Clustering versus faceted categories for information exploration, By MA Hearst, in Communications of the ACM, Volume 49 , Issue 4 (April 2006)
Based on usability perspective, this paper reveals the complex of two grouping mechanisms: clustering and faceted classification.
Traditional top-down and predefined methods like clustering approaches have benefits in their algorithms and automaticabilities while in some bottom up user-oriented methods, the hierarchical faceted categories (HFC) as the author has proposed in particular, is in favour of locating user interest through some manual setting of category hierarchies which are associated with multiple facets.
This paper first discusses some advantages and disadvantage of clustering. Simple clustering algorithms for designers and clarifying vague queries for users by returning the dominant themes as results are main reasons lead designers to take the clustering approach. However, empirical evidence does not support these usabilities. Second, the author explains why clustering method is not a useful and effective tool in information exploration and proposes the hierarchical faceted categories (HFC) approach with an introduction to their prototype: The Flamenco Open Source faceted classification project.
Based on usability perspective, this paper reveals the complex of two grouping mechanisms: clustering and faceted classification.
Traditional top-down and predefined methods like clustering approaches have benefits in their algorithms and automaticabilities while in some bottom up user-oriented methods, the hierarchical faceted categories (HFC) as the author has proposed in particular, is in favour of locating user interest through some manual setting of category hierarchies which are associated with multiple facets.
This paper first discusses some advantages and disadvantage of clustering. Simple clustering algorithms for designers and clarifying vague queries for users by returning the dominant themes as results are main reasons lead designers to take the clustering approach. However, empirical evidence does not support these usabilities. Second, the author explains why clustering method is not a useful and effective tool in information exploration and proposes the hierarchical faceted categories (HFC) approach with an introduction to their prototype: The Flamenco Open Source faceted classification project.
Table 1 shows the comparision of clustering and faceted classification
