Domain-Specific Keyword Extraction Using Joint Modeling of Local and Global Contextual Semantics

Domain-specific keyword extraction is a vital task in the field of text mining. There are various research tasks, such as spam e-mail classification, abusive language detection, sentiment analysis, and emotion mining, where a set of domain-specific keywords (aka lexicon) is highly effective. Existing works for keyword extraction list all keywords rather than domain-specific keywords from a document corpus. Moreover, most of the existing approaches perform well on formal document corpuses but fail on noisy and informal user-generated content in online social media. In this article, we present a hybrid approach by jointly modeling the local and global contextual semantics of words, utilizing the strength of distributional word representation and contrasting-domain corpus for domain-specific keyword extraction. Starting with a seed set of a few domain-specific keywords, we model the text corpus as a weighted word-graph. In this graph, the initial weight of a node (word) represents its semantic association with the target domain calculated as a linear combination of three semantic association metrics, and the weight of an edge connecting a pair of nodes represents the co-occurrence count of the respective words. Thereafter, a modified PageRank method is applied to the word-graph to identify the most relevant words for expanding the initial set of domain-specific keywords. We evaluate our method over both formal and informal text corpuses (comprising six datasets), and show that it performs significantly better in comparison to state-of-the-art methods. Furthermore, we generalize our approach to handle the language-agnostic case, and show that it outperforms existing language-agnostic approaches.

Graph Neural Collaborative Topic Model for Citation Recommendation

Due to the overload of published scientific articles, citation recommendation has long been a critical research problem for automatically recommending the most relevant citations of given articles. Relational topic models (RTMs) have shown promise on citation prediction via joint modeling of document contents and citations. However, existing RTMs can only capture pairwise or direct (first-order) citation relationships among documents. The indirect (high-order) citation links have been explored in graph neural network–based methods, but these methods suffer from the well-known explainability problem. In this article, we propose a model called Graph Neural Collaborative Topic Model that takes advantage of both relational topic models and graph neural networks to capture high-order citation relationships and to have higher explainability due to the latent topic semantic structure. Experiments on three real-world citation datasets show that our model outperforms several competitive baseline methods on citation recommendation. In addition, we show that our approach can learn better topics than the existing approaches. The recommendation results can be well explained by the underlying topics.

Development of a simulation dump truck dynamics model to estimate loads acting on its carrier and load platform when loading and dumping bulk loads

It is important for mining dump trucks to minimize the weight of the carrier and the load platform while maintaining a sufficient level of their rigidness and strength. This requirement significantly affects the weight of the transported material, the cost of transportation and, consequently, the economic efficiency of mining operations. Processes of loading and dumping of bulk loads, which is transported by dump trucks, make a significant contribution to reducing the service life of the carrier. Therefore, proper consideration of the bulk load dynamics is an important and relevant task. Contemporary systems for calculating the dynamics of solids allow for joint modeling with applications designed to calculate the dispersed body dynamics. This approach helps to obtain adequate loads in the pivots and force links of the model, to analyze the loading of the load platform, to asses the durability of the dump truck elements, to define the geometry of the load platform. In order to perform the simulation, it is required to develop a mathematical model of a dump truck, including all its key elements and subsystems, a model of the bulk load, and a model of the load platform. The purpose of the study is to develop a mathematical model of a mine dump truck to determine the loads in the pivots and force links connected to the carrier and the load platform for the strength calculations and durability analysis. The calculations are made with the combined use of the solids dynamics calculation system and the application to calculate the dynamics of dispersed bodies.

Jointly modeling marine species to inform the effects of environmental change on an ecological community in the Northwest Atlantic

Single species distribution models (SSDMs) are typically used to understand and predict the distribution and abundance of marine fish by fitting distribution models for each species independently to a combination of abiotic environmental variables. However, species abundances and distributions are influenced by abiotic environmental preferences as well as biotic dependencies such as interspecific competition and predation. When species interact, a joint species distribution model (JSDM) will allow for valid inference of environmental effects. We built a joint species distribution model of marine fish and invertebrates of the Northeast US Continental Shelf, providing evidence on species relationships with the environment as well as the likelihood of species to covary. Predictive performance is similar to SSDMs but the Bayesian joint modeling approach provides two main advantages over single species modeling: (1) the JSDM directly estimates the significance of environmental effects; and (2) predicted species richness accounts for species dependencies. An additional value of JSDMs is that the conditional prediction of species distributions can use not only the environmental associations of species, but also the presence and abundance of other species when forecasting future climatic associations.

Longitudinal Relationship of Liver Injury with Inflammation Biomarkers in Covid-19 Hospitalized Patients Using a Joint Modeling Approach

Background: The mechanisms underlying liver disease in patients with COVID-19 are not entirely known.Aim: To investigate, by means of novel statistical techniques, the changes over time in the relationship between inflammation markers and liver damage markers in relation to survival in COVID-19.Methods: The study included 221 consecutive patients admitted to the hospital during the first COVID-19 wave in Spain. Generalized additive mixed models were used to investigate the influence of time and inflammation markers on liver damage markers in relation to survival. Joint modeling regression was used to evaluate the temporal correlations between inflammation markers (serum C-reactive protein [CRP], interleukin-6, plasma D-dimer, and blood lymphocyte count) and liver damage markers, after adjusting for age, sex, and therapy.Results: The patients who died showed a significant elevation in serum aspartate transaminase (AST) and alkaline phosphatase levels over time. Conversely, a decrease in serum AST levels was observed in the survivors, who showed a negative correlation between inflammation markers and liver damage markers (CRP with serum AST, alanine transaminase [ALT], and gamma-glutamyl transferase [GGT]; and D-dimer with AST and ALT) after a week of hospitalization. Conversely, most correlations were positive in the patients who died, except lymphocyte count, which was negatively correlated with AST, GGT, and alkaline phosphatase. These correlations were attenuated with age. Conclusion: The patients who died during COVID-19 infection displayed a significant elevation of liver damage markers, which is correlated with inflammation markers over time. These results are consistent with the role of systemic inflammation in liver damage during COVID-19.

Associations between Visual Acuity and Cognitive Decline in Older Adulthood: A 9-Year Longitudinal Study

Objective: Emerging evidence suggests low vision may be a modifiable risk factor for cognitive decline. We examined effects of baseline visual acuity (VA) on level of, and change in, cognitive test performance over 9 years. Method: A population-based sample of 1,621 participants (average age 77 years) completed a comprehensive neuropsychological evaluation and VA testing at baseline and reassessed at nine subsequent annual visits. Linear regression modeled the association between baseline VA and concurrent cognitive test performance. Joint modeling of a longitudinal sub-model and a survival sub-model to adjust for attrition were used to examine associations between baseline VA and repeated cognitive test performance over time. Results: Better baseline VA was associated cross-sectionally with younger age, male sex, greater than high school education, and higher baseline neuropsychological test scores on both vision-dependent (B coefficient range −0.163 to −0.375, p = .006 to <.001) and vision-independent tests (−0.187 to −0.215, p = .003 to .002). In longitudinal modeling, better baseline VA was associated with slower decline in vision-dependent tests (B coefficient range −0.092 to 0.111, p = .005 to <.001) and vision-independent tests (−0.107 to 0.067, p = .007 to <.001). Conclusions: Higher VA is associated with higher concurrent cognitive abilities and slower rates of decline over 9 years in both vision-dependent and vision-independent tests of memory, language, and executive functioning. Findings are consistent with emerging literature supporting vision impairment in aging as a potentially modifiable risk factor for cognitive decline. Clinicians should encourage patient utilization of vision assessment and correction with the added aim of protecting cognition.

Lateral Stability Performance of Articulated Narrow-Track Tractors

A tractor losing lateral stability starts to rollover. It is a matter of fact that tractor lateral rollover accidents are one of the most frequent causes of death and injuries for farmers. Consequently, tractors are fitted with a specific protective structure to minimize the consequences for the driver during the rollover (ROPS). The narrow-track tractor, designed to operate in vineyards and orchards, is a tractor category with a very narrow track width and the risk of rollover is higher. The aim of the study was to evaluate the compact narrow-track tractor types commercially available, designed to mount a cantilever engine in the forward position with effects on the Center of Gravity (CoG) because more than 50% of the tractor weight is loaded on the front axle, and, specifically, the articulated narrow-track tractors where the stability is affected by the pivot point connecting the two tractor bodies. As a consequence of the typical tractor design of articulated tractors, during the steering action the line passing through the front and rear tire contact points on the ground changes, influencing the tractor’s stability. The approach of the research was based on reproducing the lateral stability tractor condition by developing a kinematic model, with the goal to virtually simulate the tractor behavior and to calculate the lateral stability angle for articulated tractors. The innovative contribution of this paper was the tractor articulation joint modeling, assuming a virtual pivot point to reproduce two relatives’ rotations between the front and rear bodies of the tractor: vertical (yaw angle) and longitudinal (roll angle) rotations. The lowest value of the stability angle was 39.3°, measured at −35° yaw angle. The model at the tractor design stage will allow adjusting of the tractor parameters to improve the lateral stability performance.