Citation Content Analysis Of Web References In Humanities And Social Science Research Papers

This study introduced content-based citation analysis into web citation analysis. It aimed to find out the problems existing in citation location, multiple citation scenario, and the semantic features of web citation by analysis the web citation content appeared in humanities and social science research papers. And give improvement advice to scholars to correctly and effectively cite web resources.This study developed a content-based web citation analysis framework, it consisted of three aspects of citation content:citation location, multiple citation scenario, and the semantic features of citation content. By apply the framework to web citations collected from humanities and social science research papers, both the numeric and functional features of web citation can be understand. The research sample is consisted of 36,465 research papers published from 2011 to 2015, the papers are collected from 132 humanities and social science journals indexed in both SpringerLink platform and SSCI. A self designed citation content extraction program is used to automatically identify, extract, and store web citation content. The content-based web citation analysis framework is applied to 68,934 web references listed in the end of the paper and 100,961 web citation content appeared in the paper, and the results revealed scholar's preferences towards citation location, multiple citation scenario, and semantic features of citation content.In citation location analysis, the syntax structure of research paper is thoroughly examined, including papers'full word count, section count, section word count, and section title. Web citation content's location distributions is presented based on papers' syntax structure features. Results showed the most of the sample articles in humanities and social science are followed the IMRaD structure format; about a third of the web citation content are located in the beginning of the paper, i.e. introduction, background, or literature review sections; about half of the web citation content distributed in the main argumentative sections, i.e. method, results, or discussion sections; there is only a few web citation content located in the end of the papers. Inside papers'different sections, scholars' tend to cite web resources intensively at the very beginning of the introduction section and the very end of the discussion section.In the multiple citation discussion, the relationship between web references count and web citation content count are examined, the utility level is assigned to each web references based on the number of times they mentioned in the paper, and the location distribution of web references with different utility level is examined. Results showed the multiple citation count is varied in the papers, the more web references listed in the references section, the more web citation content mentioned in the paper's main body, most of the high utility level web references existed in the minority of the papers; web references with medium utility level is more likely to gather in the same section than web references with high utility level, and most medium utility level web references mentioned in the beginning of the paper are appeared in the end of the paper as well; high utility level web references' is more likely to scattered over different sections, that indicate certain web reference is fit in different argumentative content and has multiple citation functions.In web citation semantic features, each word in citation content is annotated by word classes, and the TF-IDF algorithm is applied to set weight to each word. Work noun and action verb are used to represent the citation function features, connector and adjective are used to represent the citation opinion features. Results showed the citation function of web resources is provide research example or research program information or data about health, children, education, or population; scholars cite web resources to describe, explain the research topic, or develop their own theory or argumentation based on web resources; web citation content's opinion features are mainly presented in the form of transition connection or sequence connection, web resources are cited as research evidence, scholars cite web resources to reasoning or disproving certain research topic; the language used in argumentation is mostly positive and strong negative words are not common in web citation content.Combining three aspects of citation content analysis mentioned above with web references' discipline distribution, domain distribution, file type distribution, and accessibility features, web citation's overall status is discussed. The empirical results from analysis web citations from government website sources showed government sources web citation is highly cited in public health field, most of the web citation is made to web page files or pdf files, scholars tend to cite government website sources intensively at the beginning of the paper as research background or current study status; most government website sources appeared in the paper as statement or fact, and the language used in citation content is positive, various aspects of government website resources are acknowledged in different research fields.Content-based web citation analysis revealed four features in humanities and social science research, including the density distribution of web citation, the matching between web reference and web citation, the least effort web citation behavior, and the style and format of web reference. Accordingly, there are three improvement proposal is presented to solve the problems, scholars are expected to understand and comply with the academic writing and formatting style guide, and choose web references with caution, publishing agency and other information service institution should guide and reinforce the scholars to properly cite web resources, academics should take advantages of web recourse's merit and avoid its disadvantages.