Nature Podcast April 4, 2021 Trained on billions of words from books, articles and websites, GPT-3 was the latest in a series of ‘large language model’ AIs that are used by companies around the world to improve search results, answer questions, or propose computer code. However, these large language models are not without their issues. Their training is based on the statistical relationships between the words and phrases, which can lead them to generate toxic or dangerous outputs. Preventing responses like these is a huge challenge for researchers, who are attempting to do so by addressing biases in training data, […]
Tag Archives: Bibliometrics
Not-for-profit publisher makes big move toward open access science
EurekAlert March 10, 2021 Canadian Science Publishing (CSP)–a not-for-profit publisher of peer-reviewed STEM journals announced a new open access publishing agreement with the University of California (UC) that will offer unlimited open access publication for UC researchers publishing with its journals. CSP is exploring how to shift from subscription-based business models to models that make it easier and more affordable for researchers to publish their work as open access. Under the cost sharing model, the UC libraries will automatically pay the first $1,000 of the article processing charge (APC) for all UC authors who choose to publish in a CSP journal. […]
Major Physics Publisher Goes Double Blind
American Physical Society September 18, 2020 To increase fairness IOP Publishing (IOPP) has announced a major shift in its peer-review methods, which, they say could offer better chances of impartial evaluation. By the end of 2021, IOPP journals will make their default peer-review option “double-blind”, neither reviewers nor authors know each other’s identities. Most scientific journals operate in single-blind mode: Reviewers know who has written the paper they are scrutinizing, but the authors do not know who the reviewers are. Knowing the authors’ identities could be useful to see a new result within the context of previous work. But this […]
Spotting cutting-edge topics in scientific research using keyword analysis
Science Daily October 24, 2019 An international team of researchers (Japan, Taiwan) used bibliometric analyses targeting data of about 30 million published articles from 1970 to 2017 on PubMed to test their hypothesis that existing emerging topics contribute to the generation of new emerging topics in that field. They collected emerging keywords from medical subject headings attached to each article and performed co-word analyses of each emerging keyword 1-year prior to it becoming an emerging keyword. About 75% of total emerging keywords, at 1-year prior to becoming identified as emerging, co-appeared with other emerging keywords in the same article. Most […]
Russian website reportedly selling science article authorships
Phys.org July 22, 2019 Several websites are reporting the entity accused of selling authorships is International Publisher. Translations made by Science Chronicle suggest the group behind the site is selling authorships on finished articles listed in Scopus—and some listed by Web of Science. There are apparently authorship guarantees and tiered pricing. 73 of the papers are to be published in India-based journals, 54 based in Venezuela, 48 in the U.S., 33 in Russia and 28 in Pakistan—the names of the journals are not given. Retractionwatch is reporting that they have found evidence of 10,000 researchers paying to have their names […]
Global Artificial Intelligence Patent Survey
Inside Big Data February 22, 2019 Corresponding to the rise of 4IR digital technologies, the number of international AI based patent filings has expanded rapidly over the last few years, mostly concentrated in the United States and Asia. According to a 2016 study, approximately 75% of all AI-related patent publications in the world come from three jurisdictions: China, Japan, and the United States. Although the majority of AI-related patents are filed in these countries, Europe is also seeing substantial increases in such patent filings…read more.
Melding of concepts from different scientific fields
Science Daily March 9, 2018 Researchers in Japan developed a new methodology to analyze citations in papers that used two specified terms and tracked the changes over time. They mapped individual papers and connected these with papers they had cited, resulting in publications citing the same paper being close to each other. They found that only a few publications were required for the fusion between agent based modeling and individual based modeling and these weren’t necessarily the most cited papers. According to them three things are needed for fusion to occur: researchers being aware of issues in different fields; common […]