Emotions widely affect human decision-making. This fact is taken into account by affective computing with the goal of tailoring decision support to the emotional states of individuals. However, the accurate recognition of emotions within narrative documents presents a challenging undertaking due to the complexity and ambiguity of language. Performance improvements can be achieved through deep learning; yet, as demonstrated in this paper, the specific nature of this task requires the customization of recurrent neural networks with regard to bidirectional processing, dropout layers as a means of regularization, and weighted loss functions. In addition, we propose sent2affect, a tailored form of transfer learning for affective computing: here the network is pre-trained for a different task (i.e. sentiment analysis), while the output layer is subsequently tuned to the task of emotion recognition. The resulting performance is evaluated in a holistic setting across 6 benchmark datasets, where we find that both recurrent neural networks and transfer learning consistently outperform traditional machine learning. Altogether, the findings have considerable implications for the use of affective computing.

No abstract

The aim of this study was to examine the capability of structure-from-motion photogrammetry in defining the geometry of cliffs and undercuts in rocks of complex geomorphology. A case site was chosen along pocket beaches near the village of Stara Baška on the Adriatic Sea island of Krk, Gulf of Kvarner, Croatia, where cliff erosion of 5 m in breccias was identified by comparison of aerial photographs from 1960 and 2004. The 3D point cloud was derived from approx. 800 photos taken on 9 January 2014 by a single camera from various elevations and angles, and processed using the online software ReCap (Autodesk). Data acquisition was found to be quick and the method easy to implement. The difference between the georeferenced 3D cloud points and an RTK-GPS survey was 7 cm, i.e. within the limits of RTK-GPS precision. Quantifying the spatial variation in undercut geometries revealed that the deepest and largest (17 m 3 ) undercut was in the south-eastern sector of the beach. Reconstructing the detailed geomorphology of this 3.8-m-deep undercut convincingly demonstrates the high efficiency of the method. Such assessments of spatiotemporal changes in undercut and overhang volumes can prove useful for evaluations of cliff erosion risk. Coupled with the low cost and relatively simple application, this is evidently an attractive technique for meaningful geotechnical and coastal engineering monitoring in the future on the island of Krk and, for that matter, also on other Adriatic islands and in similar settings worldwide.

Predicting context-dependent and non-literal utterances like sarcastic and ironic expressions still remains a challenging task in NLP, as it goes beyond linguistic patterns, encompassing common sense and shared knowledge as crucial components. To capture complex morpho-syntactic features that can usually serve as indicators for irony or sarcasm across dynamic contexts, we propose a model that uses character-level vector representations of words, based on ELMo. We test our model on 7 different datasets derived from 3 different data sources, providing state-of-the-art performance in 6 of them, and otherwise offering competitive results.

No abstract