Monitoring tree regeneration in forest areas disturbed by resource extraction is a requirement for sustainably managing the boreal forest of Alberta, Canada. Small remotely piloted aircraft systems (sRPAS, a.k.a. drones) have the potential to decrease the cost of field surveys drastically, but produce large quantities of data that will require specialized processing techniques. In this study, we explored the possibility of using convolutional neural networks (CNNs) on this data for automatically detecting conifer seedlings along recovering seismic lines: a common legacy footprint from oil and gas exploration. We assessed three different CNN architectures, of which faster region-CNN (R-CNN) performed best (mean average precision 81%). Furthermore, we evaluated the effects of training-set size, season, seedling size, and spatial resolution on the detection performance. Our results indicate that drone imagery analyzed by artificial intelligence can be used to detect conifer seedling in regenerating sites with high accuracy, which increases with the size in pixels of the seedlings. By using a pre-trained network, the size of the training dataset can be reduced to a couple hundred seedlings without any significant loss of accuracy. Furthermore, we show that combining data from different seasons yields the best results. The proposed method is a first step towards automated monitoring of forest restoration/regeneration.

In this work we address the problem of argument search. The purpose of argument search is the distillation of pro and contra arguments for requested topics from large text corpora. In previous works, the usual approach is to use a standard search engine to extract text parts which are relevant to the given topic and subsequently use an argument recognition algorithm to select arguments from them. The main challenge in the argument recognition task, which is also known as argument mining, is that often sentences containing arguments are structurally similar to purely informative sentences without any stance about the topic. In fact, they only differ semantically. Most approaches use topic or search term information only for the first search step and therefore assume that arguments can be classified independently of a topic. We argue that topic information is crucial for argument mining, since the topic defines the semantic context of an argument. Precisely, we propose different models for the classification of arguments, which take information about a topic of an argument into account. Moreover, to enrich the context of a topic and to let models understand the context of the potential argument better, we integrate information from different external sources such as Knowledge Graphs or pre-trained NLP models. Our evaluation shows that considering topic information, especially in connection with external information, provides a significant performance boost for the argument mining task.

In our project ReMLAV, funded within the DFG Priority Program RATIO (http://www.spp-ratio.de/), we focus on relational and fine-grained argument mining. In this article, we first introduce the problems we address and then summarize related work. The main part of the article describes our research on argument mining, both coarse-grained and fine-grained methods, and on same-side stance classification, a relational approach to the problem of stance classification. We conclude with an outlook.

No abstract

Many interesting use cases of research data classifiers presuppose that a research data item can be mapped to more than one field of study, but for such classification mechanisms, reproducible evaluations are lacking. This paper closes this gap: It describes the creation of a training and evaluation set comprised of labeled metadata, evaluates several supervised classification approaches, and comments on their application in scientometric research. The metadata were retrieved from the DataCite index of research data, pre processed, and compiled into a set of 613,585 records. According to our experiments with 20 general fields of study, multi layer perceptron models perform best, followed by long short-term memory models. The models can be used in scientometric research, for example to analyze interdisciplinary trends of digital scholarly output or to characterize growth patterns of research data, stratified by field of study. Our findings allow us to estimate errors in applying the models. The best performing models and the data used for their training are available for re use.