Volume 7 (2022): Edition 2 (April 2022)

Given the information overload of scientific literature, there is an increasing need for computable biomedical knowledge buried in free text. This study aimed to develop a novel approach to extracting and measuring uncertain biomedical knowledge from scientific statements.

Taking cardiovascular research publications in China as a sample, we extracted subject–predicate–object triples (SPO triples) as knowledge units and unknown/hedging/conflicting uncertainties as the knowledge context. We introduced information entropy (IE) as potential metric to quantify the uncertainty of epistemic status of scientific knowledge represented at subject-object pairs (SO pairs) levels.

The results indicated an extraordinary growth of cardiovascular publications in China while only a modest growth of the novel SPO triples. After evaluating the uncertainty of biomedical knowledge with IE, we identified the Top 10 SO pairs with highest IE, which implied the epistemic status pluralism. Visual presentation of the SO pairs overlaid with uncertainty provided a comprehensive overview of clusters of biomedical knowledge and contending topics in cardiovascular research.

The current methods didn’t distinguish the specificity and probabilities of uncertainty cue words. The number of sentences surrounding a given triple may also influence the value of IE.

Our approach identified major uncertain knowledge areas such as diagnostic biomarkers, genetic polymorphism and co-existing risk factors related to cardiovascular diseases in China. These areas are suggested to be prioritized; new hypotheses need to be verified, while disputes, conflicts, and contradictions need to be settled.

We provided a novel approach by combining natural language processing and computational linguistics with informetric methods to extract and measure uncertain knowledge from scientific statements.

Performers may generate loyalty partly through eliciting illusory personal connections with their audience, parasocial relationships (PSRs), and individual illusory exchanges, parasocial interactions (PSIs). On social media, semi-PSIs are real but imbalanced exchanges with audiences, including through comments on influencers’ videos, and strong semi-PSIs are those that occur within PSRs. This article introduces and assesses an automatic method to detect videos with strong PSI potential.

Strong semi-PSIs were hypothesized to occur when commenters used a variant of the pronoun “you”, typically addressing the influencer. Comments on the videos of UK female influencer channels were used to test whether the proportion of you pronoun comments could be an automated indicator of strong PSI potential, and to find factors associating with the strong PSI potential of influencer videos. The highest and lowest strong PSI potential videos for 117 influencers were classified with content analysis for strong PSI potential and evidence of factors that might elicit PSIs.

The you pronoun proportion was effective at indicating video strong PSI potential, the first automated method to detect any type of PSI. Gazing at the camera, head and shoulders framing, discussing personal issues, and focusing on the influencer associated with higher strong PSI potential for influencer videos. New social media factors found include requesting feedback and discussing the channel itself.

Only one country, genre and social media platform was analysed.

The method can be used to automatically detect YouTube videos with strong PSI potential, helping influencers to monitor their performance.

This is the first automatic method to detect any aspect of PSI or PSR.

Researchers are more likely to read and cite papers to which they have access than those that they cannot obtain. Thus, the objective of this work is to analyze the contribution of the Open Access (OA) modality to the impact of hybrid journals.

The “research articles” in the year 2017 from 200 hybrid journals in four subject areas, and the citations received by such articles in the period 2017–2020 in the Scopus database, were analyzed. The hybrid OA papers were compared with the paywalled ones. The journals were randomly selected from those with share of OA papers higher than some minimal value. More than 60 thousand research articles were analyzed in the sample, of which 24% under the OA modality.

We obtain at journal level that cites per article in both hybrid modalities (OA and paywalled) strongly correlate. However, there is no correlation between the OA prevalence and cites per article. There is OA citation advantage in 80% of hybrid journals. Moreover, the OA citation advantage is consistent across fields and held in time. We obtain an OA citation advantage of 50% in average, and higher than 37% in half of the hybrid journals. Finally, the OA citation advantage is higher in Humanities than in Science and Social Science.

Some of the citation advantage is likely due to more access allows more people to read and hence cite articles they otherwise would not. However, causation is difficult to establish and there are many possible bias. Several factors can affect the observed differences in citation rates. Funder mandates can be one of them. Funders are likely to have OA requirement, and well-funded studies are more likely to receive more citations than poorly funded studies. Another discussed factor is the selection bias postulate, which suggests that authors choose only their most impactful studies to be open access.

For hybrid journals, the open access modality is positive, in the sense that it provides a greater number of potential readers. This in turn translates into a greater number of citations and an improvement in the position of the journal in the rankings by impact factor. For researchers it is also positive because it increases the potential number of readers and citations received.

Our study refines previous results by comparing documents more similar to each other. Although it does not examine the cause of the observed citation advantage, we find that it exists in a very large sample.

Due to the incompleteness nature of knowledge graphs (KGs), the task of predicting missing links between entities becomes important. Many previous approaches are static, this posed a notable problem that all meanings of a polysemous entity share one embedding vector. This study aims to propose a polysemous embedding approach, named KG embedding under relational contexts (ContE for short), for missing link prediction.

ContE models and infers different relationship patterns by considering the context of the relationship, which is implicit in the local neighborhood of the relationship. The forward and backward impacts of the relationship in ContE are mapped to two different embedding vectors, which represent the contextual information of the relationship. Then, according to the position of the entity, the entity's polysemous representation is obtained by adding its static embedding vector to the corresponding context vector of the relationship.

ContE is a fully expressive, that is, given any ground truth over the triples, there are embedding assignments to entities and relations that can precisely separate the true triples from false ones. ContE is capable of modeling four connectivity patterns such as symmetry, antisymmetry, inversion and composition.

ContE needs to do a grid search to find best parameters to get best performance in practice, which is a time-consuming task. Sometimes, it requires longer entity vectors to get better performance than some other models.

ContE is a bilinear model, which is a quite simple model that could be applied to large-scale KGs. By considering contexts of relations, ContE can distinguish the exact meaning of an entity in different triples so that when performing compositional reasoning, it is capable to infer the connectivity patterns of relations and achieves good performance on link prediction tasks.

ContE considers the contexts of entities in terms of their positions in triples and the relationships they link to. It decomposes a relation vector into two vectors, namely, forward impact vector and backward impact vector in order to capture the relational contexts. ContE has the same low computational complexity as TransE. Therefore, it provides a new approach for contextualized knowledge graph embedding.