• Publicationsclose
• Research dataclose
• Part of book or chapter of book close
• CH close
• Archive ouverte UNIGE close
• Hyper Article en Ligne close
• Digital Humanities and Cultural Heritage close

Ever-increasing amount of multimedia available online necessitates the development of new techniques and methods that can overcome the semantic gap problem. The said problem, encountered due to major disparities between inherent representational characteristics of multimedia and its semantic content sought by the user, has been a prominent research direction addressed by a great number of semantic augmentation approaches originating from such areas as machine learning, statistics, natural language processing, etc. In this paper, we review several of these recently developed techniques that bring together low-level representation of multimedia and its semantics in order to improve the efficiency of access and retrieval. We also present a distance-based discriminant analysis (DDA) method that defines the design of a basic building block classifier for distinguishing among a selected number of semantic categories. In addition to that, we demonstrate how a set of DDA classifiers can be grouped into a hierarchical ensemble for prediction of an arbitrary set of semantic classes.

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

The aims of this chapter are to outline the key experimental methods used by neuroscientists to evaluate the impact of game-based training strategies on attention and working memory function in humans. The chapter is organized into sections that detail methods that probe the major facets of attention followed by methods that probe working memory. Attention methods are subdivided into sections on (1) spatial attention, (2) temporal attention, (3) combined spatial and temporal attention, (4) feature-/object-based attention, (5) sustained attention, (6) interaction between top-down and bottom-up attention, and (7) attention resource allocation. Methods that assess working memory function are then presented in two major sections, (9) visual and visuospatial working memory and (10) verbal working memory. We conclude with a discussion of prospects for further research and applications.

Cet article propose une réflexion sur la manière dont la médecine fabrique les ‘grands hommes' de son panthéon, et sur la façon dont elle représente le progrès scientifique à travers eux. L'exemple retenu est celui de Théophile de Bordeu (1722-1776), adulé comme un fondateur de la médecine moderne par ses contemporains, avant de sombrer dans l'oubli. Son cas montre qu'une icône du progrès en dit moins sur elle-même que sur l'époque qui la fabrique. Le corpus étudié réunit différents genres de la consécration scientifique (épîtres encomiastiques, éloges académiques, notices historiques, avant-propos aux rééditions de ses œuvres, entrées de dictionnaires, etc.).

BabelDr (http://babeldr.unige.ch/) is a joint project of Geneva's Faculty of Translation and Interpretation (FTI) and University Hospitals (HUG), that has been active since July 2015. The goal is to develop methods that allow rapid prototyping of medium-vocabulary web-enabled medical speech translators, with particular emphasis on languages spoken by victims of the current European refugee crisis. A demonstrator system freely available on the project site translates spoken French medical examination questions into four languages.

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.

Fossil lacustrine sediments were found at about 15 m above sea level, directly above beach deposits dating from Marine Isotope Stage (MIS) 11 on a tectonically stable coastline in Eleuthera (Bahamas). This new discovery brings further support to the hypothesis suggesting that a major collapse of polar ice sheets occurred during this interglacial period. Two stratigraphic sections (GW1/GW2), including several carbonate units separated by paleosols, were logged near Glass Window (northern Eleuthera). Both sections comprise a basal unit made of cross-bedded, bio-peloidal limestone, including an early generation of fibrous rim cement of marine origin and a late phase of drusy calcite mosaic. A second unit displays well-defined planar bedding at both sites, whereas faint trough cross-stratifications are also visible at GW1. It consists of a bio-peloidal grainstone containing Halimeda fragments and lithoclasts derived from the underlying unit, and is further characterized by an early generation of isopachous fibrous cement. The upper part of this second unit shows one thick laminated crust with desiccation polygons and numerous spherulites of cyanobacterial origin, and is capped by a paleosol. The top of the GW2 section is represented by a karstified oolitic/peloidal grainstone, not found at GW 1, whereas the upper part of GW1 includes one bioclastic and one oolitic limestone unit separated by a paleosol. Based on sedimentary structures, petrographic composition, and amino-acid ratios borrowed from previous authors, the basal and upper units exposed at both sections can be identified as eolianites dating from MIS 11, 9, 7 and 5e, respectively. Occurring between +12 and +14.5 m, the basal part of the second unit can be interpreted as beach deposit, whereas its upper part, at +15 m, likely corresponds to lacustrine or pond sediments, both dating from MIS 11. Considering the subsidence rate affecting Eleuthera, these beach and pond deposits, and the phreatic cements in the basal eolian units unavoidably imply a sea-level highstand at ca. 20 m during MIS 11. This exceptional event was probably caused by the collapse and melting of polar ice-sheets during this very warm climatic period. In the context of present-day global warming, a similar scenario could be repeated in the near future.

This paper offers an overview of the history of medial NP-adjuncts from Old English to Present-Day English. In Present-Day English, adverbs are perfectly grammatical in a position between the subject and the main verb ('He recently left for London') whereas NP-adjuncts are at best stylistically marked in this position ('(*)He tomorrow leaves for London'). The paper shows that while medial placement of NP-adjuncts has been considerably less frequent as compared to adverbs ever since around 1500, the contrast was initially much stronger in clauses with finite main verbs than in clauses with finite auxiliaries. It is only in the 19th century that medial placement becomes equally marked in both contexts. These developments are accounted for in terms of processing constraints disfavouring the use of medial NP-adjuncts and a structural reanalysis of NP-medial adjuncts in Late Modern English.

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.

Listeners can understand novel lexical forms without apparent difficulty. This ability to analyze and interpret an unfamiliar input string raises some important psycholinguistic questions. We are led to ask how this parsing is actually accomplished and what its role is in the recognition of familiar word forms. The Standard psycholinguistic answer to the latter question has been that the human parsing abilities at the lexical level are of only minor importance. Indeed, in modeling lexical processing, psycholinguists have not been particularly concerned with morphological productivity and its implications for lexical processing and storage. This neglect of productivity is clearly apparent in the default view of language comprehension which is assumed to be based upon two radically different processing mechanisms. The first is exploited during word recognition and involves retrieving information from a permanent memory store, the lexicon. The second mechanism allows the integration of the semantic and syntactic information associated with the individually recognized words and their order. These latter processes parse and construct novel sentential representations. Thus, the mechanisms that are capable of generating new linguistic structure are typically reserved for the post-lexical processes.