Study Of Distribution Law Of Microblog Altmetrics Based On Large Scale Dataset

The economic and cultural development of the nation and society heavily rely on science and technology.Above all,scholarly communication and scientific evaluation are two fundamental elements in advancing and guiding the scientific development.The Internet was born in the academic community and,in the big data era,has profoundly influenced scientific communication as well as the scientific evaluation situation.Altmetrics,well known for its aim of measuring the full impact of scientific product,is thus increasingly playing an important role and attracting widely focus.However,the current basic research are far from enough in supporting the urgent need from application oriented research(like various evaluation researchand information retrieval).The situation requires more studies to be focused on the nature,attributes and value of altmetrics indicators and sources.Microblog altmetrics has among all the highest coverage and is a very typical type of altmetrics indicator,so it is selected as study object.This study,based on large scale dataset,has systematically and completely revealed the count distribution,author distribution,language distribution and value distribution of microblog altmetrics,in order to better understand its meaning and pattern,promote the scientific use of its data and indicators,and provide reference for future studies of other types of altmetrics indicators.Most analysis in the dissertation are based on dataset provided by Altmetric.com,including 5.18 million records of scientific product,involving over 20 million scientific tweets and 14 other types of altmetrics indicators.Bibliographic data of different time span are collected from Scopus.According to specific research question,each chapter has used slightly different dataset.For description of various distribution and law at different level,Pandas and Matplotlib packages from Python are adopted to do statistical analysis and visualization analysis.For analyzing motivation distribution,author identity and language preference of scientific tweets,content analysis that combines quantitative analysis and qualitative analysis is used.In addition,co-occurrence analysis and concept model etc.are used.The general research process is:firstly,research both at home and abroad are reviewed to propose the research questions(Chapter 0);secondly,related theories are introduced and concept models are constructed to provide basis for understanding the research questions(Chapter 1);thirdly,empirical studies are done to answer these research questions(Chapter 2,Chapter 3,Chapter 4 and Chapter 5);fourthly,conclusions of the whole research are summarized(Chapter 6).To be specific,core content of the dissertation are divided into five parts:Chapter 1 summarizes related theories of microblog altmetrics and constructs several concept models.Definition of altmetrics is given and distinguished from social media metrics and article level metrics.Relationship between altmetrics and big data is also presented.Several core concepts like microblog altmetrics indicators,scientific tweets and scientific Sina Weibo are identified and defined.Theories from citation analysis and sociology are introduced as theoretic basis.Concept model is constructed to demonstrate the inner relationship between the four studies of distribution law and the research logic of each study of distribution law.Chapter 2 explores the count distribution law of microblog altmetrics with empirical study.Relative coverage of microblog altmetrics is calculated and compared with other altmetrics indicators;Immediacy distribution is measured and proves that microblog altmetrics has significant immediacy advantage over citation indicators;distribution of microblog altmetrics in article level,source level and discipline level are revealed.The skewness turns out to be even higher than citation indicators.Disciplines with the highest attention are identified;Altmetric attention score is compared and articles with scientific weibo are found to have much higher attention.Chapter 3 explores author distribution law of microblog altmetrics with empirical study.Productivity of scientific tweet authors are calculated and described.Authors are divided into 20 levels according to the activeness.Identity and motivation of scientific tweet authors with the highest productivity are analyzed.Number of scientific tweets from authors with different degree of activeness are compared.Result shows that value of different scientific tweets should not be equaled.Author distribution from journal level and discipline level are calculated.Disciplines with the largest number of authors are identified.Geographic distribution of scientific tweet author is measured and mapped.Principal countries and cities of scientific tweet authors are highlighted.Samples are selected based on location and frequency of scientific tweets to analyzed author's identity.Chapter 4 explores language distribution of microblog altmetrics with empirical study.Language distribution of scientific publications and scientific tweets are compared and results show that English has become lingua franca in informal scholarly communication.Different languages demonstrate strength in certain disciplines.Distribution of scientific tweets from countries around the world are compared to reveal the impact that national culture has on scientific tweet distribution;All countries are divided into three categories according to the percentage of English scientific tweets.Samples are selected to analyze the author's language preference.Author's language preference is divided into 4 categories based on whether the language of scientific tweets is in accordance with other scientific tweets.Chapter 5 explores value distribution law of microblog altmetrics with empirical study.Two particular context factors,i.e.count type and user category,and their influence on microblog altmetrics are studied;with large scale and systematic analysis,the correlation between microblog altmetrics indicators andcitation indicators is calculated and compared with that of other altmetrics indicators;co-occurrence analysis is adopted to cluster altmetrics indicators into three categories among which microblog altmetrics is in the core position.Timeline analysis demonstrates that microblog altmetrics is probably the prior event of other events;content analysis of scientific weibo shows that percentage of scientific weibo that reflects the scholarly impact is low.It is probably the ultimate reason why the correlation with citation indicators is low.Meanwhile,microblog altmetrics reflect entertainment impact and societal impact.Conclusion is the most frequently mentioned element in scientific weibo.