Specific User Research Articles

Identifying the Drivers Behind the Dissemination of Online Misinformation: A Study on Political Attitudes and Individual Characteristics in the Context of Engaging With Misinformation on Social Media

The increasing dissemination of online misinformation in recent years has raised the question which individuals interact with this kind of information and what role attitudinal congruence plays in this context. To answer these questions, we conduct surveys in six countries (BE, CH, DE, FR, UK, and US) and investigate the drivers of the dissemination of misinformation on three noncountry specific topics (immigration, climate change, and COVID-19). Our results show that besides issue attitudes and issue salience, political orientation, personality traits, and heavy social media use increase the willingness to disseminate misinformation online. We conclude that future research should not only consider individual’s beliefs but also focus on specific user groups that are particularly susceptible to misinformation and possibly caught in social media “fringe bubbles.”

A Method for Modeling the Individual Convenient Zone of a Human

When designing products to fit a specific user, it is essential to know the user’s upper limb range and strength capabilities at each point of the range space. This is particularly relevant when those capabilities are atypical, e.g., in cases of nonstandard body dimensions, disability, or old age. In this paper, we describe a new method to measure and model the strength capabilities at each point of any person’s upper limb range and then present this information in the form of an Individual Convenient Zone (ICZ) model, which is helpful in virtual product prototyping (CAD) for a specific user. The proposed new method includes creating a database of multiple, detailed, spatial-force characteristics, quickly identifying and modeling the ICZ of any human, and analyzing the ergonomics of a product using a digital human model in combination with the ICZ model. The paper also describes an example of how the proposed methodology can be used to customize kitchen furniture design to the ICZ of a specific senior. The expected result of incorporating ICZ into the design is a better fit between the designed product and the user’s needs, supporting user-centered design methodology. Using ICZ enables the involvement of end-users in product design (participatory design). This is particularly important when designing for people with mobility impairments who are more sensitive to nonergonomic solutions. The ICZ modeling method described in this article may have broader applications beyond kitchen furniture design; it could be used to design workspaces and other similar areas where humans reside and perform manipulation activities.

Target Speaker Extraction by Fusing Voiceprint Features

It is a critical problem to accurately separate clean speech in the multispeaker scenario for different speakers. However, in most cases, smart devices such as smart phones interact with only one specific user. As a consequence, the speech separation models adopted by these devices only have to extract the target speaker’s speech. A voiceprint, which reflects the speaker’s voice characteristics, provides prior knowledge for the target speech separation. Therefore, how to efficiently integrate voiceprint features into the existing speech separation models to improve their performance for the target speech separation is an interesting problem not fully explored. This paper attempts to solve this issue to some extent and our contributions are as follows. First, two different voiceprint features (i.e., MFCCs and d-vector) are explored in the performance enhancement for three speech separation models. Second, three different feature fusion methods are proposed to efficiently fuse the voiceprint features with the magnitude spectrograms originally used in the speech separation models. Third, a target speech extraction method which utilizes the fused features is proposed for two speaker-independent models. Experiments demonstrate that the speech separation models integrated with voiceprint features using three feature fusion methods can effectively extract the target speaker’s speech.

Analysis on development trend of biomass energy distributed utilization

Distributed biomass energy technology has strong adaptability to the types of raw materials, flexible project scale, can meet the needs of special users, better economy in small scale, easier commercial development, in line with the characteristics of biomass resources and China’s national conditions. The distributed utilization of biomass energy mainly includes biomass briquette fuel and biogas. The key technologies include biomass briquette fuel processing and combustion, large and medium-sized biogas engineering technology, biomass gasification pyrolysis and gas utilization. At present, China’s distributed biomass energy technology is mainly in the stage of technological improvement and application demonstration. It is expected that by 2030, most of the key technologies will be basically mature and have the conditions for industrialization. The main development direction of China’s distributed biomass energy industry is the replacement of traditional coal-fired gas, urban/rural clean living energy supply, and rural ecological environmental protection. The pollution caused by burning coal/fuel oil, and at the same time centering on the national new urbanization strategy, provide sustainable clean energy for the construction of new rural areas, and improve the level of rural ecological and environmental protection. At present, the main bottleneck restricting the development of distributed biomass energy industry is economy and reliability. The state should increase investment in technological innovation and policy support, convert the environmental and social benefits of biomass energy into cost benefits, and promote biomass energy. The development of the industry can be distributed and utilized.

Using the Electronic Health Record User Context in Clinical Decision Support Criteria.

Computerized clinical decision support (CDS) used in electronic health record systems (EHRs) has led to positive outcomes as well as unintended consequences, such as alert fatigue. Characteristics of the EHR session can be used to restrict CDS tools and increase their relevance, but implications of this approach are not rigorously studied. To assess the utility of using "login location" of EHR users-that is, the location they chose on the login screen-as a variable in the CDS logic. We measured concordance between user's login location and the location of the patients they placed orders for and conducted stratified analyses by user groups. We also estimated how often login location data may be stale or inaccurate. One in five CDS alerts incorporated the EHR users' login location into their logic. Analysis of nearly 2 million orders placed by nearly 8,000 users showed that concordance between login location and patient location was high for nurses, nurse practitioners, and physician assistance (all >95%), but lower for fellows (77%) and residents (55%). When providers switched between patients in the EHR, they usually did not update their login location accordingly. CDS alerts commonly incorporate user's login location into their logic. User's login location is often the same as the location of the patient the user is providing care for, but substantial discordance can be observed for certain user groups. While this may provide additional information that could be useful to the CDS logic, a substantial amount of discordance happened in specific user groups or when users appeared not to change their login location across different sessions. Those who design CDS alerts should consider a data-driven approach to evaluate the appropriateness of login location for each use case.

Toward privacy-aware federated analytics of cohorts for smart mobility

Location-based Behavioral Analytics (LBA) holds a great potential for improving the services available in smart cities. Naively implemented, such an approach would track the movements of every citizen and share their location traces with the various smart service providers—similar to today's Web analytics systems that track visitors across the web sites they visit. This study presents a novel privacy-aware approach to location-based federated analytics that removes the need for individuals to share their location traces with a central server. The general approach is to model the behavior of cohorts instead of modeling specific users. Using a federated approach, location data is processed locally on user devices and only shared in anonymized fashion with a server. The server aggregates the data using Secure Multiparty Computation (SMPC) into service-defined cohorts, whose data is then used to provide cohort analytics (e.g., demographics) for the various smart service providers. The approach was evaluated on three real-life datasets with varying dropout rates, i.e., clients not being able to participate in the SMPC rounds. The results show that our approach can privately estimate various cohort demographics (e.g., percentages of male and female visitors) with an error between 0 and 8 percentage points relative to the actual cohort percentages. Furthermore, we experimented with predictive models for estimating these cohort percentages 1-week ahead. Across all three datasets, the best-performing predictive model achieved a Pearson's correlation coefficient above 0.8 (strong correlation), and a Mean Absolute Error (MAE) between 0 and 10 (0 is the minimum and 100 is the maximum). We conclude that privacy-aware LBA can be achieved using existing mobile technologies and federated analytics.

Handy-Type Tactile Sensor for Object Recognition Using Convolutional Neural Networks

Tactile sensation obtained from touch is one of the most important factors that determine the impression of an object. A method for identifying tactile textures is required because individual differences exist in feeling of tactile textures. In this study, we propose a handy-type tactile sensor for object recognition using convolutional neural networks (CNNs). The sensor consists of a three-axis pressure sensor and an optical motion sensor for a mouse and can detect time-series data. The object is identified using CNN from the time-series data, namely, pressure and speed of the sensor, when the sensor is moved by a user using a hand. Thus, the sensor system configuration is simple without needing a drive device, and it can be possibly constructed at a low cost. Fifteen types of objects were identified using the prototype sensor. The total average correct recognition rate by one specific user in this study was 77%. Further, the total average recognition rate by four separate users without considering each individual use of the sensor system was 48%. Although the problem in individual identification remained, this result demonstrated the potential for identification application. The proposed sensor system can be used as a functional and useful device.

Social behavior analysis for improving the positive energy transition

The improvement of energy transition programs towards renewable energy sources and efficiency of energy savings programs should focus towards new energy behavioral characteristics for reducing energy consumption and maintaining people well-being. This paper deals with the results of a survey campaign focused on social energy behavior of customers after complete liberalization of the electricity market and facing the increase of electricity prices occurring in the last year. The energy behavioral assistance programs require indicators that can be quantified. The standardization of these results will determine useful information to be derived from the survey poll. The available data on energy consumption from energy utilities could help researchers and companies in estimating the efficiency of the actions towards energy transition. In addition, the up-scaling of various energy behavioral programs and life-cycle of the obtained energy savings should be monitored and evaluated. The energy consumption can be reduced using various behavior assistance programs, tuned for specific end users. The customer can diffuse the good practices involving more and more users of different age, involving and transforming the end-user in an active part. The analysis of the adopted methodology is very positive considering the obtained results with respect to the established objectives.

A Communication Study of Young Adults and Online Dependency during the COVID-19 Pandemic

People use digital media and the Internet daily. The time that young people spend connected to digital devices will increase as technology advances, which could have severe health risks and behavioral dependence implications. In the context of the current pandemic, in which socializing, studying, and working is changing, this question has become particularly relevant. Therefore, we propose to investigate the perceptions of young university adults—understood as generation Z—regarding their digital media practices, particularly during the pandemic, and to study the consequences of a permanent connection to these formats on the development of an addiction to the Internet. Our quantitative method approach applied questionnaires to a sample of 407 young university adults studying in Portugal. Through this survey, we could target a specific user group, quantify their consumption, and measure their online experiences. The results point to an increase in dependence on the Internet during confinement. While it is true that carrying out academic work is one of the reasons for this growth, it is noteworthy that the respondents sought the online world in a significan way to pass the time and escape their routine. In addition, they confess to experiencing some consequences of excessive use, such as sleep disturbances and adverse emotional reactions—such as instabilities, depression, and nervousness—when not online. We conclude that the pandemic has intensified online dependence. However, above all that, it is necessary to look at the mental and general health consequences that this excessive use has brought, which may not be visible or manifested by the youth in the short term, and may come to have consequences in the long term.

Lithuanisation of Software Metaphors

The majority of software that is localised, i.e., adapted or Lithuanised as a product to the Lithuanian market, language, culture, social norms and values, is originally engineered and designed in the United States of America. Despite the fact that internationalisation has to be carried out as a process of neutralising and standardising the software for further localisation of the product, it is usually carried out incompletely or in some cases might be unperformed at all, if companies have not considered the product to be sold globally. Thus, the software contains attributes, for example, writing conventions, legal norms and rules, culture-specific items, metaphors and other that reflect the source culture and language of the software. Software related texts, online help documents, resource files, messages and other contain various metaphors that are related to applications, malfunctioning, software components and menu items and hardware. At the same time, conceptual metaphors represent the worldview, and its perception is pertinent only to some specific language users. The idea of localisation is both to adapt products to target users in terms of their culture and language and make them feel that the product was designed and created for them. Thus, taking in mind that conceptual metaphors represent the perception of the worldview, the crucial question arises, if it is possible to adapt, i.e., Lithuanise, conceptual metaphors to such an extent that they would reflect the worldview and perception of the target language and culture the tropes are used to represent, categorise and depict. At the same time, it becomes interesting to examine how the trope is being changed in the target language as well as analyse the capacities of the language to change, adapt to and embrace new things it has to name.

A meta-learning BCI for estimating decision confidence

Objective. We investigated whether a recently introduced transfer-learning technique based on meta-learning could improve the performance of brain–computer interfaces (BCIs) for decision-confidence prediction with respect to more traditional machine learning methods. Approach. We adapted the meta-learning by biased regularisation algorithm to the problem of predicting decision confidence from electroencephalography (EEG) and electro-oculogram (EOG) data on a decision-by-decision basis in a difficult target discrimination task based on video feeds. The method exploits previous participants’ data to produce a prediction algorithm that is then quickly tuned to new participants. We compared it with with the traditional single-subject training almost universally adopted in BCIs, a state-of-the-art transfer learning technique called domain adversarial neural networks, a transfer-learning adaptation of a zero-training method we used recently for a similar task, and with a simple baseline algorithm. Main results. The meta-learning approach was significantly better than other approaches in most conditions, and much better in situations where limited data from a new participant are available for training/tuning. Meta-learning by biased regularisation allowed our BCI to seamlessly integrate information from past participants with data from a specific user to produce high-performance predictors. Its robustness in the presence of small training sets is a real-plus in BCI applications, as new users need to train the BCI for a much shorter period. Significance. Due to the variability and noise of EEG/EOG data, BCIs need to be normally trained with data from a specific participant. This work shows that even better performance can be obtained using our version of meta-learning by biased regularisation.

Empirical analysis and modelling social network user interaction behavior and time characteristics based on selection preference

User interaction plays an important role in the formation of network public opinion. Many studies have found the power-law characteristics and multimode characteristics of individual complex behavior in social network, but it is difficult to systematically describe and explain the physical phenomena of individual interaction by mathematical methods. To deeply analyze the causes of this phenomenon, it is inseparable from the detailed study of micro individuals and their interaction process. In social network, the reposting behavior of a specific user reflects his individual preference regarding the process of interaction. The frequency of interaction between friends with high preference is much higher than that of friends with low preference. For the first time, we try to study the interaction behavior of users from the perspective of micro individuals by considering user selection preferences. We establish a mathematical expression model of user interaction behavior in social network. The model give the mathematical description dual-mode characteristics of the physical phenomenon of bimodal characteristics. Through the theoretical analysis and simulation experiments of the model, we analyze the distribution of user interaction behavior and the time characteristics of user interaction process. The empirical analysis of data gathered from Weibo also confirms the conclusion of theoretical research.

RESEARCH ON THE IMPLEMENTATION AND PRACTICE OF WECHAT PUBLIC PLATFORM TEACHING RESOURCES BASED ON PSYCHOLOGICAL AND EMOTIONAL REGULATION — TAKING PRIMARY SCHOOL FOREIGN LANGUAGE (ENGLISH) CURRICULUM AS AN EXAMPLE

BackgroundNowadays, wechat applications are very popular in China. Through the text, pictures, sound and video on the wechat official account platform, all-round communication between individuals and between individuals and the media has been realized. Wechat official account platform is an important functional module of wechat application. On this platform, enterprises and individuals can create an official account to communicate and interact with specific users. Its forms of communication and interaction are very diverse, including text, sound, pictures, pictures, text messages and videos. This paper applies the wechat official account platform to foreign language (English) teaching for primary school students. Developed foreign language (English) learning resources suitable for wechat official account platform.Subjects and Methods330 public primary school teachers were investigated by stratified sampling, including the experimental primary school in F City, the First Affiliated Primary School of s Normal University and the experimental primary school in district a of F City. From the deep understanding of the wechat official account platform, the strong support for the wechat official account platform, the diverse forms of the wechat official account platform, the diverse contents of the wechat official account platform, and the diversity of the wechat official account platform, we conducted a questionnaire survey, and realized the implementation of the wechat official account and the practical teaching resource platform from the perspective of psychology. In addition, the study also found that foreign language anxiety was negatively correlated with foreign language achievement. Similarly, there is a significant negative correlation between foreign language anxiety and self-efficacy. State anxiety, trait anxiety and foreign language learning anxiety are all positively correlated. The “anxiety” component of learners' personality characteristics plays a great role in the generation of foreign language learning anxiety. The correlation coefficient between LCAS and trait anxiety was 0.408, while the correlation coefficient between LCAS and state anxiety was 0.395. Both reached a significant level (P < 0.01). Self efficacy is positively correlated with foreign language achievement. Research shows that foreign language anxiety, gender and self-efficacy can be used as comprehensive variables to measure foreign language learning achievement.ResultsBased on the teaching concept of “teacher led and student-centered”, the traditional classroom teaching and network assisted learning were combined. Through the primary school foreign language (English) learning resource platform based on wechat official account platform \\ primary school foreign language (English) reading resource platform based on wechat official account platform \\ primary school foreign language (English) counseling and Q & a platform based on wechat official account platform, Let students realize resource-based autonomous learning in order to achieve the best effect of foreign language learning.ConclusionFrom the survey data, the teachers of the surveyed schools have realized the advantages of oral English teaching on wechat public platform. Wechat public platform has built a bridge for communication between teachers and students and realized free dialogue between teachers and students. Therefore, the platform has become an oral English training base. Carry out students' English teaching and develop English learning resources suitable for wechat public platform. Based on the concept of “double master”, these two teaching methods are combined with traditional classroom teaching and network assisted learning. Based on wechat public primary school English learning resource platform, wechat public primary school English reading resource platform or wechat public primary school English counseling platform, let primary school students realize autonomous learning and optimize primary school English learning.AcknowledgementsThis project is supported by Zhejiang Industry-Education Integration Engineering Base (Zhejiang Development and Reform Commission society [2020] No. 319).

Telecom Big Data Based User Offloading Self-Optimisation in Heterogeneous Relay Cellular Systems

This paper proposes a telecom big data based user offloading self-optimisation (TBDUOS) scheme. Its aim is to assist telecom operators to effectively balancing the load distribution with achieving good service performance and customer management in heterogeneous relay cellular systems. To achieve these objectives, in the cell-level offloaded traffic analysis stage, the optimal offloaded traffic is calculated to minimise the total blocking probability. In the user-level offloading stage, the user portrait is drawn and the K-MEANS algorithm is employed to manage the users clustering in the heavily loaded cell, and finally shifting users to assistant cells. Simulation results show the TBDUOS scheme can effectively reduce the handover failure and call dropping of specific users, especially voice/stream users, high consumption users, high level users. The TBDUOS scheme can also reduce the blocking probability.

Reactive synthesis of software robots in RPA from user interface logs

Robotic Process Automation (RPA) is an emerging technology in the field of Business Process Management (BPM) that enables the automation of intensive repetitive tasks (or simply routines). RPA solutions access the user interface (UI) layer of software (SW) applications and provide a virtual workforce of SW robots that are able to mimic human keyboard and mouse interactions with a UI as if a real person was doing them. To take full advantage of this technology, organizations leverage the support of skilled human experts that preliminarily observe how routines are executed on the UI of the involved SW applications, and then implement the executable RPA scripts required to automate the routines enactment by SW robots on a target computer system. However, the current practice is time-consuming and error-prone, as it strongly relies on the ability of the human experts to correctly interpret the routines (and their variants) to automate. In this paper, to tackle this issue, we use a design science research method to develop an approach, called SmartRPA, which is able to interpret the UI logs keeping track of many routine executions, and to automatically synthesize SW robots that emulate the most suitable routine variant for any specific intermediate user input that is required during the routine execution. The approach is implemented as an open-source tool and evaluated with four non-functional requirements employing both syntectic and real-world data.

Nationwide, Operational Sentinel-1 Based InSAR Monitoring System in the Cloud for Strategic Water Facilities in Hungary

The intensive development of both interferometric technology and sensors in recent years allows Interferometric Synthetic Aperture Radar (InSAR)-based applications to be accessible to a growing number of users. InSAR-based services now cover entire countries and soon even the whole of Europe. These InSAR systems require massive amounts of computer processing power and significant time to generate a final product. Most, if not all, of these projects have a limited “monitoring component”, aimed at historical analysis but are rarely, if ever, updated. Consequently, the results do not necessarily meet every purpose or specific user requirement. It is now clear that the increasing computing capacity and big data provided by the sensors have initiated the development of new InSAR services. However, these systems are only useful when linked to specific real-world operational problems. Continuous monitoring of a country’s ageing water management infrastructure has become an increasingly critical issue in recent years, in addition to the threats posed by climate change. Our article provides a comprehensive overview of a nationwide, dedicated, operational InSAR application, which was developed to support the operational work of the Hungarian Disaster Management Service (HDMS). The objective was to provide monthly monitoring of 63 water facilities, including 83 individual objects, distributed throughout Hungary, in combination with the development of a near real-time warning system. Our work involved the compilation of a completely new InSAR System as a Service (SaaS) which incorporates user requirements, preparatory work, the compilation of the Sentinel-1 automatic processing pipeline, the installation of corner reflectors, a special early warning system, and a dedicated user interface. The developed system can automatically start to evaluate the S1 measurements within 24 h of downloading the data into the system storage forward the results toward the warning system before the next image arrives. Users are provided with detailed information on the stability of 70% of the 83 water facility objects monitored through the dedicated user interface. The additional early warning system currently operates as a preliminary “spatial decision support system”, but the HDMS is willing to make it fully operational over the next few years.

Historical and future weather data for dynamic building simulations in Belgium using the regional climate model MAR: typical and extreme meteorological year and heatwaves

Abstract. Increasing temperatures due to global warming will influence building, heating, and cooling practices. Therefore, this data set aims to provide formatted and adapted meteorological data for specific users who work in building design, architecture, building energy management systems, modelling renewable energy conversion systems, or others interested in this kind of projected weather data. These meteorological data are produced from the regional climate model MAR (Modèle Atmosphérique Régional in French) simulations. This regional model, adapted and validated over Belgium, is forced firstly, by the ERA5 reanalysis, which represents the closest climate to reality and secondly, by three Earth system models (ESMs) from the Sixth Coupled Model Intercomparison Project database, namely, BCC-CSM2-MR, MPI-ESM.1.2, and MIROC6. The main advantage of using the MAR model is that the generated weather data have a high resolution (hourly data and 5 km) and are spatially and temporally homogeneous. The generated weather data follow two protocols. On the one hand, the Typical Meteorological Year (TMY) and eXtreme Meteorological Year (XMY) files are generated largely inspired by the method proposed by the standard ISO15927-4, allowing the reconstruction of typical and extreme years, while keeping a plausible variability of the meteorological data. On the other hand, the heatwave event (HWE) meteorological data are generated according to a method used to detect the heatwave events and to classify them according to three criteria of the heatwave (the most intense, the longest duration, and the highest temperature). All generated weather data are freely available on the open online repository Zenodo (https://doi.org/10.5281/zenodo.5606983, Doutreloup and Fettweis, 2021) and these data are produced within the framework of the research project OCCuPANt (https://www.occupant.uliege.be/ (last access: 24 June 2022)​​​​​​​ – ULiège).

Lightweight ID-based broadcast signcryption for cloud–fog-assisted IoT

Balancing the security and efficiency of the internet of things (IoT) is difficult due to its numerous heterogeneous devices, different trust levels of stakeholders, and asymmetric resource distribution among layers. Identity-based broadcast encryption (IBBE) is a promising technique to ensure data confidentiality, simplify the certification management issue, and achieve one-to-many data sharing while reducing overheads simultaneously. However, broadcast encryption (BE) schemes without source authentication are not applicable in the real world as they cannot prevent invalid messages from malicious users. Meanwhile, the computation cost of IBBE is heavy and not suitable for constrained devices. This paper proposes a lightweight ID-based broadcast signcryption scheme (LIBBSC), which takes advantage of fog nodes to temporarily store ciphertext and to partially designcrypt for end users. A fog node transforms the ciphertext based on the specific user’s public identity key and then sends the outdesigncrypted ciphertext to the end-user (this step transforms the broadcast to the one-to-one communication). Finally, only the designated end-user with the associated secret key can rapidly designcrypt with minimal computational cost. Our scheme is selective-ID secure against chosen-ciphertext attacks in the random oracle model. And the implementation and evaluation of the proposed scheme show that it is efficient and suitable for IoT devices.

Preference-Driven Texture Modeling Through Interactive Generation and Search.

Data-driven texture modeling and rendering has pushed the limit of realism in haptics. However, the lack of haptic texture databases, difficulties of model interpolation and expansion, and the complexity of real textures prevent data-driven methods from capturing a large variety of textures and from customizing models to suit specific output hardware or user needs. This work proposes an interactive texture generation and search framework driven by user input. We design a GAN-based texture model generator, which can create a wide range of texture models using Auto-Regressive processes. Our interactive texture search method, which we call "preference-driven," follows an evolutionary strategy given guidance from user's preferred feedback within a set of generated texture models. We implemented this framework on a 3D haptic device and conducted a two-phase user study to evaluate the efficiency and accuracy of our method for previously unmodeled textures. The results showed that by comparing the feel of real and generated virtual textures, users can follow an evolutionary process to efficiently find a virtual texture model that matched or exceeded the realism of a data-driven model. Furthermore, for 4 out of 5 real textures, ≥ 80% of the preference-driven models from participants were rated comparable to the data-driven models.

DAC Stacking: A Deep Learning Ensemble to Classify Anxiety, Depression, and Their Comorbidity From Reddit Texts.

Depression is the most incapacitating disease worldwide, and it has an alarming comorbidity rate with anxiety. The use of social networks to expose personal difficulties has enabled works on the automatic identification of specific mental conditions, particularly depression. In spite of many solutions proposed for the automatic recognition of depression, fewer exist for anxiety and its comorbidity with depression. In this paper, we propose DAC Stacking, a solution that leverages stacking ensembles and Deep Learning (DL) to automatically identify depression, anxiety, and their comorbidity, using data extracted from Reddit. The stacking is composed of single-label binary classifiers, that either distinguish between specific disorders and control users (experts), or between pairs of target conditions (differentiating). A meta-learner explores these base classifiers as a context for reaching a multi-label decision. We assessed alternative ensemble topologies, exploring roles for base models, DL architectures, and word embeddings. All base classifiers and ensembles outperformed the baselines for depression and anxiety (f-measures near 0.79). The ensemble topology with the best performance (Hamming Loss of 0.29 and Exact Match Ratio of 0.46) combines base classifiers of three DL architectures, and includes expert and differentiating base models. The analysis of the influential classification features according to SHAP revealed the strengths of our solution and provided insights on the challenges for the automatic classification of the addressed mental conditions.