How to use scite to help with your research (scite webinar, Sep 29, 2021)
TLDRThe webinar 'How to use scite to help with your research' introduces scite, a tool aimed at enhancing the reliability of research by analyzing citation context. It demonstrates how scite can be utilized to examine individual articles, researchers, and the next generation of citations beyond traditional metrics. The tool uses machine learning to classify citation statements as supporting, mentioning, or contrasting evidence. The presentation also covers the browser extension, reference check, and the new citation statement search, emphasizing scite's potential to improve research reliability and interpretation.
Takeaways
- 😀 The webinar introduces 'Scite', a tool designed to enhance the reliability and interpretability of research by providing detailed citation information.
- 📈 Scite offers a browser extension for various browsers that displays citation information right beside research articles online, allowing for immediate context on how work has been cited.
- 🔍 Scite uses machine learning to classify citation statements into supporting, mentioning, or contrasting evidence, providing a deeper understanding of the research's impact.
- 🤖 The platform has partnered with over a dozen publishers to access content not freely available and uses an open-source tool called Grobid to extract in-text citations from articles.
- 🔗 Scite's citation information is linked by DOI, allowing users to easily navigate back to the full text of the articles and explore their metadata, abstracts, and citation contexts.
- 📚 The tool provides a new way to visualize citation networks, showing not just the number of citations but the relationships and contexts in which research is cited.
- 👥 Scite also offers features to analyze citations at the researcher, journal, and affiliation levels, providing dashboards for a comprehensive view of citation patterns.
- 🗂️ Users can create custom dashboards based on topics of interest, DOI lists, or library syncs, and set email alerts for new citations or publications.
- 🔎 Scite has a 'Reference Check' tool that helps authors, editors, and peer reviewers analyze the references of a manuscript for editorial concerns or citation patterns.
- 📘 The platform is designed to work with a wide range of disciplines, from STEM to social sciences and humanities, although the value of citation types may vary by field.
- 🔑 The webinar concludes by acknowledging the historical context of Scite's approach, citing the work of Eugene Garfield, and highlighting the importance of community feedback for ongoing validation and improvement.
Q & A
What is the purpose of the Scite tool as described in the webinar?
-Scite is designed to make research more reliable by providing a deeper understanding of how research articles have been cited. It goes beyond just citation counts and offers insights into the context and nature of citations, helping users to better interpret and utilize research findings.
How does Scite classify citation statements into supporting, mentioning, or contrasting?
-Scite uses a deep learning model that has been trained on approximately 50,000 citation statements annotated by experts. The model classifies each citation statement as providing supporting evidence, merely mentioning the cited work, or providing contrasting evidence.
What is the browser extension developed by Scite, and what does it show?
-Scite has developed a browser extension available for Chrome, Firefox, Safari, and Microsoft Edge. This extension displays Scite's citation information virtually anywhere a research article is read online. It shows the number of paper citations and a breakdown of citation statements into supporting, mentioning, and contrasting categories.
How does Scite access full-text articles to extract citation statements?
-Scite accesses full-text articles through open access sources like PubMed Central and partnerships with over a dozen publishers. The content is generally received in PDF or XML formats, which Scite processes using an open-source tool called Grobid to extract in-text citations and surrounding text.
What is the Scite Index, and how is it different from traditional impact factors?
-The Scite Index is a metric that measures the ratio of supporting citations to the total number of supporting plus contrasting citations for a given journal. Unlike traditional impact factors that focus on citation counts, the Scite Index provides a view of the sentiment of citations, indicating whether a journal's articles are generally supported or contrasted in the literature.
How can Scite be used to analyze citations for individual researchers?
-Scite allows users to view individual researcher profiles to see how their work has been cited, including the number of citations, the types of citations (supporting, mentioning, contrasting), and other metrics such as open access article counts or first-to-cite status. This helps researchers understand their citation patterns and impact on the field.
What is the Reference Check tool, and how does it help authors and editors?
-The Reference Check tool is a feature that allows users to upload a PDF or DOCX document, and it automatically identifies in-text citations, matches them to references, and checks them against Scite's database for citation context and editorial information. This helps authors and editors to ensure that the references they cite are reliable and relevant, and it can flag any potential issues such as retracted articles.
How does Scite handle cases where it cannot reliably match citation statements to subject areas?
-When Scite cannot reliably match citation statements to subject areas, it categorizes them under a general 'Other' section. This ensures that users still have access to citation information, even if it's not filtered by specific subject areas.
What is the process for Scite to extract and classify citation statements from full-text articles?
-Scite's process involves using the Grobid tool to extract in-text citations and surrounding text from full-text articles. A deep learning model then classifies the extracted citation statements based on their context as supporting, mentioning, or contrasting. The model's accuracy is regularly validated and improved through user feedback and expert review.
How can Scite be used to create custom dashboards for specific topics or groups of articles?
-Users can create custom dashboards on Scite by searching for specific topics or uploading a list of DOIs for a group of articles. Scite then aggregates the citation data for these articles, allowing users to explore the citation context, authors, and other metrics in a customized view.
Outlines
📚 Introduction to the Webinar and Site's Purpose
Josh Nicholson, the co-founder and CEO of Site, introduces a webinar aimed at discussing the importance of research reliability and the role of Site in enhancing it. He emphasizes the impact of scientific research, the challenges of information overload, and the lack of reproducibility in studies. Site is presented as a tool to make research more reliable by providing a deeper understanding of citations and their context. The webinar offers a special discount for participants and encourages interaction through questions and feedback.
🔍 Deep Dive into Site's Citation Analysis Features
This paragraph explains how Site uses a browser extension to provide citation information while reading research articles online. It discusses the breakdown of citation statements into supporting, mentioning, and contrasting evidence. The process involves machine learning to classify these statements and partnerships with publishers for full-text access. The paragraph also covers the technical challenges of extracting in-text citations and the use of the Grobid tool for this purpose.
🤖 Machine Learning Models for Citation Context
Josh describes the use of a deep learning model to classify citation statements into supporting, mentioning, or contrasting categories. The model was trained on a dataset of 50,000 citation statements annotated by experts. The accuracy is tested using a holdout set of 10,000 citation statements. The challenges of scientific writing and the limitations of the model are acknowledged, but the value of citation context is emphasized for understanding research.
📈 Visualization and Analysis of Citation Networks
The paragraph discusses the visualization of citation networks to understand the relationships between papers and the significance of node sizes representing citation counts. It explains the ability to filter citations and explore the conversation among papers, providing a more nuanced view of research support and contrast within the academic community.
🏫 Insights into Institutional and Journal Citations
Josh presents the features of Site that allow users to analyze citations at the institutional and journal levels. He discusses the dashboards available for universities, journals, and other affiliations, providing insights into citation patterns, editorial notices, and the overall support or contrast of published works within these entities.
🎯 Targeted Topic Analysis and Custom Dashboards
This section covers how Site can be used to analyze specific topics, such as 'chromosome segregation,' by searching for the most supported research and filtering results based on various criteria. Users can create custom dashboards to aggregate and monitor information on topics of interest, enhancing the ability to stay updated through email alerts.
📝 Reference Check Tool for Manuscripts
Josh introduces the Reference Check tool designed to help authors, editors, and peer reviewers analyze the references of a manuscript. The tool identifies in-text citations and matches them against metadata and citation information to provide insights into how referenced works have been cited in the literature, including any editorial concerns such as retractions.
📊 Representation of Social Sciences and Humanities
The discussion addresses the representation of social sciences and humanities in Site's data, comparing the incidence of supporting and contrasting citations across different fields. Josh explains that while the value of citation context is consistent across disciplines, the utility of citation types may vary, and ongoing research is being conducted to analyze this further.
🗂 New Citation Statement Search Feature
Josh introduces a new feature that allows users to search citation statements across the entire Site database, providing a powerful method to find expert analyses and opinions on a wide range of topics. This search capability is not limited to academic articles but can also be applied to content from sources like The New York Times, enhancing accessibility to research insights.
👥 Acknowledging the Pioneers of Citation Analysis
In conclusion, Josh acknowledges the pioneers in the field of citation analysis, such as Eugene Garfield, who envisioned the automation of citation indexing. He emphasizes the importance of building upon previous work and highlights the technical and content challenges overcome in developing Site's capabilities.
🤝 Final Q&A and Closing Remarks
The final paragraph wraps up the webinar with an open invitation for questions and feedback. Josh and Dom encourage participants to reach out via email for any follow-up questions and emphasize the team's commitment to continuous improvement based on user input. They express gratitude for the participants' time and interest.
Mindmap
Keywords
💡scite
💡research
💡reproducibility
💡citation
💡deep learning model
💡bibliometrics
💡browser extension
💡open access
💡publishers
💡retractions
💡reference check
Highlights
Introduction to Scite as a tool for making research more reliable through citation analysis.
Scite's browser extension that provides citation information while reading research articles online.
The importance of understanding citation context beyond just the number of citations.
How Scite classifies citation statements into supporting, mentioning, or contrasting evidence.
The use of machine learning to analyze over 900 million citation statements at scale.
Partnerships with publishers to access content that is not freely available.
Utilization of the open-source tool Grobid for extracting in-text citations from articles.
The ability to filter citations by year, section of the citing paper, and type of publication.
Visualizing citation networks to understand the conversation among different research papers.
Scite's Reference Check tool for evaluating the references of a manuscript pre-publication.
Integration of Scite with major submission systems to prevent citation of retracted articles.
The potential for Scite to analyze institutions and universities based on citation data.
Scite's new citation statement search for finding expert analyses and interpretations on various topics.
Acknowledgment of previous work in the field that inspired the development of Scite.
The ongoing process of validation and feedback from the research community to improve Scite's classification accuracy.
Scite's plans for future development based on user feedback and needs.
Closing remarks encouraging users to reach out with questions, feedback, or suggestions for improvement.