Testing an Automated Accuracy Assessment Method on Bibliographic Data

書目資料準確性評估自動化之測試研究

本研究探討資料品質文獻所提及的自動化資料準確性評估法，以瞭解其用於評估書目資料時的適切性。本研究用來測試的書目資料為兩位諾貝爾化學獎得主10年內之出版品，書目資料檢索自Web of Science與Scopus；在準確性評估上，分別以自動化與人工兩種評估法進行書目資料準確性測試，之後再跟原始出版品比對，以瞭解人工與自動化評估的高下。研究結果顯示，人工評估法的準確性得分較高，自動化評估法還需要納入更多能反映書目資料特質的評估規則，始能提高準確性。在兩組書目資料的測試中，單一分欄資料準確性的評估，都比整體書目記錄評估的表現要好。本研究之貢獻在於增進對書目資料準確度標準評估法的探討，並說明了資料準確性在引文比對過程中的重大影響。

This study investigates automated data accuracy assessment as described in data quality literature for its suitability to assess bibliographic data. The data samples comprise the publications of two Nobel Prize winners in the field of Chemistry for a 10-year-publication period retrieved from the two bibliometric data sources, Web of Science and Scopus. The bibliographic records are assessed against the original publication (gold standard) and an automatic assessment method is compared to a manual one. The results show that the manual assessment method reflects truer accuracy scores. The automated assessment method would need to be extended by additional rules that reflect specific characteristics of bibliographic data. Both data sources had higher accuracy scores per field than accumulated per record. This study contributes to the research on finding a standardized assessment method of bibliographic data accuracy as well as defining the impact of data accuracy on the citation matching process.