727 research outputs found
Sort by
Managerial bonuses and subordinate mistreatment [Dataset]
No full textCan performance bonuses increase the likelihood that managers coerce their subordinates into exerting high levels of effort when doing so promotes neither efficiency nor equity? We consider a laboratory setting in which managers compete to obtain a large bonus at the end of the experiment and the probability of obtaining it depends partly on the effort exerted by the subordinates. We find that managers are more than twice as likely to punish subordinates exerting “fair” levels of effort if they compete for individual bonuses than when there are no bonuses. This is not the case when they compete for pooled bonuses – giving managers discretionary power over how to share them with their subordinates – as most subordinates willingly exert maximal effort. A model in which individuals are assumed to care about equity captures subordinates’ behavior well. Managers’ willingness to coerce subordinates into exerting unfairly high levels of effort, on the other hand, cannot be readily accounted for by any of the existing models of social preferences
KGE Algorithms
No full text An updated method for link prediction that uses a regularization factor that models relation argument types
Abstract (Kotnis and Nastase, 2017):
Learning relations based on evidence from knowledge repositories
relies on processing the available relation instances. Knowledge
repositories are not balanced in terms of relations or entities – there
are relations with less than 10 but also thousands of instances, and
entities involved in less than 10 but also thousands of relations.
Many relations, however, have clear domain and range, which we
hypothesize could help learn a better, more generalizing, model.
We include such information in the RESCAL model in the form
of a regularization factor added to the loss function that takes into
account the types (categories) of the entities that appear as arguments
to relations in the knowledge base. Tested on Freebase, a frequently
used benchmarking dataset for link/path predicting tasks, we note
increased performance compared to the baseline model in terms of
mean reciprocal rank and hits@N, N = 1, 3, 10. Furthermore, we
discover scenarios that significantly impact the effectiveness of the
type regularizer
Lauchheim II.2. Katalog der Gräber 301–600
No full textMit rund 1300 Gräbern aus dem Zeitraum vom späten 5. bis zum späten 7. Jahrhundert ist das Gräberfeld von Lauchheim 'Wasserfurche' (Ostalbkreis) bis heute der größte bekannte merowingerzeitliche Bestattungsplatz Süddeutschlands. In den Jahren 1986 bis 1996 wurde das fast vollständig erhaltene Gräberfeld großflächig ausgegraben. Seit 2009 widmete sich ein multidisziplinäres Forschungsprojekt der Dokumentation und Auswertung der Lauchheimer Gräber und Grabbeigaben. Dabei kamen in großem Maßstab innovative Dokumentationsmethoden mit dem Ziel einer möglichst vollständigen Datenerfassung zum Einsatz. Als Ergebnis dieses Projekts werden die Befunde und Funde aus Lauchheim in einem detaillierten Katalog mit umfangreichem Tafelteil vorgelegt und damit der weiteren wissenschaftlichen Auswertung zugänglich gemacht. Der vorliegende zweite von insgesamt vier Teilbänden des Katalogs umfasst die Gräber 301–600. Als Datensupplement werden Abbildungen der Perlen aus den Gräbern bereitgestellt. Die Abbildungen liegen als separate tif-Dateien sowie zusammengefasst in einer pdf-Datei vor. Die Dateinamen der einzelnen Bilddateien setzen sich aus der Grabnummer, der Objektnummer und einer Bildnummer zusammen
The MSC Data Set
No full textFrom this page you can download resources we created for modal sense classification as reported in Zhou et al. (2015), Marasović et al. (2016) and Marasović and Frank (2015) (see "Related Publication" below):
Heuristically sense-annotated training data acquired from EUROPARL and OpenSubtitles (EPOS_E, English). The dataset was used for:
the EMNLP 2015 Workshop submission "Semantically enriched models for modal sense classification" by Mengfei Zhou, Anette Frank,Annemarie Friedrich, and Alexis Palmer
the LiLT submission "Modal Sense Classification At Large: Paraphrase-Driven Sense Projection, Semantically Enriched Classification Models and Cross-Genre Evaluations" by Ana Marasović, Mengfei Zou, Alexis Palmer, Anette Frank
the RepL4NLP submission "Multilingual Modal Sense Classification using a Convolutional Neural Network" by Ana Marasović and Anette Frank.
Composition of training and testing used for the classification experiments. The dataset was used for:
the EMNLP 2015 Workshop submission "Semantically enriched models for modal sense classification" by submission Mengfei Zhou, Anette Frank,Annemarie Friedrich, and Alexis Palmer
the RepL4NLP submission "Multilingual Modal Sense Classification using a Convolutional Neural Network" by Ana Marasović and Anette Frank.
Manually annotated subsection of MASC (English). The dataset was used for the LiLT submission "Modal Sense Classification At Large: Paraphrase-Driven Sense Projection, Semantically Enriched Classification Models and Cross-Genre Evaluations" by Ana Marasović, Mengfei Zou, Alexis Palmer, Anette Frank.
Heuristically modal sense annotated training data and manually annotated test data from EUROPARL and OpenSubtitles (EPOS_G, German). The dataset was used for the RepL4NLP submission "Multilingual Modal Sense Classification using a Convolutional Neural Network" by Ana Marasović and Anette Frank.
</p
Sentiment Compound Data (DE)
No full textThis dataset contains gold standards that are required for building a classifier that automatically extracts opinion (noun) compounds
Abstract Anaphora Resolution [Source Code]
No full textAbstract Anaphora Resolution (AAR) aims to find the interpretation of nominal expressions (e.g., this result, those two actions) and pronominal expressions (e.g., this, that, it) that refer to abstract-object-antecedents such as facts, events, plans, actions, or situations.
The folder Silver Data contains the code for processing the silver training data described in Marasović et al. (2017). For more information read Silver Data/README.
The folder Gold Data contains the code for processing the gold training and evaluation data. Use Gold Data/process_aar_data.py to prepare the ASN corpus (Kolhatkar et al, 2013) and the CoNLL-12 shared task data (Jauhar et al, 2015). Read arrau_csn/instructions_arrau_construction.txt for processing of the ARRAU corpus (Poesio et al, 2018).
The implementation for training and evaluating models presented in Marasović et al. (2017) maybe be found in the folder EMNLP 2017. The readme contains the information on how to run the training and evaluation scripts.
The implementation for training and evaluating models presented in the thesis may be found in the Thesis folder.
<p
Negative Sampling for Learning Knowledge Graph Embeddings
No full textReimplementation of four KG factorization methods and six negative sampling methods.
Abstract
Knowledge graphs are large, useful, but incomplete knowledge repositories. They encode knowledge through entities and relations which define each other through the connective structure of the graph. This has inspired methods for the joint embedding of entities and relations in continuous low-dimensional vector spaces, that can be used to induce new edges in the graph, i.e., link prediction in knowledge graphs. Learning these representations relies on contrasting positive instances with negative ones. Knowledge graphs include only positive relation instances, leaving the door open for a variety of methods for selecting negative examples. In this paper we present an empirical study on the impact of negative sampling on the learned embeddings, assessed through the task of link prediction. We use state-of-the-art knowledge graph embeddings -- \rescal , TransE, DistMult and ComplEX -- and evaluate on benchmark datasets -- FB15k and WN18. We compare well known methods for negative sampling and additionally propose embedding based sampling methods. We note a marked difference in the impact of these sampling methods on the two datasets, with the "traditional" corrupting positives method leading to best results on WN18, while embedding based methods benefiting the task on FB15k
Selectional Preference Embeddings (EMNLP 2017)
No full textJoint embeddings of selectional preferences, words, and fine-grained entity types.
The vocabulary consists of:
verbs and their dependency relation separated by "@", e.g. "sink@nsubj" or "elect@dobj"
words and short noun phrases, e.g. "Titanic"
fine-grained entity types using the FIGER inventory, e.g.: /product/ship or /person/politician
The files are in word2vec binary format, which can be loaded in Python with gensim like this:
from gensim.models import KeyedVectors
emb_file = "/path/to/embedding_file"
emb = KeyedVectors.load_word2vec_format(emb_file, binary=True)
</p
Zur absoluten Chronologie der Einzelgrabkultur in Norddeutschland und Nordjütland [Supplement]
No full textMit der Absicht ein typochronologisches Modell der geschweiften Becher der norddeutschen Einzelgrabkultur und damit eine detaillierte chronologische Differenzierung für die Diskussion historischer Prozesse zu entwickeln, erfolgt zunächst eine absolutchronologische Einordnung der Bestattungen der Einzelgrabkultur in Nordjütland (Dänemark), an die sich die Einbindung von radiometrischen Datierungen in typologische Analysen und eine Modellierung der 14C-Daten aus Bestattungen für den norddeutschen Raum anschließt. Das Ergebnis ist eine absolutchronologische Typentwicklung der geschweiften Becher in Norddeutschland zwischen 2900 und 2150 cal BC mit einer absolutchronologischen
Fixierung der Grabinventare der Einzelgrabkultur in Nordjütland zwischen 2750
und 2330 cal BC
Current knowledge on correlations between highly prevalent dental conditions and chronic diseases: an umbrella review [Dataset]
No full textIntroduction: There are existing studies investigating relationships between chronic systemic and dental conditions, but it remains unclear how such knowledge can be used in clinical practice. The present report provides an overview on existing systematic reviews, identifying and evaluating the most frequently reported dental-chronic disease correlations and common risk factors.
Methods: A systematic review on existing systematic reviews published between 1995 and 2017 was conducted (umbrella review) and indexed into four databases. The focus was set on the three most prevalent dental conditions and ten chronic systemic diseases with the highest burden of disease in Germany. Two independent reviewers assessed all articles for eligibility, methodological quality using the AMSTAR criteria, and extracted data from the included studies.
Results: 32 out of the initially identified 1,249 systematic reviews were included for qualitative synthesis. The dental conditions with most frequently observed correlations to chronic systemic diseases was periodontitis. The chronic systemic disease with most frequently observed correlations with a dental condition was diabetes mellitus type 2. Most dental-chronic disease correlations were found between periodontitis and diabetes mellitus type 2 and periodontitis and cardiovascular disease. Frequently reported common risk factors included smoking, age, sex and overweight. Using the AMSTAR criteria, 2 studies were assessed as low quality, 26 studies moderate and 4 studies high quality.
Conclusion: The quality of included systematic reviews was heterogenous. The most frequently reported correlations were found for periodontitis with diabetes mellitus type 2 and for cardiovascular disease with diabetes mellitus type 2. Yet the strength of evidence for these and other disease correlations remains limited. The evidence to assess the causality of these disease correlations remains unclear. Future research should therefore focus on the causality of disease links in order to provide more decisive evidence with respect to the design of intersectoral care processes