運用文字探勘與XBRL技術提升企業資訊擷取與整合效益之研究

Enhancing Effectiveness of Business Information Retrieval and Integration via Text Mining and XBRL Technology

周濟群(Chi-Chun Chou) ； 連子杰(Zih-Jie Lian)

《當代會計》 12卷1期 (2011/05) Pp. 85-114

https://doi.org/10.6675/JCA.2011.12.1.04

摘要

除傳統財務報表資訊外，企業財報附註或企業年報中的非財務資訊等補充性資訊，亦包含許多重要的企業公開資訊，且兩者間應具有資訊整合的綜效。然而，受限於兩者資料格式之不一致，投資人必須透過繁瑣且錯誤率較高的人工處理程序，方能擷取、整合或流通此兩類資訊。因此，市場參與者開始重視如何利用資訊技術提升企業資訊擷取與整合效益。本研究應用設計科學研究方法，探討不同的企業報告資訊應如何整合與塑模，首先運用本體知識概念分類方法與文字探勘技術，處理股東會年報中與企業策略相關之非財務性資訊，以協助使用者有效率地分類閱讀這些非結構化文字資訊。其次則運用可延伸企業報導語言XBRL作為財務報表資料來源，以解決傳統財報格式不易處理之問題。最後，在資訊整合方面，則透過連結機制之設計以整合非財務與財務資訊。本研究並運用雛型系統設計法驗證所提出的企業資訊整合架構確實可行，能協助投資人能於短時間內擷取、整合或流通企業資訊，提升決策品質。

關鍵字

企業資訊整合 ； 概念階層 ； 文字探勘 ； 可延伸企業報告語言 ； 資訊擷取

並列摘要

In addition to the information provided in traditional financial statements, the supplementary data contained in footnotes and non-financial disclosures in the annual report proves to be decision relevant. They have the potential to provide a synergy for information integration between data sources. However, due to the format inconsistency between two data sources, investors confront inefficient and ineffective process for retrieving, integrating and exchanging information. Hence, market participants are starting to put an emphasis on utilizing information technology to advance information retrieval and integration. This research proposes an ”annual report knowledge construct” framework to enhance the effectiveness of data retrieval and integration through the design science methodology. It uses text mining technology for reclassifying the semi-structured or unstructured business strategy-related non-financial data. In addition, financial data in XBRL format is used to mitigate the format inconsistency problem. Finally, this paper develops a linking mechanism to integrate financial and non-financial data. This paper also adopts the prototyping method to prove the concept of the business information integration framework.

並列關鍵字

business information integration ； concept hierarchy ； text mining ； XBRL ； information retrieval