Comprehension Technology For Science And Technology Literature Abstracts

Co-word analysis and literature concept tree are commonly used functions in literature analysis,while existing co-word analysis functions often incorrectly associate unrelated entity pairs,and literature concept tree functions incorrectly exclude some similar concepts.Therefore,this thesis addresses the shortcomings of the existing Chinese term co-occurrence analysis,proposes comprehension technology for science and technology literature abstracts and its processing flow,and investigates the information extraction and literature research object identification methods of scientific and technical literature abstracts in the processing flow,with the following four main works:(1)To address the shortcomings of term co-word analysis and literature concept tree function in which term association is not considered semantically,the comprehension technology for science and technology literature abstracts is proposed.The basic concepts of comprehension technology for science and technology literature abstracts and its processing flow,the concepts and connotations of scientific and technical terms,research methods and research objects involved in the process are elaborated to provide support for the subsequent research on abstract information extraction techniques and research object identification methods.(2)In order to meet the needs of comprehension technology for science and technology literature abstracts for different scientific and technical fields,and in view of the fact that the existing domain terminology identification is only carried out in specific fields,the technology of scientific and technical terminology and research method information extraction in scientific and technical literature abstracts is studied.This thesis recognizes terms from scientific and technical literature abstracts according to their textual characteristics,selects UIE based on ERNIE for fine-tuning,constructs prompt learning templates and a multi-answer extraction structure to assist in fine-tuning,and experimental results on the self-annotated dataset and the CMe EE-V2 dataset demonstrate the feasibility and effectiveness of the information extraction technique in this thesis.(3)In order to achieve the correct association of scientific and technical terms of scientific and technical literature abstracts at the semantic level,a rule-based research object identification method is proposed in conjunction with the structure of scientific and technical terms.The method is based on the extraction information of literature titles and abstracts,and investigates the alignment rules of similar scientific and technical terms with the same research objects in the titles and abstracts,using the abstract information to identify the research objects of the literature from the literature titles,and thus associating the scientific and technical terms in the abstracts.The test results on the literature extraction information show the effectiveness of the method,and the test results are analyzed.(4)A system for understanding abstracts of scientific and technical literature was designed and implemented.Based on the scientific and technical literature abstract understanding and its processing flow proposed in this thesis,a convenient scientific and technical literature abstract understanding system integrating literature management functions was designed and implemented to provide a one-stop service platform for researchers to read and export literature data,manage literature and analyses literature functions,in view of the fragmentation in the use of existing literature management software and literature analysis software.This thesis investigates the shortcomings of the existing Chinese term co-occurrence analysis,proposes and implements a more accurate comprehension technology for science and technology literature abstracts for term and its relationship identification,designs and implements a scientific and technical literature abstract understanding system that integrates literature management,and provides a more convenient one-stop service platform for literature management and analysis for the majority of researchers.