727 research outputs found
Sort by
Methane production by three widespread marine phytoplankton species: release rates, precursor compounds, and potential relevance for the environment [Dataset]
No full textWe investigated marine algae as a source of methane (CH4) in oxic surface waters of oceans. The widespread haptophytes Emiliania huxleyi, Phaeocystis globosa and Chrysochromulina sp. were investigated. We performed CH4 production and stable carbon isotope measurements and provide unambiguous evidence that all three investigated marine algae are involved in the production of CH4 under oxic conditions. In cultures that were treated with 13C-labeled hydrogen carbonate, 13C values of CH4 increased with incubation time, resulting from the conversion of 13C–hydrogen carbonate to 13CH4. The addition of 13C-labeled dimethyl sulfide, dimethyl sulfoxide, and methionine sulfoxide – known algal metabolites that are ubiquitous in marine surface layers – resulted in the occurrence of 13C-enriched CH4 in cultures of E. huxleyi, clearly indicating that methylated sulfur compounds are also precursors of CH4. We propose that haptophyte mediated CH4 production could be a common and important process in marine surface waters
COREC – A neural multi-label COmmonsense RElation Classification system
No full textWe examine the learnability of Commonsense knowledge relations as represented in CONCEPTNET. We develop a neural open world multi-label classification system that focuses on the evaluation of classification accuracy for individual relations. Based on an in-depth study of the specific properties of the CONCEPTNET resource such as relation ambiguity or argument heterogeneity, we investigate the impact of different relation representations and model variations. Our analysis reveals that the complexity of argument types and relation ambiguity are the most important challenges to address. We design a customized evaluation method to address the incompleteness of the resource that can be expanded in future work.</p
LibriVoxDeEn - A Corpus for German-to-English Speech Translation and Speech Recognition
No full textThis dataset is a corpus of sentence-aligned triples of German audio, German text, and English translation, based on German audio books. The corpus consists of over 100 hours of audio material and over 50k parallel sentences. The speech data are low in disfluencies because of the audio book setup. The quality of audio and sentence alignments has been checked by a manual evaluation, showing that that speech alignment is in general very high. The sentence alignment quality is comparable to well-used parallel translation data and can be adjusted by cutoffs on the automatic alignment score. To our knowledge, this corpus is to date the largest resource for end-to-end speech translation for German
Erzählte Inschriften in der Literatur des Mittelalters (Projektdatenbank)
No full textDiese Datenpublikation entstammt dem Teilprojekt C05 (»Inschriftlichkeit. Reflexionen materialer Textkultur in der Literatur des 12. bis 17. Jahrhunderts«) des Sonderforschungsbereichs 933 (»Materiale Textkulturen«, Förderzeitraum: 2011–2023). Im Rahmen des Teilprojekts wurden erzählte Inschriften in der mittelalterlichen Literatur gesammelt und auf einer Website öffentlich zugänglich gemacht. Diese Datenpublikation enthält den Datensatz, auf dem diese Website beruht. Das Projekt ist seit Ende 2023 abgeschlossen.
Als Inschrift zählt Geschriebenes, das a) nicht zu den gängigen Formen von Schriftlichkeit gehört (also nicht mit üblichen Hilfsmitteln auf Pergament und Papier geschrieben ist) und/oder Geschriebenes, b) bei dem von einer intensiven Verbindung von Inhalt und Schriftträger auszugehen ist. Der Ausgangspunkt der Sammlung sind deutschsprachige erzählte Inschriften vor 1700. In der zweiten und dritten Förderperiode des SFB (2015–2023) wurden zudem verstärkt erzählte Inschriften in altnordischen, alt- und mittelenglischen, mittellateinischen und altfranzösischen Textstellen gesammelt.
Belege für die englische und skandinavische Literatur des Mittelalters recht vollständig sein. Ebenfalls recht vollständig abgedeckt ist die deutschsprachige höfische Literatur des 12.-14. Jahrhunderts. Der Bereich der geistlichen und historiographischen Literatur wurde nicht systematisch durchgesehen.
Gesammelt wurden die Daten vom Projektleiter (Ludger Lieb), von den Projektmitarbeiter*innen (Frank Krabbes, Astrid Lembke, Michael R. Ott, Laura Velte, Philipp Friedhofen und Dennis Disselhoff), den studentischen Hilfskräften des Teilprojekts sowie von den Mercator Fellows (Gastwissenschaftler*innen): Katja Schulz (Altnordisch), Christine Neufeld (Alt- und Mittelenglisch), Christiane Conrad von Heydendorff (Altfranzösisch) und Dennis Pulina (Mittellatein). </p
Age differences in emotion perception in a multiple target setting: An eye-tracking study [Dataset]
No full textResearch focusing on the association between age and emotion perception has revealed inconsistent findings, with some support for an age-related positivity effect, as predicted by socioemotional selectivity theory. We used the mood-of-the-crowd (MoC) task to investigate whether older adults judge a crowd consisting of happy and angry expressions to be dominated by happy faces more frequently. The task was to decide whether an array of faces included more angry or more happy faces. Accuracy, reaction times, and gaze movements were analyzed to test the hypothesis, derived from socioemotional selectivity theory, that age would be associated with a bias toward judging crowds as happy, and with longer and more numerous fixations on happy expressions. Seventy-six participants took part in the study representing three different age groups (young, middle-aged, old). Contrary to the hypothesis, older participants more often judged the emotional crowd to be angry compared to younger participants. Furthermore, whereas fixations were longer for happy faces than for angry faces in younger adults, this difference was not present in older adults. A decline in inhibitory processing in older adults as well as higher cognitive demands of the task are discussed as possible explanations for these findings
Twitter Titling Corpus
No full textThe Twitter Titling Corpus contains 4002 stance-annotated tweets collected between 20 June 2017 and 30 August 2017 mentioning 6 presidents. Each tweet is annotated for the naming form used to refer to the president, for the purpose of a study on the relation between naming variation and stance (cited below).
This data is to be used for research purposes only.
Columns
tweet_id: id of the tweet
president: person entity mentioned in the tweet who was president at the time of collection
country: country the president was president of at the time of collection
stance: positive, neutral or negative sentiment towards the president
naming form: form used to refer to president out of
first-name (FN)
last-name (LN)
first-name last-name (FNLN)
title last-name (TLN)
title first-name last-name (TFNLN)
</ul
Covariance resampling for particle filter -- state and parameter estimation for soil hydrology [dataset]
No full textThis repository contains the data for the paper
'Covariance resampling for particle filter -- state and parameter estimation for soil hydrology'. All plots contained in the paper can be reproduced using this data
Genre-sensitive Neural Situation Entity classifier (DE, EN)
No full textThis is a Classifier for situation entity types as described in Becker et al., 2017. These clause types depend on a combination of syntactic-semantic and contextual features. We explore this task in a deeplearning framework, where tuned word representations capture lexical, syntactic and semantic features. We introduce an attention mechanism that pinpoints relevant context not only for the current instance, but also for the larger context. The advantage of our neural model is that it avoids the need to reproduce linguistic features for other languages and is thus more easily transferable.
We provide code for the basic local model (GRU), the local model with attention (GRU+attention), and our best performing context model which uses labels of previous clauses and genre information (GRU+attention+label+genre).
The data we used for our experiments can be found here, and we used the same train-dev-test split: https://github.com/annefried/sitent/tree/master/annotated_corpus </p
Capitalizing on the (False) Consensus Effect: Two Tractable Methods to Elicit Private Information [Dataset]
No full textWe propose and experimentally test two tractable methods to incentivize the elicitation of private information: Benchmark and Coordination. Both mechanisms capitalize on the false consensus effect, a well-documented phenomenon that follows Bayesian reasoning. That is, individuals use their own type when predicting the type of others. Since it is not feasible to incentivize the elicitation of private information using facts, when these are not verifiable, we incentivize the respondent to reveal her perceptions about others and use that statement to predict the subject´s private information. The stronger the relationship between a subject´s type and her perception about the type of others, the more effective the mechanisms are in revealing the subject´s privately held information. In an experiment, we apply the mechanisms to reveal beliefs about probabilities. On the aggregate level, both mechanisms accurately reveal mean first-order beliefs of the population. On the subject level, the modal difference between probabilities elicited in either mechanism and actual first-order beliefs is zero. The results indicate that subjects strongly anchor their statements in Benchmark and Coordination on their private information
Norms in the Lab: Inexperienced versus Experienced Participants
No full textUsing coordination games, we study whether social norm perception differs between inexperienced and experienced participants in economic laboratory experiments. We find substantial differences between the two groups, both regarding injunctive and descriptive social norms in the context of participation in lab experiments. By contrast, social norm perception for the context of daily life does not differ between the two groups. We therefore conclude that learning through experience is more important than selection effects for understanding differences between the two groups. We also conduct exploratory analyses on the relation between lab and field norms and find that behaving unsocial in an experiment is considered substantially more appropriate than in daily life. This appears inconsistent with the hypothesis that social preferences measured in lab experiments are inflated and indicates a distinction between revealed social preferences as measured commonly and the elicitation of normatively appropriate behavior