基于学术文献全文内容的细粒度算法实体抽取与评估研究

在大数据时代，算法在学术研究中的提出、改进和应用，对推动不同学科的发展起到重要作用。在基于数据驱动型的研究成果中包含大量的算法，为研究特定领域算法实体提供重要的数据基础。本研究以自然语言处理领域为例，利用该领域学术文献全文内容，开展特定领域算法实体抽取和评估研究。具体而言，首先，本项目构建规范的全文语料和训练语料；随后，利用弱监督机器学习方法从文献中自动抽取算法实体和其他相关实体，并识别算法实体关系、不同实体间关系，在减少人工标注量的同时，保证结果的准确性；接着，统计算法实体被提及频次，识别算法提及位置与动机、算法所在文献的主题和任务；最后，本研究综合上述特征，对算法实体进行多维度关联分析，并构建算法知识库，从而对特定领域的算法进行综合评估。本研究的结果，能为相关科研人员提供科学、客观的视角，帮助他们全面了解不同算法在特定领域的影响力与应用情况，从而为他们选择与使用算法提供坚实的基础。

In the era of big data, the proposal, improvement and application of algorithms in academic research have played an important role in promoting the development of different disciplines. Academic documents in the field of data-driven research contain a large number of algorithms，which are important basic data for researching algorithm entities in specific fields. To this end, this article takes the field of natural language processing (NLP) as an example, and uses the full text of academic papers in NLP to carry out the extraction and evaluation of algorithm entities in a specific field. Specifically, this research manually constructs a structured full-text corpus and training corpus based on content analysis. Then, we use weakly supervised machine learning methods to automatically extract algorithm entities and other entities from academic papers, and explore the relationship between different entities. The method reduce the workload of manual annotation while ensuring the accuracy of the results. At the same time, this research utilizes the full-text content to count the mentioned frequency, identify the location and motivation of the algorithm entity, as well as the topic and task of the article where the algorithm extracted. Finally, based on the above characteristics, multi-dimensional association analysis is performed on algorithm entities, and an algorithm knowledge base is constructed to comprehensively evaluate the algorithm in a specific domain. The results of this research can provide researchers with a scientific and objective perspective and help them understand the influence and practical application of different algorithms in specific fields, thereby providing them with a solid foundation for understanding, selecting, and applying algorithms.