文摘创新点挖掘的语义识别与关联发现方法研究

The problem which is existed in the scientific field is it is becoming more and more difficult to get innovation idea for any scientific researcher by reading academic literature. The reason of the above problem are the number of academic literature is appearing explosive growth, and the scientific disciplines are divided finer and finer, and the acceleration in interdisciplinary research. In order to solve the problem, a new method named as the academic articles' abstracts semantics recognition and association discovery based on innovation points mining of academic articles' abstracts is proposed in this project. Firstly, the ontology semantic model which is used to describe the innovation point of abstract will be constructed in the proposed method. And a mining method for innovation point semantic recognition and extraction is proposed based on the above model is proposed. The innovation point semantic is described as a three tuple coarse-grained structure which comprises “problems, method, results”. So a coarse-grained knowledge database will be produced with the mining method. At the same time, a query system which is based on three tuple coarse-grained knowledge database will be built. Secondly, every knowledge unit in the above three tuple will be decomposed as SVO(subject predicate object) fine-grained semantic keywords units, and the innovation point knowledge database which is described as some predicate formulas will be produced. Finally, knowledge discovery can be realized with using semantics of predicates complementary relationship. Our method is helpful to solve the problem that is getting innovation point from academic literature are increasingly difficult. Our method can expand the knowledge service technologies which are in the field of Library and information. So our method has important theoretical significance and wide application prospect.

针对科技界出现的学术文献海量式增长、学科细分加剧以及跨学科研究加速等因素导致的学术文献创新点获取日益困难的问题。提出文摘创新点挖掘的语义识别与关联发现方法。该方法首先构建学术文摘创新点陈述的本体语义模型，并依此模型提出“问题、方法、结果”三元组结构化粗粒度创新点语义识别和提取的挖掘方法，产生粗粒度创新点知识库。建立三元组创新点粗粒度知识库的查询系统。然后，将三元组中的每一元知识分解成为“主谓宾”细粒度的语义关键词单元，并生成谓词公式表达的创新点语义知识库，最后利用谓词的语义互补关系和关联关系实现知识发现。这种方法有助于解决学术文献创新点获取日益困难问题，对拓展图书情报档案领域的知识服务技术，具有重要的理论意义和广泛的应用前景。

搜索引擎的出现促进了学术研究和文献信息查找的高度融合，但是为用户提供海量文献信息的同时也给科研工作者精准获取知识带来了很大的难度。针对当前学科细分加剧，导致出现利用关键词检索文献信息获取跨学科创新点日益困难等问题。本项目，首先根据学术文摘能凸显和简明清晰地陈述创新点要素等特点，构建了表达学术文摘创新点的本体语义结构模型，并依此模型建立了PMR（问题、方法、结果）三元组结构化创新点挖掘方法，开发了创新点知识库事实数据的关联查询与推理方法；再次，利用PMR事实数据库为用户创建了科学研究成果创新点自助查新模型工具，提升了科学研究工作者检测判断科研成果创新点的速度；最后，利用医学文摘具有针对疾病症状采用药物治疗方法的事实数据，挖掘中文急救医学数据，采用分词工具和语义识别技术，构建了医学急救领域SPO（主谓宾）语义三元组模型，实现了中文医学数据的SPO抽取理论模型和软件关键技术的开发，建立了国内中文急救医学数据知识库和SPO知识图谱模型。上述阶段性研究中开发出来的工具都在北京万方数据股份有限公司进行了初步验证。本项目在进行上述阶段性研究工作的同时，还进行了如下的相关研究并取得相关的成果：（1）围绕学术文摘创新点挖掘的语义识别与关联发现方法研究，开展了学术文摘创新点的模型和挖掘研究，进行了文本自然语言处理的技术攻关，培养硕士生1名；（2）发表与课题相关并标注本基金资助的学术论文共计7篇，其中主流国际会议（Fuzzy Systems and Data Mining V - Proceedings of FSDM）论文1篇，被EI索引1篇、相关领域内的权威期刊论文、核心期刊论文6篇；（3）申请相关专利2项。