A Visual Approach for Interactive Keyterm-Based Clustering

Abstract

The keyterm-based approach is arguably intuitive for users to direct text-clustering processes and adapt results to various applications in text analysis. Its way of markedly influencing the results, for instance, by expressing important terms in relevance order, requires little knowledge of the algorithm and has predictable effect, speeding up the task. This article first presents a text-clustering algorithm that can easily be extended into an interactive algorithm. We evaluate its performance against state-of-the-art clustering algorithms in unsupervised mode. Next, we propose three interactive versions of the algorithm based on keyterm labeling, document labeling, and hybrid labeling. We then demonstrate that keyterm labeling is more effective than document labeling in text clustering. Finally, we propose a visual approach to support the keyterm-based version of the algorithm. Visualizations are provided for the whole collection as well as for detailed views of document and cluster relationships. We show the effectiveness and flexibility of our framework, Vis-Kt , by presenting typical clustering cases on real text document collections. A user study is also reported that reveals overwhelmingly positive acceptance toward keyterm-based clustering.

Bridging the Gap Between Ethics and Practice

“…In general, exploratory user interfaces are welcomed by users who spend more time developing an understanding of the sensitivity of variables and digging more deeply into aspects that interest them, leading to greater satisfaction and compliance with recommendations [80,27,21]. Further gains come from enabling adaptable user interfaces to fit different needs and personalities [127,119].…”

Human-Centered Artificial Intelligence: Three Fresh Ideas

“…Many applications involving machine and deep learning algorithms provide post-hoc explanations of why a decision refused mortgage or parole requests. However, exploratory user interfaces using interactive visual designs offer a more likely path to successful customer adoption and acceptance (Chatzimparmpas et al, 2020;Hohman et al, 2018;Nourashrafeddin et al, 2018;Yang et al, 2020). Well-designed interactive visual interfaces will improve the work of machine learning algorithm developers and facilitate comprehension by various stakeholders.…”

Interactive keyterm-based document clustering and visualization via neural language models

“…To avoid inconsistency after several interactions, the empirical convergence problem, and interaction limitations due to complex mathematical models previously present in the literature, the system Vis-Kt (NOURASHRAFEDDIN et al, 2018;SHERKAT et al, 2018) is an alternative that uses document clustering algorithms instead of topic modeling algorithms. One of the proposed algorithms is Lexical Double Clustering (NOURASHRAFEDDIN et al, 2018), which overcame state-of-the-art algorithms such as LDA and NMF in several datasets. A second the data points are colored relative to their clusters (topics).…”

“…In the system, both algorithms are interoperable. The idea behind LDC and iKMeans is that, before finding document clusters, it is better to focus on term clusters and the keyterms that represent topics meaningful to the user (NOURASHRAFEDDIN et al, 2018).…”

“…As Nourashrafeddin et al (2018) and Sherkat et al (2018) report, the keyterm-based approach for document clustering presents greater clustering quality measures when compared to state-of-the-art algorithms, such as LDA and NMF. With that in mind, this project seeks to extend the formerly proposed keyterm-based IDC system, Vis-Kt, to enhance its document representation to a more efficient representation based on neural language models.…”