Evaluating the Use of Clustering for Automatically Organising Digital Library Collections

Abstract

Abstract. Large digital libraries have become available over the past years through digitisation and aggregation projects. These large collections present a challenge to the new user who wishes to discover what is available in the collections. Subject classification can help in this task, however in large collections it is frequently incomplete or inconsistent. Automatic clustering algorithms provide a solution to this, however the question remains whether they produce clusters that are sufficiently cohesive and distinct for them to be used in supporting discovery and exploration in digital libraries. In this paper we present a novel approach to investigating cluster cohesion that is based on identifying instruders in a cluster. The results from a human-subject experiment show that clustering algorithms produce clusters that are sufficiently cohesive to be used where no (consistent) manual classification exists.

Machine Reading Tea Leaves: Automatically Evaluating Topic Coherence and Topic Model Quality

“…Since its inception, the method of Chang et al (2009) has been used variously as a means of assessing topic models (Paul and Girju, 2010;Reisinger et al, 2010;Hall et al, 2012). Despite its wide acceptance, the method relies on manual annotation and has never been automated.…”

Improving Topic Coherence Using Entity Extraction Denoising

“…As the acceptance of topic coherence measures increases as a mean of topic model assessment (Paul and Girju, 2010;Reisinger et al, 2010;Hall et al, 2012), recent research trends focus on proposing fast and efficient models that can be scaled up to big amounts of data (Yang et al, 2015;Nguyen et al, 2015), using the whole text per document for training.…”

Hierarchical Structuring of Cultural Heritage Objects within Large Aggregations

“…It is possible to receive multiple records for the same object from the same institution. 8 A quality control failure during the data ingestion process can let duplicates be published in the Europeana portal. Clustering allows us to identify these duplicates with a high degree of accuracy; often the exact same metadata appears in many fields.…”

“…In the latter case, an automatic procedure would have difficulty making the distinction with other types of relations. 8 Derivative works These are objects which are derived from another one, such as reprint. Fig.…”

“…Recently, researchers started looking at using external knowledge bases such as Wikipedia [7] or WordNet [14] to help measuring similarities between objects. Different similarity measures were compared [8,3] but most existing work explore a single dimension of similarity, which does not take into account the multidimensionality of CH collections; it also focuses on smaller-scale collections. The extraction of FRBR-like relations, a topic researched for more than a decade [17,9], has been a clear source for inspiration for us.…”