The amount of content published on the Internet has grown exponentially in recent times. Social networks have enabled this content to reach an even wider audience. However, the freedom of communication provided by these networks can consequently facilitate the spread of offensive language and hate speech. Although social media platforms have attempted to implement mechanisms for detecting and addressing such content, it remains an ongoing challenge, particularly for languages other than English, such as Spanish. One promising approach to tackle this problem is the application of Natural Language Processing (NLP) tools, which rely on the use of language models and deep learning for text classification. In this work, an approach for detecting Spanish Hate Speech with ALBETO (SHS-ALBETO) is proposed. Experimentation is conducted with HatEval dataset. The performance of SHS-ALBETO is compared with other competing models, such as BERT, BETO, and DistilBETO, along with other proposals from the state-of-the-art. SHS-ALBETO has improved the existing results in the scientific literature, simultaneously providing reduced computing times. Additionally, analyses of the results have revealed its advantages together with challenging aspects that must be addressed to further improve the performance of this kind of approach.

The global decline in bee populations poses a critical threat to biodiversity and ecosystem stability, motivating the adoption of precision beekeeping strategies that combine sensor networks with data-driven models to optimise hive management and reduce colony losses. This study introduces a multivariate autoregressive multilayer perceptron (AMLP) model that integrates historical internal hive variables (temperatures, weight, humidity, and pressure) with external climatological data to forecast future states of these endogenous variables. Data were collected from 13 sensor-equipped hives of the BeeObserver project. The AMLP was evaluated against a standard multilayer perceptron (MLP) and a vector autoregressive (VAR) model using 10-fold rolling-window cross-validation. Forecast performance was assessed using two different error metrics for 1-and 3-day horizons. Across all hives, the AMLP reduced the mean percentage error by approximately 6%-7% relative to the MLP and up to 1.3% relative to the VAR, achieving superior predictive accuracy, with statistically significant improvements for most internal variables. By combining autoregressive lags with neural network flexibility, the AMLP captures both temporal dependencies and specific patterns while supporting incremental retraining as new data arrive. This approach provides scalable, adaptive, and real-time prediction of hive dynamics, offering a robust tool for proactive decision-making in precision beekeeping. The results demonstrate that integrating temporal and environmental information through AMLP models enhances predictive accuracy and supports timely interventions, ultimately improving colony health and resilience. These findings highlight the potential of advanced data-driven forecasting models to strengthen sustainable apiculture practices and contribute to the conservation of bee populations.

The use of keywords is increasingly being applied across diverse domains, including the movie industry, whose main platforms are adopting advanced natural language processing techniques. Algorithms for automatic extraction of keywords can provide relevant information in this domain. The most novel approaches covering several categories (statistics, graphs, word embedding, and hybrid) have been considered in a model study framework. They have been implemented, applied, and evaluated with standard datasets. In addition, a movie dataset with gold standard keywords, based on textual metadata from synopses and reviews, has been specifically developed for this scope. Keyword extraction models have been evaluated in terms of F-score and computation time. Furthermore, content analysis, both quantitative and qualitative, of the extracted keywords in the movie context has been performed. Results show a great variability in model performance and computation time among the different models. Qualitative results, in addition to F-score and computation time, demonstrate that keyword extraction works better with synopses than with reviews. The quantitative content analysis revealed that EmbedRank effectively reduces redundancy and limits the use of proper nouns, leading to high-quality keywords.

Parkinson's disease (PD) is a neurodegenerative disorder that affects more than 10 million people worldwide. Despite its prevalence, the detection of PD remains a complicated task, as no gold standard test has yet been developed to provide an accurate diagnosis. In this context, many recent studies have focused on the automatic detection and progression tracking of PD from voice-related characteristics, being feature engineering the most common approach. This work intends to address an existing research gap by introducing a novel strategy that analyzes raw voice waveforms. Despite recent advancements, one of the significant hurdles is still the lack of extensive and diverse datasets. This article also implements a data augmentation solution. Big Vocoder Slicing Adversarial Network (BigVSAN) is used to generate synthetic voice data that mimics the characteristics of real patients and healthy subjects. For the PD detection task, deep learning models such as ResNet, LSTM-FCN, InceptionTime, and CDIL-CNN are used. The experiments were performed using the speech task of sustained vowel /a/ in the PC-GITA database, which contains the recordings of healthy and PD subjects. CDIL-CNN achieves the best results, improving the accuracy by 15.87% (8.96%) compared to the model that does not use augmented data (from the best method found in the literature that uses voice waveforms). The results of this study indicate that models trained with raw waveforms showcase modest but promising performance, underlying the potential of audio analysis to improve the early detection of PD, providing a non-invasive and potentially remotely applicable method.

Introduction: The relevant prevalence of voice-related pathologies underscores the need for robust computer-aided diagnostic (CAD) systems capable of supporting early detection and continuous monitoring. Electroglottography (EGG), a non-invasive technique measuring vocal fold contact area, has proven valuable in identifying and diagnosing vocal disorders. Problem statement: Traditional diagnostic methods struggle with the dependent nature of EGG measurements within subjects, leading to challenges in managing within-subject variability and supporting multi-class classification. Objectives: This study aims to design, implement, and evaluate two ensemble-based approaches that address the dependency in EGG measurements. The goal is to enhance the detection of vocal disorders by managing within-subject variability and facilitating multi-class classification. Methods: The proposed methods utilize replicated acoustic biomarkers derived from EGG signals. Simulation-based experiments were conducted to assess the robustness and effectiveness of these methods. Additionally, experiments were performed using EGG signals from the Saarbrüecken Voice Database (SVD). Results: Simulation results indicate that integrating replicated data improves accuracy rates compared to non-replicated models. Experiments on SVD demonstrated the robustness of the proposed methodology across different vowels in classifying healthy individuals, patients with laryngitis, and those with vocal fold paralysis. Conclusion: The data ensemble-based approaches developed effectively manage the dependent nature of EGG measurements, enhancing the detection and classification of vocal disorders. These methods can be applied to other data types where replications play a key role. Future research should focus on collecting comprehensive EGG databases and further exploring multi-class classification methods to solidify EGG and machine learning as a valuable tool for non-invasive assessment of laryngeal function.

The field of multi-objective optimization is experiencing a relevant growth due to its successful applications numerous real-life problems. Two prominent trends are indicator-based and decomposition-based search strategies. However, the performance of these search strategies depends on the specific problem to solve. In the context of automatic summarization, hybridization of these techniques is an interesting and challenging proposal, which aims to improve the performance. For this reason, an Indicator and Decomposition-based Hybrid Evolutionary Algorithm (IDHEA) has been designed, implemented, and tested for addressing the automatic summarization problem. The proposed hybrid multi-objective approach has integrated the fundamentals of both indicator-based and decomposition-based techniques to solve this particular problem. Experimentation has been conducted using Document Understanding Conferences (DUC) dataset. The performance has been assessed by means of multi objective metrics such as hypervolume, inverted generational distance, distance to ideal point, and set coverage, while summary quality has been evaluated with Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. The proposed approach has outperformed standard algorithms in multi-objective evaluation, in addition to improving the existing results in the scientific literature in terms of summary quality.

Nowadays, automatic text summarization task is a matter that has acquired special relevance in numerous contexts. Particularly, sentiment analysis and opinion mining need summarization methods to quickly analyze public opinion about any event. In this way, the aim of the sentiment-oriented summarization approach is to produce a summary reflecting the sentiment of the authors’ opinions, covering the main content, and reducing the redundancy. In this work, a Sentiment-Oriented Dominance-based Bee Algorithm (SODBA) has been designed, developed, and applied for solving this problem. Experimentation has been conducted with datasets provided by Document Understanding Conferences. The evaluation of the results has been carried out by using the Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics and the Pearson correlation coefficient. The reported results have outperformed those obtained in the scientific literature in terms of ROUGE metrics. Moreover, SODBA has been applied to the tweets concerning the COVID-19 pandemic to obtain the summaries of the days with the most positive and the most negative sentiment.

An early detection of Parkinson’s Disease (PD) is important for patients’ quality of life and current detection methods are not optimal in this task. The search of an automatic, low-cost, non-invasive method would be appropiate to delevope a computer-aided diagnosis system. Machine learning models based on data such as neural networks are at the core of many of these systems. In this work 5 convolutional neural network architectures have been compared by their capability on classifying spectrograms generated with PCGita database. For this classification task, transfer learning using Saarbrücken Voice Database and a data augmentation technique have been addresed with 14 models from the 5 architectures. After performing 5-fold cross-validation, results show that VGG16 architecture is able to distinguish between healthy and PD spectrograms with around 85% of global accuracy. Further research is needed to explore the potential of this technology with multicondition training in medical environments.

In recent years, there have been relevant advances in precision beekeeping. These advances are mainly focused on proposing sensor systems that collect crucial information for bee welfare, creating integrated architectures that allow beekeepers to monitor the current state of their hive through real-time data. However, there is a lack of predictive models that would allow beekeepers to anticipate specific events that endanger bee welfare and lead to a decline in productivity. Specifically, predictive approaches accounting for the high correlation among internal variables of beehives have not been developed to date. To address this research gap, multivariate predictive models, including auto-regressive state-space and time series models, have been implemented and applied to four different hives from the we4bee project. These models aim to predict the internal variables of beehives (four different temperatures, humidity, and weight) by utilizing the meteorological conditions to which the hives are exposed. A cross-validation adapted to time series data was employed for model generalization assessment. Prediction models based on vector time series exhibited superior performance in forecasting internal hive variables compared to multivariate auto-regressive state-space models. Overall, the approach based on the vector error correction model yielded the best balance between fit, prediction, and computational cost. The VEC-based approach produces predictions with maximum mean absolute errors of 177 (312)g in weight, 3.366 (3.802)% in humidity, and 1.122 (1.685)°C in temperature at 1 (3)-days ahead when dealing with beehives exhibiting a high degree of correlation in their internal variables. Moreover, the VEC-based approach requires less than a second to perform the time series fitting process, which makes it particularly interesting for application in big data environments. The integration of such models into a decision support system would meet the need of beekeepers to anticipate potential threats to the welfare of their bee colonies, streamlining their monitoring processes while eliminating the need for continuous inspections.

Despite significant advances on automatic detection of Parkinson’s disease (PD) based on speech, several open challenges still need to be addressed before a validated computer-aided diagnosis system can be used practically. One of these challenges lies in considering the potential corruption of speech caused by environmental noises, which may be nonstationary and exhibit varied characteristics. Speech features automatically extracted from diadochokinetic (DDK) tests have shown utility in assessing articulatory aspects of speech impairment in PD. The authors propose an automatic PD detection system based on a multicondition training (MCT) framework. The approach considers various types of realistic acoustic noise in addition to DDK recordings and uses machine learning for feature selection and classification. For each experiment, the noise addition process did not artificially increase the dataset size, as each subject’s recordings were either affected by a single noise type or had no injected noise. To compare with this MCT-based approach, an alternative method is examined where training involves speech samples affected by uniform noise conditions. This method, referred to as single-condition training (SCT), involves training with features either from the original waveforms or from waveforms altered by noise addition, ensuring uniformity by using the same type of realistic noise across all the training samples. The benefit of the MCT approach is demonstrated by showing the results obtained in classification tests to discriminate patients affected by PD from healthy individuals. The experiments performed were based on an in-house voice recording database composed of 30 individuals diagnosed with PD and 30 healthy controls. The speech samples were recorded using a smartphone as a data collection device so that the samples were not affected by speech compression algorithms. Both approaches (SCT and MCT) were tested against each specific type of noise under consideration. The mean accuracy rates showed improvements of 1.68%, 5.18%, and 4.39% for/pa/,/ta/, and/ka/ syllables, respectively, when using MCT compared with SCT. To the best of the authors’ knowledge, this is the first strategy published in the literature to deal with the potential corruption of speech by environmental noise in automatic PD detection aid systems based on DDK tests.

Currently, automatic multi-document summarization is an interesting subject in numerous fields of study. As a part of it, query-focused summarization is becoming increasingly important in recent times. These methods can automatically produce a summary based on a query given by the user, including the most relevant information from the query at the same time as the redundancy among sentences is reduced. This can be achieved by developing and applying a multi-objective optimization approach. In this paper, an Indicator-based Multi-Objective Variable Neighborhood Search (IMOVNS) algorithm has been designed, implemented, and tested for the query-focused extractive multi-document summarization problem. Experiments have been carried out with datasets from Text Analysis Conference (TAC). The results were evaluated using the Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. IMOVNS has greatly improved the results presented in the scientific literature, providing improvement percentages in ROUGE metric reaching up to 69.24% in ROUGE-1, up to 57.70% in ROUGE-2, and up to 77.37% in ROUGE-SU4 scores. Hence, the proposed IMOVNS offers a promising solution to the query-focused summarization problem, thus highlighting its efficacy and potential for enhancing automatic summarization techniques.

Morality is a topic that people are increasingly concerned about. Morality is observed and measured during public acts or when developing and consuming products, such as movies. The Moral Foundations Theory (MFT) was developed to rigorously perform these measurements with the support of the Moral Foundations Dictionary (MFD). In this paper, a Word Embedding-based Moral Foundation Assignment (WEMFA) approach has been designed, implemented, and applied to the movie domain for multiple assignment of moral foundations. WEMFA may use any dictionary, and it has been applied to a movie collection generated from movie synopses. A comparison between WEMFA and MoralStrength, the only approach found in the scientific literature, has been carried out. The proposed approach provided a percentage improvement of 41.7% with respect to the best version of MoralStrength, which uses an extension of the original MFD almost 10 times larger in number of terms. In addition, an extension of the original MFD (MFD24) has been built by adding 14 new moral foundations to the 10 original ones, enriching the moral context. WEMFA provided a mean accuracy of 78% with MFD24 despite the increment of the number of moral foundations. Besides, new extended dictionaries or even totally different ones can be used with WEMFA, since it does not need any training.

Objective: Voice analysis based systems offer low-cost, highly available automatic diagnostic aid for Parkin-son's disease (PD) detection anywhere a smartphone with a broadband connection is available. However, reliability depends on factors affecting the communication channel. In this paper the effects of recording device mismatch are analyzed. Multicondition training (MCT) is proposed to improve robustness against that mismatch. Methods: An experiment on 30 PD patients and 30 healthy subjects was designed. 3 vocalizations of sustained \a\ were recorded using a smartphone. These recordings, along with a simulation of 8 additional smartphones, were analyzed. Acoustical features were extracted and averaged per patient and recording device. Machine learning was used to distinguish healthy from PD patients by using different combinations of train-test smartphones. Results: By using the same device for training and testing, a 10% best-worse mean accuracy drop is observed. The gap among different devices reaches 37%. MCT retains 90% of the maximum accuracy and exceeds a 20% mean accuracy while lowers dispersion of the aggregated results obtained with single condition. Smartphone position shows a direct impact on performance. Conclusion: Recording device has a major effect on results. It is also found that positioning of the recording device might also be influential. Using MCT appears to improve robustness. Significance: Results support the use of mobile devices to create an automated PD detection test. It is also encouraged to consider the use of MCT to obtain more robust and reliable results across different devices.

Streaming services are increasingly leveraging Artificial Intelligence (AI) technologies for improved content cataloging, user experiences in content discovery, and personalization. A significant challenge in this domain is the automated assignment of microgenres to movies. This study introduces and evaluates approaches based on clustering, topic modeling, and word embedding to address this task. The evaluation employs a preprocessed dataset containing movie-related data—title tags, synopses, genres, and reviews—alongside a predefined microgenre list. Comparisons of three activation functions (binary step, ramp, and sigmoid) gauge their effectiveness in augmenting microgenre tags. Results demonstrate the superiority of the word embedding approach over clustering and topic modeling in terms of mean accuracy. Even more, the word embedding approach stands as the sole fully automated solution. Analysis indicates that incorporating review-based tags introduces noise and undermines accuracy. Besides, the word embedding approach yields optimal outcomes using the sigmoid function, effectively doubling assigned tags while maintaining matching quality. This sheds light on the potential of word embedding methods within the movie domain.

Currently, the explosive growth of the information available on the Internet makes automatic text summarization systems increasingly important. A particularly relevant challenge is the update summarization task. Update summarization differs from traditional summarization in its dynamic nature. While traditional summarization is static, that is, the document collections about a specific topic remain unchanged, update summarization addresses dynamic document collections based on a specific topic. Therefore, update summarization consists of summarizing the new document collection under the assumption that the user has already read a previous summarization and only the new information is interesting. The multiobjective number-one-selection genetic algorithm (MONOGA) has been designed and implemented to address this problem. The proposed algorithm produces a summary that is relevant to the user’s given query, and it also contains updates information. Experiments were conducted on Text Analysis Conference (TAC) datasets, and Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics were considered to assess the model performance. The results obtained by the proposed approach outperform those from the existing approaches in the scientific literature, obtaining average percentage improvements between 12.74% and 55.03% in the ROUGE scores.

The number of scientific publications in different knowledge fields has considerably grown in recent times. This makes difficult for researchers to synthesize all the scientific-technical advances, so automatic summarization methods of scientific papers would be helpful. These methods generate a summary from a reference paper with its most relevant contributions. More specifically, citation-based summarization considers the citation contexts to the reference paper in subsequent publications. For the first time, this problem has been formulated as a multi-objective optimization problem, optimizing the content coverage and the redundancy reduction in a simultaneous way. A Decomposition-based Multi-Objective optimization algorithm for Citation-based Summarization (DMOCS) has been designed, developed, and applied for solving this problem. The results obtained by the proposed approach have improved the existing ones in the scientific literature between 17.47% and 133.50%, increasing the ROUGE percentage improvements when the N-gram is larger. Besides, an exhaustive analysis of the different parts of a scientific paper has been performed, showing that the citations from the citing papers with their corresponding spans in the reference paper impact in the quality of a citation-based summary.

Movie recommender systems are increasingly present in our daily lives, offering content of interest from streaming providers. Objectives in addition to the liking probability can be proposed to provide movie recommendations. However, there is a lack of recommenders that are aware of the benefit and that address the multi-objective nature of the problem. A profit-aware recommender system based on swarm intelligence in a multi-objective environment (multi-objective artificial bee colony, MOABC) has been designed, imple-mented, and applied. The proposed approach incorporates new intelligent operators that try to improve both objectives (liking probability and profit) simultaneously in each itera-tion instead of exclusively using randomness. This increases the quality of the recommen-dations with respect to the state-of-the-art algorithms. This new proposal has been evaluated using MovieLens datasets, covering different sizes (large, medium, and small). The experiments show that the MOABC performs better than collaborative filtering (CF, a standard in recommender systems) and Non-dominated Sorting Genetic Algorithm II (NSGA-II, the only multi-objective proposal in scientific literature that is profit aware) in terms of accuracy and global profit. Furthermore, statistical analysis shows that the pro-posed approach generates better and more robust results, also showing that the multi -objective nature of the problem must be exploited.(c) 2023 Elsevier Inc. All rights reserved.

Precision beekeeping combines technology and statistics aimed at managing an apiary effectively and reducing the risk of situations that can lead to bee population losses. Databases of the we4bee project of three sensorised beehives were considered for analysis. They contain interior sensor data (temperature, relative humidity, and weight) and data of meteorological events. Static and dynamic vector autoregressive models and linear and nonlinear regression models were constructed to predict the hives' internal variables. They were compared by 100-fold cross-validation adapted for time series. In general, the dynamic vector autoregressive model provided the best predictions, with a feasible computational cost. Only in some specific cases did the static vector autoregressive version produces smaller errors, although the differences were not statistically significant. Generalised additive and dynamic linear models always provided less accurate results than the dynamic vector autoregressive model. There is a need of integrating accurate predictive models, such as the dynamic vector autoregressive one. This predictive model can be integrated into a decision support system to alert the beekeeper of out-of-the-ordinary situations in the hives, and thus aid in their efficient management.

Motivated by a longitudinal oral health study, the Signal-Tandmobiel® study, an inhomogeneous mixed hidden Markov model with continuous state-space is proposed to explain the caries disease process in children between 6 and 12 years of age. The binary caries experience outcomes are subject to misclassification. We modelled this misclassification process via a longitudinal latent continuous response subject to a measurement error process and showing a monotone behaviour. The baseline distributions of the unobservable continuous processes are defined as a function of the covariates through the specification of conditional distributions making use of the Markov property. In addition, random effects are considered to model the relationships among the multivariate responses. Our approach is in contrast with a previous approach working on the binary outcome scale. This method requires conditional independence of the possibly corrupted binary outcomes on the true binary outcomes. We assumed conditional independence on the latent scale, which is a weaker assumption than conditional independence on the binary scale. The aim of this article is therefore to show the properties of a model for a progressive longitudinal response with misclassification on the manifest scale but modelled on the latent scale. The model parameters are estimated in a Bayesian way using an efficient Markov chain Monte Carlo method. The model performance is shown through a simulation-based example, and the analysis of the motivating dataset is presented.

Voice pathologies have a direct impact on human communication. One of the most common voice disorders is Reinke’s edema. Speech analysis algorithms applied to voice recordings in combination with machine learning techniques are explored to develop non-invasive low-cost tools to help diagnose this pathology. Different approaches have been compared to discriminate subjects affected by Reinke’s edema from healthy ones. Several classification methods have been used, such as decision trees, k-nearest neighbours, neural networks, support vector machines, Bayesian classification, regression analysis and linear discriminant. The experiments are based on two different databases. One of them is the commercial database Massachusetts Eye and Ear Infirmary (MEEI), recorded under highly controlled acoustical conditions, while the other one is an in-house database, recorded in a more realistic environment. The best results have been obtained by using the model based on neural networks, that achieved an overall accuracy of 100% on MEEI database and 95.49% on the in-house one. These are competitive results in comparison with those presented in the scientific literature and show the potential of these techniques to be employed within a support system for the diagnosis of Reinke’s edema.