@phdthesis{ChujfiLaRoche2020,
  author    = {Chujfi-La-Roche, Salim},
  title     = {Human Cognition and natural Language Processing in the Digitally Mediated Environment},
  school      = {Universit{\"a}t Potsdam},
  pages     = {148},
  year      = {2020},
  abstract  = {Organizations continue to assemble and rely upon teams of remote workers as an essential element of their business strategy; however, knowledge processing is particular difficult in such isolated, largely digitally mediated settings. The great challenge for a knowledge-based organization lies not in how individuals should interact using technology but in how to achieve effective cooperation and knowledge exchange. Currently more attention has been paid to technology and the difficulties machines have processing natural language and less to studies of the human aspect—the influence of our own individual cognitive abilities and preferences on the processing of information when interacting online. This thesis draws on four scientific domains involved in the process of interpreting and processing massive, unstructured data—knowledge management, linguistics, cognitive science, and artificial intelligence—to build a model that offers a reliable way to address the ambiguous nature of language and improve workers' digitally mediated interactions. Human communication can be discouragingly imprecise and is characterized by a strong linguistic ambiguity; this represents an enormous challenge for the computer analysis of natural language. In this thesis, I propose and develop a new data interpretation layer for the processing of natural language based on the human cognitive preferences of the conversants themselves. Such a semantic analysis merges information derived both from the content and from the associated social and individual contexts, as well as the social dynamics that emerge online. At the same time, assessment taxonomies are used to analyze online comportment at the individual and community level in order to successfully identify characteristics leading to greater effectiveness of communication. Measurement patterns for identifying effective methods of individual interaction with regard to individual cognitive and learning preferences are also evaluated; a novel Cyber-Cognitive Identity (CCI)—a perceptual profile of an individual's cognitive and learning styles—is proposed. Accommodation of such cognitive preferences can greatly facilitate knowledge management in the geographically dispersed and collaborative digital environment. Use of the CCI is proposed for cognitively labeled Latent Dirichlet Allocation (CLLDA), a novel method for automatically labeling and clustering knowledge that does not rely solely on probabilistic methods, but rather on a fusion of machine learning algorithms and the cognitive identities of the associated individuals interacting in a digitally mediated environment. Advantages include: a greater perspicuity of dynamic and meaningful cognitive rules leading to greater tagging accuracy and a higher content portability at the sentence, document, and corpus level with respect to digital communication.},
  language  = {en}
}
@article{KrestelChikkamathHeweletal.2021,
  author    = {Krestel, Ralf and Chikkamath, Renukswamy and Hewel, Christoph and Risch, Julian},
  title     = {A survey on deep learning for patent analysis},
  series = {World patent information},
  volume    = {65},
  journal   = {World patent information},
  publisher = {Elsevier},
  address   = {Amsterdam},
  issn      = {0172-2190},
  doi       = {10.1016/j.wpi.2021.102035},
  pages     = {13},
  year      = {2021},
  abstract  = {Patent document collections are an immense source of knowledge for research and innovation communities worldwide. The rapid growth of the number of patent documents poses an enormous challenge for retrieving and analyzing information from this source in an effective manner. Based on deep learning methods for natural language processing, novel approaches have been developed in the field of patent analysis. The goal of these approaches is to reduce costs by automating tasks that previously only domain experts could solve. In this article, we provide a comprehensive survey of the application of deep learning for patent analysis. We summarize the state-of-the-art techniques and describe how they are applied to various tasks in the patent domain. In a detailed discussion, we categorize 40 papers based on the dataset, the representation, and the deep learning architecture that were used, as well as the patent analysis task that was targeted. With our survey, we aim to foster future research at the intersection of patent analysis and deep learning and we conclude by listing promising paths for future work.},
  language  = {en}
}
@article{MadrugadeBritoOttoKuhlicke2021,
  author    = {Madruga de Brito, Mariana and Otto, Danny and Kuhlicke, Christian},
  title     = {Tracking topics and frames regarding sustainability transformations during the onset of the COVID-19 crisis},
  series = {Sustainability / Multidisciplinary Digital Publishing Institute (MDPI)},
  volume    = {13},
  journal   = {Sustainability / Multidisciplinary Digital Publishing Institute (MDPI)},
  number    = {19},
  publisher = {MDPI},
  address   = {Basel},
  issn      = {2071-1050},
  doi       = {10.3390/su131911095},
  pages     = {19},
  year      = {2021},
  abstract  = {Many researchers and politicians believe that the COVID-19 crisis may have opened a "window of opportunity " to spur sustainability transformations. Still, evidence for such a dynamic is currently lacking. Here, we propose the linkage of "big data " and "thick data " methods for monitoring debates on transformation processes by following the COVID-19 discourse on ecological sustainability in Germany. We analysed variations in the topics discussed by applying text mining techniques to a corpus with 84,500 newspaper articles published during the first COVID-19 wave. This allowed us to attain a unique and previously inaccessible "bird's eye view " of how these topics evolved. To deepen our understanding of prominent frames, a qualitative content analysis was undertaken. Furthermore, we investigated public awareness by analysing online search behaviour. The findings show an underrepresentation of sustainability topics in the German news during the early stages of the crisis. Similarly, public awareness regarding climate change was found to be reduced. Nevertheless, by examining the newspaper data in detail, we found that the pandemic is often seen as a chance for sustainability transformations-but not without a set of challenges. Our mixed-methods approach enabled us to bridge knowledge gaps between qualitative and quantitative research by "thickening " and providing context to data-driven analyses. By monitoring whether or not the current crisis is seen as a chance for sustainability transformations, we provide insights for environmental policy in times of crisis.},
  language  = {en}
}
@article{RischKrestel2020,
  author    = {Risch, Julian and Krestel, Ralf},
  title     = {Toxic comment detection in online discussions},
  series = {Deep learning-based approaches for sentiment analysis},
  journal   = {Deep learning-based approaches for sentiment analysis},
  editor    = {Agarwal, Basant and Nayak, Richi and Mittal, Namita and Patnaik, Srikanta},
  publisher = {Springer},
  address   = {Singapore},
  isbn      = {978-981-15-1216-2},
  issn      = {2524-7565},
  doi       = {10.1007/978-981-15-1216-2_4},
  pages     = {85 -- 109},
  year      = {2020},
  abstract  = {Comment sections of online news platforms are an essential space to express opinions and discuss political topics. In contrast to other online posts, news discussions are related to particular news articles, comments refer to each other, and individual conversations emerge. However, the misuse by spammers, haters, and trolls makes costly content moderation necessary. Sentiment analysis can not only support moderation but also help to understand the dynamics of online discussions. A subtask of content moderation is the identification of toxic comments. To this end, we describe the concept of toxicity and characterize its subclasses. Further, we present various deep learning approaches, including datasets and architectures, tailored to sentiment analysis in online discussions. One way to make these approaches more comprehensible and trustworthy is fine-grained instead of binary comment classification. On the downside, more classes require more training data. Therefore, we propose to augment training data by using transfer learning. We discuss real-world applications, such as semi-automated comment moderation and troll detection. Finally, we outline future challenges and current limitations in light of most recent research publications.},
  language  = {en}
}