Building a 'Virtual Library': continuing a global collaboration to strengthen research capacity within Nepal and other low- and middle-income countries

To fill the gap in health research capacity-building efforts, we created the ‘Virtual Library’ (VL) – a web-based repository of context-relevant resources for health researchers in low- and middle-income countries (LMICs). This paper describes the participatory process used to systematically develop the VL, and describes how our interprofessional team – representing both an LMIC (Nepal) and a high-income country (HIC) (USA, US) – engaged in shared meaning-making. A team of researchers and clinicians representing a range of subdisciplines from Nepal and the US created a replicable search strategy and standardized Resource Screening Guide (RSG) to systematically assess resources to be included within the VL. Descriptive methods were used to summarize findings from the RSG and lessons learned from the collaborative process. Collectively, 14 team members reviewed 564 potential resources (mean = 40, SD = 22.7). Mean RSG score was 7.02/10 (SD = 2). More than 76% of resources met each of the four quality criteria (relevant; reputable, accessible; understandable). Within the published VL, 298 resources were included, organized by 15 topics and 45 sub-topics. Of these, 223 resources were evaluated by the RSG; 75 were identified by team member expertise. The collaborative process involved regular meetings, iterative document revisions, and peer review. Resource quality was better than expected, perhaps because best practices/principles related to health research are universally relevant, regardless of context. While the RSG was essential to systematize our search and ensure reproducibility, team member expertise was valuable. Pairing team members during peer-review led to bi-directional knowledge sharing and was particularly successful. This work reflects a highly collaborative global partnership and offers a model for future health research capacity-building efforts. We invite engagement with the Virtual Library <https://lmicresearch.org> as one supportive pillar of infrastructure to develop individual and institutional research capacity.

Small world of the miRNA science drives its publication dynamics

Many scientific articles became available in the digital form which allows for querying articles data, and specifically the automated metadata gathering, which includes the affiliation data. This in turn can be used in the quantitative characterization of the scientific field, such as organizations identification, and analysis of the co-authorship graph of those organizations to extract the underlying structure of science. In our work, we focus on the miRNA science field, building the organization co-authorship network to provide the higher-level analysis of scientific community evolution rather than analyzing author-level characteristics. To tackle the problem of the institution name writing variability, we proposed the k-mer/n-gram boolean feature vector sorting algorithm, KOFER in short. This approach utilizes the fact that the contents of the affiliation are rather consistent for the same organization, and to account for writing errors and other organization name variations within the affiliation metadata field, it converts the organization mention within the affiliation to the K-Mer (n-gram) Boolean presence vector. Those vectors for all affiliations in the dataset are further lexicographically sorted, forming groups of organization mentions. With that approach, we clustered the miRNA field affiliation dataset and extracted unique organization names, which allowed us to build the co-authorship graph on the organization level. Using this graph, we show that the growth of the miRNA field is governed by the small-world architecture of the scientific institution network and experiences power-law growth with exponent 2.64 ± 0.23 for organization number, in accordance with network diameter, proposing the growth model for emerging scientific fields. The first miRNA publication rate of an organization interacting with already publishing organization is estimated as 0.184 ± 0.002 year-1.

Aggregating large-scale databases for PubMed author name disambiguation

OBJECTIVE

PubMed has suffered from the author ambiguity problem for many years. Existing studies on author name disambiguation (AND) for PubMed only used internal metadata for development. However, some of them are incomplete (eg, a large number of names are only abbreviated and their full names are not available) or less discriminative. To this end, we present a new disambiguation method, namely AggAND, by aggregating information from external databases.

MATERIALS AND METHODS

We address this issue by exploring Microsoft Academic Graph, Semantic Scholar, and PubMed Knowledge Graph to enhance the built-in name metadata, and extend the internal metadata with some external and more discriminative metadata.

RESULTS

Experimental results on enhanced name metadata demonstrate comparable performance to 3 author identifier systems, as well as show superiority over the original name metadata. More importantly, our method, AggAND, incorporating both enhanced name and extended metadata, yields F1 scores of 95.80% and 93.71% on 2 datasets and outperforms the state-of-the-art method by a large margin (3.61% and 6.55%, respectively).

CONCLUSIONS

The feasibility and good performance of our methods not only help better understand the importance of external databases for disambiguation, but also point to a promising direction for future AND studies in which information aggregated from multiple bibliographic databases can be effective in improving disambiguation performance. The methodology shown here can be generalized to broader bibliographic databases beyond PubMed. Our code and data are available online (https://github.com/carmanzhang/PubMed-AND-method).

Large Scale Subject Category Classification of Scholarly Papers With Deep Attentive Neural Networks

Subject categories of scholarly papers generally refer to the knowledge domain(s) to which the papers belong, examples being computer science or physics. Subject category classification is a prerequisite for bibliometric studies, organizing scientific publications for domain knowledge extraction, and facilitating faceted searches for digital library search engines. Unfortunately, many academic papers do not have such information as part of their metadata. Most existing methods for solving this task focus on unsupervised learning that often relies on citation networks. However, a complete list of papers citing the current paper may not be readily available. In particular, new papers that have few or no citations cannot be classified using such methods. Here, we propose a deep attentive neural network (DANN) that classifies scholarly papers using only their abstracts. The network is trained using nine million abstracts from Web of Science (WoS). We also use the WoS schema that covers 104 subject categories. The proposed network consists of two bi-directional recurrent neural networks followed by an attention layer. We compare our model against baselines by varying the architecture and text representation. Our best model achieves micro-F1 measure of 0.76 with F1 of individual subject categories ranging from 0.50 to 0.95. The results showed the importance of retraining word embedding models to maximize the vocabulary overlap and the effectiveness of the attention mechanism. The combination of word vectors with TFIDF outperforms character and sentence level embedding models. We discuss imbalanced samples and overlapping categories and suggest possible strategies for mitigation. We also determine the subject category distribution in CiteSeerX by classifying a random sample of one million academic papers.

A Digital Library for Increasing Awareness About Living Donor Kidney Transplants: Formative Study

Background

It is not common for people to come across a living kidney donor, let alone consider whether they would ever donate a kidney themselves while they are alive. Narrative storytelling, the sharing of first-person narratives based on lived experience, may be an important way to improve education about living donor kidney transplants (LDKTs). Developing ways to easily standardize and disseminate diverse living donor stories using digital technology could inspire more people to consider becoming living donors and reduce the kidney shortage nationally.

Objective

This paper aimed to describe the development of the Living Donation Storytelling Project, a web-based digital library of living donation narratives from multiple audiences using video capture technology. Specifically, we aimed to describe the theoretical foundation and development of the library, a protocol to capture diverse storytellers, the characteristics and experiences of participating storytellers, and the frequency with which any ethical concerns about the content being shared emerged.

Methods

This study invited kidney transplant recipients who had received LDKTs, living donors, family members, and patients seeking LDKTs to record personal stories using video capture technology by answering a series of guided prompts on their computer or smartphone and answering questions about their filming experience. The digital software automatically spliced responses to open-ended prompts, creating a seamless story available for uploading to a web-based library and posting to social media. Each story was reviewed by a transplant professional for the disclosure of protected health information (PHI), pressuring others to donate, and medical inaccuracies. Disclosures were edited.

Results

This study recruited diverse storytellers through social media, support groups, churches, and transplant programs. Of the 137 storytellers who completed the postsurvey, 105/137 (76.6%) were white and 99/137 (72.2%) were female. They spent 62.5 min, on average, recording their story, with a final median story length of 10 min (00:46 seconds to 32:16 min). A total of 94.8% (130/137) of storytellers were motivated by a desire to educate the public; 78.1% (107/137) were motivated to help more people become living donors; and 75.9% (104/137) were motivated to dispel myths. The ease of using the technology and telling their story varied, with the fear of being on film, emotional difficulty talking about their experiences, and some technological barriers being reported. PHI, most commonly surnames and transplant center names, was present in 62.9% (85/135) of stories and was edited out.

Conclusions

With appropriate sensitivity to ensure diverse recruitment, ethical review of content, and support for storytellers, web-based storytelling platforms may be a cost-effective and convenient way to further engage patients and increase the curiosity of the public in learning more about the possibility of becoming living donors.

12-Year Use of a Digital Reference Library (VitalBook) at a U.S. Dental School: Students' and Alumni Perceptions

Digital textbooks are being used to reduce production and storage costs of printed copies, enhance usage, and include search capabilities, but the use of digital texts is not universally accepted. In 2001, the New York University College of Dentistry introduced a digital reference library, the VitalBook. Beginning in 2005, the college annually surveyed senior students and, from 2012, also surveyed alumni on their opinions and extent of use of the VitalBook. The aim of this study was to evaluate 12 years of students' perspectives and three years of alumni perspectives on the value of the VitalBook to their dental educational experience. Students were asked how frequently they used the VitalBook, if it was a good investment, if they would use it after graduation, and if they would recommend it to others. Alumni were asked the last three questions. This study reports the results from 4,105 students over 12 years (average response rate 95.3%) and 184 alumni over three years (average response rate 17.4%). The results indicated that students used the VitalBook on average 24% of their study time, but they were split regarding the other questions. The majority opinion in 2005 was negative on all questions. These opinions shifted to become more favorable to a peak in 2010, but declined since then to a more negative overall view of the VitalBook. A split opinion among students continued through 2016, with fewer recommending it although more considered it a good investment with plans to use it after graduation. Alumni mirrored their responses as students. These results suggest that, as more flexible and dynamic digitized reference systems emerge, the use of student-paid traditional digitized textbooks may become an even less favored choice.

Transforming Education Research Through Open Video Data Sharing

Open data sharing promises to accelerate the pace of discovery in the developmental and learning sciences, but significant technical, policy, and cultural barriers have limited its adoption. As a result, most research on learning and development remains shrouded in a culture of isolation. Data sharing is the rare exception (Gilmore, 2016). Many researchers who study teaching and learning in classroom, laboratory, museum, and home contexts use video as a primary source of raw research data. Unlike other measures, video captures the complexity, richness, and diversity of behavior. Moreover, because video is self-documenting, it presents significant potential for reuse. However, the potential for reuse goes largely unrealized because videos are rarely shared. Research videos contain information about participants' identities making the materials challenging to share. The large size of video files, diversity of formats, and incompatible software tools pose technical challenges. The Databrary (databrary.org) digital library enables researchers who study learning and development to store, share, stream, and annotate videos. In this article, we describe how Databrary has overcome barriers to sharing research videos and associated data and metadata. Databrary has developed solutions for respecting participants' privacy; for storing, streaming, and sharing videos; and for managing videos and associated metadata. The Databrary experience suggests ways that videos and other identifiable data collected in the context of educational research might be shared. Open data sharing enabled by Databrary can serve as a catalyst for a truly multidisciplinary science of learning.

PhysiomeSpace: digital library service for biomedical data

Every research laboratory has a wealth of biomedical data locked up, which, if shared with other experts, could dramatically improve biomedical and healthcare research. With the PhysiomeSpace service, it is now possible with a few clicks to share with selected users biomedical data in an easy, controlled and safe way. The digital library service is managed using a client-server approach. The client application is used to import, fuse and enrich the data information according to the PhysiomeSpace resource ontology and upload/download the data to the library. The server services are hosted on the Biomed Town community portal, where through a web interface, the user can complete the metadata curation and share and/or publish the data resources. A search service capitalizes on the domain ontology and on the enrichment of metadata for each resource, providing a powerful discovery environment. Once the users have found the data resources they are interested in, they can add them to their basket, following a metaphor popular in e-commerce web sites. When all the necessary resources have been selected, the user can download the basket contents into the client application. The digital library service is now in beta and open to the biomedical research community.