論文

In interlibrary loan (ILL) practice, the bibliographic identification task is critical. However, automating this matching process is not straightforward due to the lack of standardized data formats across libraries and the diverse nature of title and author information. In addition, false positives, where different bibliographic records are mistakenly identified as the same, can lead to serious problems such as misdelivery of materials and incorrect use of information resources. Although human judgment is effective in resolving such complex issues, it comes at a high cost. To achieve low-cost bibliographic identification, a hybrid approach that identifies high-risk pairs and requires only those pairs to human judgment while processing the rest mechanically is considered effective. However, the conditions that identify this risk have not been clarified. To address this gap, we propose a hybrid bibliographic identification framework that efficiently combines rule-based filtering, machine learning, and human judgment, and clarify the extraction conditions for pairs that require human judgment within this framework. Experiments on a real-world library dataset of 30,000 bibliographic pairs revealed that false positives from machine processing are concentrated in the pairs where titles are similar but author information is dissimilar. Applying human judgment to the pairs, we achieved a 99.7% accuracy rate with only 14 false positives. The proportion of pairs that required human judgment was limited to just 0.36% of the total. Overall, this study provides practical insights into the factors for identifying pairs that require human judgment, and rethinking the bibliographic knowledge processing at the point of human-machine collaborative intelligence in the age of AI.

Entity matching is an issue of interest in information integration and data cleaning. Since the representations of the same entity vary, it is often impossible to fully automate the entity matching and require human inputs. However, to guarantee high-quality entity matching, how to integrate human resources into the entity matching while minimizing the cost of human resources? In this paper, we propose BUBBLE, a novel human-in-the-loop entity matching framework hybridizing Bayesian inference and crowdsourcing. To guarantee entity matching quality, Bayesian inference is conducted to determine whether the matching requires crowdsourcing. We show that we can define Bayesian error rate for this problem. For optimization, we use metric learning to select the candidate matching pairs by nearest-neighbor search in the learned embedding space, and we construct a k-nearest neighbor graph to avoid the redundant matching. We applied BUBBLE to a bibliographic data matching problem on the National Diet Library. The experimental results show that BUBBLE can assign tasks to humans with higher quality results compared to those of the same number of task assignments to humans. The result also shows that our optimization scheme is effective without sacrificing the quality.

Introduction. Entity matching is a fundamental operation in a wide range of information management applications and a tremendous number of methods have been proposed to address the problem. Human-in-the-loop entity matching is a human-AI collaborative approach which is effective when the data for entity matching is incomplete or requires domain knowledge. A typical human-in-the-loop approach is to allow a machine-learning-based matcher to ask humans to match entities when it cannot match them with high confidence. However, ML-based matchers cannot avoid the unknown-unknown problem, i.e., they can resolve the entities incorrectly with high confidence. Method. This paper addresses an inconsistency-based method to deal with this problem. The method asks humans to resolve the entities when we find inconsistency in the transitivity property behind entity matching. For example, if a matcher returns a positive result only for two combinations among three entities, the result is inconsistent. Analysis. This paper shows an implementation of our idea in similarity-based blocking method and Bayesian inference and explains the result of an extensive set of experiments that reveals how and when the method is effective. Results. The result showed that the inconsistency-based sampling selects very different entity pairs compared to other sampling strategies and that a simple hybrid strategy performs well in many practical situations. Conclusion. The results indicate our approach complements any existing matcher that can cause the unknown-unknown problem in entity matching.

Since the mid-1990s, numerous digital archives have been established in Japan. Regrettably, some have been discontinued because they became unmanageable. Moreover, budget and management system cuts have impacted the longevity of existing digital archives. This study aims to identify factors influencing the longevity of these archives by analyzing the current state of Japanese digital archives and conducting interviews with pertinent stakeholders. In our investigation into the survival of digital archives, we manually confirmed the continued web accessibility of 512 digital archives that were surveyed by the National Diet Library in 2009 and for which contact information was provided. In addition, we distributed surveys to 348 institutions responsible for maintaining digital archives. The results of our web publication status survey revealed that, of the 453 digital archives made publicly accessible online in 2009, 431 (or 95.1%) continued to be published in 2018. In addition, 38 (or 64.4%) of the 59 that were not initially made available in 2009 had begun to be published online. However, the survey responses painted a different picture. Among the 192 responding institutions, 151 (or 78.6%) reported that they continue to operate their archives, whereas 41 (or 21.4%) reported that they no longer do so. In light of these results, it is evident that a substantial number of digital archives that are no longer actively operating maintain a web presence. In addition, interviews with key figures engaged in Japanese digital archives were conducted. Results reveal a number of reasons for the discontinuation of digital archives, including budgetary and human resource concerns, and technical issues like copyright handling and adapting to changes in system and data formats. In addition to these factors related to the construction and operation of digital archives, the magnitude of the external response to digital archives was identified to significantly affect their continuity. Specifically, it was suggested that the discontinuation of digital archives that receive little response from within and outside the institution could pose significant continuity issues.