基于文本语境约束的XML商业报告多维分析模型研究

The XML-based business reports have two characters, the XLink mechanism for cross-documents element linking and the huge text information to explain the structured table. To address the complications resulted from the two characters, this proposal focuses on the multi-dimensional analytical model for the mass XML-based business reports contextualized with textual. First, in order to fetch data for multi-dimensional analytical model, the proposal studies the method to setup index for cross-documents linking arcs and the corresponding navigation algorithm. Second, get associated textual contexts using keyword query method, and then build the domain ontology based on the dimension values of structured table and extract ontology instances from textual contexts using ontology rule-based reasoning. The relevance model is also proposed to compute the correlations between text and query conditions, extracted facts and textual contexts. Lastly, a dimensional hypercube is setup based on fact correlation, context correlation and dimensions of structured business report table. Through operations including selection, projection and aggregation based on definition of data model and fundamental operating algebra, the mass XBRL data can be processed more efficiently.

针对XML商业报告采用XLink机制进行跨文档元素关联、报告中融合大量文本信息对结构化表格进行说明的特性，研究基于文本语境约束的海量XML商业报告的多维分析模型。首先，研究支持跨文档链接弧的索引构建方法和导航算法，为多维分析模型提供数据获取机制。其次，通过关键字查询方法获取关联的文本语境，进而结合结构化表格的维度值、基于领域本体规则推理从文本中自动抽取实例，形成关注事实数据。通过建立相关度模型，获取文本与查询条件相关度、关注事实与文本语境相关度。最后，在事实相关度、语境相关度和商业报告结构化表格维度的基础上建立维度超立方体，通过定义其数据模型和基础操作代数实现选择、投影和聚合等操作，从而解决海量XBRL数据的高效分析处理问题。

针对XBRL商业报告采用XLink机制进行跨文档元素关联、报告中融合大量文本信息对结构化表格进行说明的特性，研究海量XBRL商业报告基于语义的融合和数据分析。采用XBRL维度技术，提出基于空超立方体实现非维度化基础元素的维度化解决方案和基于空维度去除冗余元素的解决方案，建立多维数据模型。基于语义网技术，提出并实现了基于本体的XBRL语义元模型和知识库的构建，设计实现了一种基于语义的XBRL数据到OWL数据的转换方法，具体包括：基于语义网技术构建XBRL财务报告和事实数据语义元模型、 解析XBRL分类标准构建其XML信息集及其生成算法、解析XBRL实例文档构建实例文档信息集及其生成算法、设计XBRL分类标准到OWL本体转换算法、设计XBRL实例文档到OWL实体的转换算法。在语义层面实现数据格式的融合。