How comparable are parallel corpora? Measuring the distribution of general vocabulary and connectives

In this paper, we question the homogeneity of a large parallel corpus by measuring the similarity between various sub-parts. We compare results obtained using a general measure of lexical similarity based on χ2 and by counting the number of discourse connectives. We argue that discourse connectives provide a more sensitive measure, revealing differences that are not visible with the general measure. We also provide evidence for the existence of specific characteristics defining translated texts as opposed to non-translated ones, due to a universal tendency for explicitation.

Shallow Dialogue Processing Using Machine Learning Algorithms (or not)

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.

Towards Automatic Identification of Discourse Markers in Dialogs: the Case of 'Like'

This article discusses the detection of discourse markers (DM) in dialog transcriptions, by human annotators and by automated means. After a theoretical discussion of the definition of DMs and their relevance to natural language processing, we focus on the role of like as a DM. Results from experiments with human annotators show that detection of DMs is a difficult but reliable task, which requires prosodic information from soundtracks. Then, several types of features are defined for automatic disambiguation of like: collocations, part-of-speech tags and duration-based features. Decision-tree learning shows that for like, nearly 70% precision can be reached, with near 100% recall, mainly using collocation filters. Similar results hold for well, with about 91% precision at 100% recall.

Prüfung der Grundsätze, welche über die Peräquation der Kriegslasten bisher sind aufgestellt worden

von Friedrich Heinrich Hatzfeld

Pollen records and age determinations from 2 profiles at Lake Nikolay, Lena Delta

Radiocarbon-dated pollen, rhizopod, chironomid and total organic carbon (TOC) records from Nikolay Lake (73°20'N, 124°12'E) and a pollen record from a nearby peat sequence are used for a detailed environmental reconstruction of the Holocene in the Lena Delta area. Shrubby Alnus fruticosa and Betula exilis tundra existed during 10,300-4800 cal. yr BP and gradually disappeared after that time. Climate reconstructions based on the pollen and chironomid records suggest that the climate during ca. 10,300-9200 cal. yr BP was up to 2-3 °C warmer than the present day. Pollen-based reconstructions show that the climate was relatively warm during 9200-6000 cal. yr BP and rather unstable between ca. 5800-3700 cal. yr BP. Both the qualitative interpretation of pollen data and the results of quantitative reconstruction indicate that climate and vegetation became similar to modern-day conditions after ca. 3600 cal. yr BP. The chironomid-based temperature reconstruction suggests a relatively warm period between ca. 2300 and 1400 cal. yr BP, which corresponds to the slightly warmer climate conditions reconstructed from the pollen. Modern chironomid and rhizopod assemblages were established after ca. 1400 cal. yr BP.

Age determinations from Bykovsky Peninsula, Siberia

A wide variety of environmental records is necessary for analysing and understanding the complex Late Quaternary dynamics of permafrost-dominated Arctic landscapes. A NE Siberian periglacial key region was studied in detail using sediment records, remote sensing data, and terrain modelling, all incorporated in a geographical information system (GIS). The study area consists of the Bykovsky Peninsula and the adjacent Khorogor Valley in the Kharaulakh Ridge situated a few kilometres southeast of the Lena Delta. In this study a comprehensive cryolithological database containing information from 176 sites was compiled. The information from these sites is based on the review of previously published borehole data, outcrop profiles, surface samples, and our own field data. These archives cover depositional records of three periods: from Pliocene to Early Pleistocene, the Late Pleistocene and the Holocene. The main sediment sequences on the Bykovsky Peninsula consist of up to 50 m thick ice-rich permafrost deposits (Ice Complex) that were accumulated during the Late Pleistocene. They were formed as a result of nival processes around extensive snowfields in the Kharaulakh Ridge, slope processes in these mountains (such as in the Khorogor Valley), and alluvial/proluvial sedimentation in a flat accumulation plain dominated by polygonal tundra in the mountain foreland (Bykovsky Peninsula). During the early to middle Holocene warming, a general landscape transformation occurred from an extensive Late Pleistocene accumulation plain to a strongly thermokarst-dominated relief dissected by numerous depressions. Thermokarst subsidence had an enormous influence on the periglacial hydrological patterns, the sediment deposition, and on the composition and distribution of habitats. Climate deterioration, lake drainage, and talik refreezing occurred during the middle to late Holocene. The investigated region was reached by the post-glacial sea level rise during the middle Holocene, triggering thermo-abrasion of ice-rich coasts and the marine inundation of thermokarst depressions.

Reconstructed glaciation limits in Eurasia (digitized from map Fig.1)

Based on field investigations in northern Russia and interpretation of offshore seismic data, we have made a preliminary reconstruction of the maximum ice-sheet extent in the Barents and Kara Sea region during the Early/Middle Weichselian and the Late Weichselian. Our investigations indicate that the Barents and Kara ice sheets attained their maximum Weichselian positions in northern Russia prior to 50 000 yr BP, whereas the northeastern flank of the Scandinavian Ice Sheet advanced to a maximum position shortly after 17000 calendar years ago. During the Late Weichselian (25 000-10000 yr BP), much of the Russian Arctic remained ice-free. According to our reconstruction, the extent of the ice sheets in the Barents and Kara Sea region during the Late Weichselian glacial maximum was less than half that of the maximum model which, up to now, has been widely used as a boundary condition for testing and refining General Circulation Models (GCMs). Preliminary numerical-modelling experiments predict Late Weichselian ice sheets which are larger than the ice extent implied for the Kara Sea region from dated geological evidence, suggesting very low precipitation.

Pollen record with radiocarbon datings from Faddeyevskiy Island, Laptev Sea

Pollen, plant macrofossil, loss-on-ignition and radiocarbon analyses of a 1.4-m section in thermokarst topography from Faddeyevskiy Island (75°20'N, 143°50'E, 30 m elevation) provides new information on Late Pleistocene interstadial environmental history of this high Arctic region. Conventional radiocarbon dates (25,700 ± 1000, 32,780 ± 500, 35,200 ± 650 yr BP) and two AMS dates (29,950 ± 660 and 42,990 ± 1280 yr BP) indicate that the deposits accumulated during the Kargian (Boutellier) interval. Numerous mammoth (Mammuthus primigenius) remains that have been collected in vicinity of the site in this study were radio-carbon dated to 36,700-18,500 yr BP. Rare bison (Bison priscus) bones were dated to 32,200 ± 600 and 33,100 ± 320 yr BP. Poaceae, Cyperaceae, and Artemisia pollen dominate the spectra with some Ranunculaceae, Caryophyllaceae, Rosaceae, and Asteraceae. The pollen spectra reflect steppe-like (tundra-steppe) vegetation, which was dominant on the exposed shelf of the Arctic Ocean. Numerous Carex macrofossils suggest that the summer climate was at least 2°C warmer than today. The productivity of the local vegetation during the Kargian interstadial was high enough to feed the population of grazing mammals.

Holocene carbon cycle dynamics, links to model files

We are investigating the late Holocene rise in CO2 by performing four experiments with the climate-carbon-cycle model CLIMBER2-LPJ. Apart from the deep sea sediments, important carbon cycle processes considered are carbon uptake or release by the vegetation, carbon uptake by peatlands, and CO 2 release due to shallow water sedimentation of CaCO3. Ice core data of atmospheric CO2 between 8 ka BP and preindustrial climate can only be reproduced if CO2 outgassing due to shallow water sedimentation of CaCO3 is considered. In this case the model displays an increase of nearly 20 ppmv CO2 between 8 ka BP and present day. Model configurations that do not contain this forcing show a slight decrease in atmospheric CO2. We can therefore explain the late Holocene rise in CO2 by invoking natural forcing factors only, and anthropogenic forcing is not required to understand preindustrial CO2 dynamics.

Pollen record and age determination of a sediment profile from Lake Teletskoye, Russia

A high-resolution pollen record from Lake Teletskoye documents the climate-related vegetation history of the northern Altai Mountain region during the last millennium. Siberian pine taiga with Scots pine, fir, spruce, and birch dominated the vegetation between ca. AD 1050 and 1100. The climate was similar to modern. In the beginning of the 12th century, birch and shrub alder increased. Lowered pollen concentrations and simultaneous peaks in herbs (especially Artemisia and Poaceae), ferns, and charcoal fragments point to colder and more arid climate conditions than before, with frequent fire events. Around AD 1200, regional climate became warmer and more humid than present, as revealed by an increase of Siberian pine and decreases of dry herb taxa and charcoal contents. Climatic conditions were rather stable until ca. AD 1410. An increase of Artemisia pollen may reflect slightly drier climate conditions between AD 1410 and 1560. Increases in Alnus, Betula, Artemisia, and Chenopodiaceae pollen and in charcoal particle contents may reflect further deterioration of climate conditions between AD 1560 and 1810, consistent with the Little Ice Age. After AD 1850 the vegetation gradually approached the modern one, in conjunction with ongoing climate warming.