Search

Clinical text often includes numbers of various types and formats. However, most current text classification approaches do not take advantage of these numbers. This study aims to demonstrate that using numbers as features can significantly improve the performance of text classification models. This study also demonstrates the feasibility of extracting such features from clinical text. Unsupervised learning was used to identify patterns of number usage in clinical text. These patterns were analyzed manually and converted into pattern-matching rules. Information extraction was used to incorporate numbers as features into a document representation model. We evaluated text classification models trained on such representation. Our experiments were performed with two document representation models (vector space model and word embedding model) and two classification models (support vector machines and neural networks). The results showed that even a handful of numerical features can significantly improve text classification performance. We conclude that commonly used document representations do not represent numbers in a way that machine learning algorithms can effectively utilize them as features. Although we demonstrated that traditional information extraction can be effective in converting numbers into features, further community-wide research is required to systematically incorporate number representation into the word embedding process.

Burrowing is a common trait among crayfish thought to help species deal with adverse environmental challenges. Here we used in-vivo experimental data and in-silico modelling of oxygen saturation in a virtual burrow inhabited by crayfish. Except for the entrance 200 mm region, the burrow microenvironment becomes anoxic, on average, within 8 hours, and 12-hour day-night multiple cycles were not sufficient for refreshing the burrow microenvironment even with temporary lack of crayfish. We asked whether the ecological category of crayfish burrowing activity is reflected in the physiological ability to cope with hypoxia and anoxia. As dissolved oxygen declined, respiration patterns of primary burrowers differed from those of secondary and tertiary burrowers, showing also the highest variability in anoxia tolerance. Secondary burrowers showed consistent tolerance with all species exhibiting a mean survival of > 3h anoxic conditions. Tertiary burrowers were variable, exhibiting moderate to zero tolerance of anoxia. The adaptive mechanisms to cope with hypoxia might be a basal legacy from the crayfish monophyletic ancestors – lobsters, traveller crustaceans often reaching deep depths in the ocean. These results challenge the current understanding of crayfish ecology, opening an evolutionary ecological perspective which might be relevant for the next generation of phylogenetical approaches.

Somatic cell count (SCC) can be used as an indicator of subclinical mastitis and its analysis in relation with the milk composition can provide useful information on the existence of some correlations or patterns. Based on milk production data recorded during 5 years (2012-2015, 2017) at the Research and Development Station for Bovine Arad we conducted a statistical analysis aiming to identify correlations between SCC and milk characteristics (protein and fat content, lactose, nonfat solids, milk quantity, pH, casein) and to find potential profiles of SCC evolution. The correlation analysis was based on 226 lactating cows for which at least 20 measurements were available. Both classical correlation coefficient (i.e. Pearson) and correlation coefficient for repeated measurements (i.e. Bland-Altman) have been computed. In both cases, a moderate negative correlation between SCC and the lactose level has been identified while no significant correlation between SCC and the other milk characteristics has been detected. However, a more accurate description of the relation between SCC and lactose was obtained using a linear mixed model. Aiming to analyze SCC profiles, an additional attribute has been added to the data based on the following encoding rule: the attribute has value 0 if SCC is smaller than 2x105cells/ml, 1 if it is larger than 2x105cells/mL and 2 if the value is missing. In this way, data vectors containing 13 values per year have been constructed for 175 cows and a dissimilarity matrix has been constructed as a first step for cluster analysis. Overall, the results have shown that lactose and SCC were negatively correlated.

Multiple causes can determine the disturbance of natural equilibrium in a population of a species, a common one being the presence of invasive competitors. It is well known that invasives can drive native species to the resettlement of trophic position, changing reproduction strategies or even daily normal behaviours. Here we investigated the hypothesis that more effective biological weapons of an intruder (Faxonius limosus) came with increased boldness behaviour, contributing to their invasion success in competition against the native species (Pontastacus leptodactylus). We tested boldness of specimens representing the two species by video-based assessment of crayfish individuals’ attempts to leave their settlement microenvironment. The experiment was followed by a series of measurements concerning chelae biometry, force, and muscle energetics. The native species was found less expressive in terms of boldness even if it has larger chelae and better muscular tissue performance. In contrast, because of better biomechanical construction of the chelae, the invasive species proved to be capable of twice superior force achievements, which expectedly explain their bolder behaviour. The agonistic behaviour strategy of the invasive crayfish species appears to be based on sheer physical superiority while the native crayfish relies on intimidation display. These findings suggest that, in interspecific interactions, the behaviour strategy of the invasive crayfish species is based on sheer physical superiority while the native crayfish relies on intimidation display.

Hate speech is an important problem in the management of user-generated content. To remove offensive content or ban misbehaving users, content moderators need reliable hate speech detectors. Recently, deep neural networks based on the transformer architecture, such as the (multilingual) BERT model, achieve superior performance in many natural language classification tasks, including hate speech detection. So far, these methods have not been able to quantify their output in terms of reliability. We propose a Bayesian method using Monte Carlo dropout within the attention layers of the transformer models to provide well-calibrated reliability estimates. We evaluate and visualize the results of the proposed approach on hate speech detection problems in several languages. Additionally, we test if affective dimensions can enhance the information extracted by the BERT model in hate speech classification. Our experiments show that Monte Carlo dropout provides a viable mechanism for reliability estimation in transformer networks. Used within the BERT model, it ofers state-of-the-art classification performance and can detect less trusted predictions. Also, it was observed that affective dimensions extracted using sentic computing methods can provide insights toward interpretation of emotions involved in hate speech. Our approach not only improves the classification performance of the state-of-the-art multilingual BERT model but the computed reliability scores also significantly reduce the workload in an inspection of ofending cases and reannotation campaigns. The provided visualization helps to understand the borderline outcomes.

Due to complex experimental settings, missing values are common in biomedical data. To handle this issue, many methods have been proposed, from ignoring incomplete instances to various data imputation approaches. With the recent rise of deep neural networks, the field of missing data imputation has oriented towards modelling of the data distribution. This paper presents an approach based on Monte Carlo dropout within (Variational) Autoencoders which offers not only very good adaptation to the distribution of the data but also allows generation of new data, adapted to each specific instance. The evaluation shows that the imputation error and predictive similarity can be improved with the proposed approach.

The long-term survival of a species requires, among other things, gene flow between populations. Approaches for the evaluation of fragmentation in the frame of freshwater habitats consider only a small amount of the information that combined demography and geography are currently able to provide. This study addresses two species of Austropotamobius crayfish in the light of population genetics, spatial ecology and protected areas of the Carpathians. Advancing the classical approaches, we defined ecological distances upon the rasterised river network as a surrogate of habitat resistance to migration, quantifying the deviations from the species´ suitability range for a set of relevant geospatial variables in each cell of the network. Molecular analyses revealed the populations of the two Austropotamobius crayfish species are clearly distinct, lacking hybridisation. Comparing pairs of populations, we found, in some cases, a strong disagreement regarding genetic and ecological distances, potentially due to human-mediated translocations or the geophysical phenomena of regressive erosion, which may have led to unexpected colonisation routes. Protected areas were found to offer appropriate local habitat conditions but failed to ensure connectivity. The methodology applied in this study allowed us to quantify the contribution of each geospatial (environmental) variable to the overall effect of fragmentation, and we found that water quality was the most important variable. A multilevel approach proved to reveal a better understanding of drivers behind the distribution patterns, which can lead to more adequate conservation measures.

As a result of social network popularity, in recent years, hate speech phenomenon has significantly increased. Due to its harmful effect on minority groups as well as on large communities, there is a pressing need for hate speech detection and filtering. However, automatic approaches shall not jeopardize free speech, so they shall accompany their decisions with explanations and assessment of uncertainty. Thus, there is a need for predictive machine learning models that not only detect hate speech but also help users understand when texts cross the line and become unacceptable. The reliability of predictions is usually not addressed in text classification. We fill this gap by proposing the adaptation of deep neural networks that can efficiently estimate prediction uncertainty. To reliably detect hate speech, we use Monte Carlo dropout regularization, which mimics Bayesian inference within neural networks. We evaluate our approach using different text embedding methods. We visualize the reliability of results with a novel technique that aids in understanding the classification reliability and errors., Comment: The final authenticated publication is available online at https://doi.org/10.1007/978-3-030-31372-2_24