Miocene), whereas the deepest units are of Chattian (Late Oligocene) or Aquitanian (Earliest Miocene) age. Accordingly, the platform aggraded 44 m of sediments in a 10 to 3 my time span, from the Chattian/Aquitanian to the Burdigalian, whereas previous surface investigations of the island showed that only 11 m of carbonates were accumulated in a 17 my-long period, between the Burdigalian and the Early Pleistocene. This new record shows that the accumulation rate of the Mayaguana Bank was much higher during the Late Paleogene/Early Miocene than during the time interval from the Middle Miocene to the Pleistocene. This decrease is likely due to vertical tectonic motions related to the late phases of the Cuban orogeny which reduced accommodation on the platform top. These results designate the Mayaguana Bank as an accurate gauge to record the elevation of sea-level highstands during the Neogene

When developing NLP tools for low-resource languages, one is often confronted with the lack of annotated data. We propose to circumvent this bottleneck by training a supervised HMM tagger on a closely related language for which annotated data are available, and translating the words in the tagger parameter files into the low-resource language. The translation dictionaries are created with unsupervised lexicon induction techniques that rely only on raw textual data. We obtain a tagging accuracy of up to 89.08% using a Spanish tagger adapted to Catalan, which is 30.66% above the performance of an unadapted Spanish tagger, and 8.88% below the performance of a supervised tagger trained on annotated Catalan data. Furthermore, we evaluate our model on several Romance, Germanic and Slavic languages and obtain tagging accuracies of up to 92%.

This paper presents a shallow dialogue analysis model, aimed at human-human dialogues in the context of staff or business meetings. Four components of the model are defined, and several machine learning techniques are used to extract features from dialogue transcripts: maximum entropy classifiers for dialogue acts, latent semantic analysis for topic segmentation, or decision tree classifiers for discourse markers. A rule-based approach is proposed for solving cross-modal references to meeting documents. The methods are trained and evaluated thanks to a common data set and annotation format. The integration of the components into an automated shallow dialogue parser opens the way to multimodal meeting processing and retrieval applications.

Recurrent neural networks (RNNs) have achieved impressive results in a variety of linguistic processing tasks, suggesting that they can induce non-trivial properties of language. We investigate here to what extent RNNs learn to track abstract hierarchical syntactic structure. We test whether RNNs trained with a generic language modeling objective in four languages (Italian, English, Hebrew, Russian) can predict long-distance number agreement in various constructions. We include in our evaluation nonsensical sentences where RNNs cannot rely on semantic or lexical cues ("The colorless green ideas I ate with the chair sleep furiously"), and, for Italian, we compare model performance to human intuitions. Our language-model-trained RNNs make reliable predictions about long-distance agreement, and do not lag much behind human performance. We thus bring support to the hypothesis that RNNs are not just shallow-pattern extractors, but they also acquire deeper grammatical competence. Accepted to NAACL 2018

International audience; This essay examines the geographic origins, the political belongings and the confessional profiles of the members of the two most important mercantile nations of eighteenth-century southern Italy: the French nation and the British factory of Naples. Taking into account their pronounced prosopographic heterogeneity, it shows how legal resources and their social uses played a crucial role in defining the boundaries of such groups.

We present the results of the VarDial Evaluation Campaign on Natural Language Processing (NLP) for Similar Languages, Varieties and Dialects, which we organized as part of the fourth edition of the VarDial workshop at EACL’2017. This year, we included four shared tasks: Discriminating between Similar Languages (DSL), Arabic Dialect Identification (ADI), German Dialect Identification (GDI), and Cross-lingual Dependency Parsing (CLP). A total of 19 teams submitted runs across the four tasks, and 15 of them wrote system description papers.