SOCIAL NETWORK ANALYSIS FOR DIGITAL HUMANITIES

Doctoral thesis (2021)

Current trends in academia show that a key factor for tackling complex problems and doing successful research is interdisciplinarity. With the increasing availability of digital tools and online databases ... [more ▼]

Current trends in academia show that a key factor for tackling complex problems and doing successful research is interdisciplinarity. With the increasing availability of digital tools and online databases, many disciplines in the humanities and social sciences are seeking to incorporate computational techniques in their research workflow. Digital humanities (DH) is a collaborative and interdisciplinary area of research that bridges computing and the humanities disciplines, bringing digital tools to humanities scholars to use, together with a critical understanding of such tools. Social network analysis is one of such tools. Social network analysis focuses on relationships among social actors and it is an important addition to standard social and behavioral research, which is primarily concerned with attributes of the social units. In this work we present the field of digital humanities and its current challenges, as well as an overview of the most recent trends in historical network research, emphasizing the advantages of using social network analysis in history and the missed opportunities. We then present the field of network analysis, providing a formalization of the concept of social network, models that explain the mechanism governing complex networks and tools such as network metrics, orbit analysis and Exponential Random Graph Model. We tackle the problem of community detection. We propose MemLPA, a new version of the label propagation algorithm, by incorporating a memory element, in order for nodes to consider past states of the network in their decision rule. We present a use case, drawn from the collaboration with a historian colleague, showing how social network analysis can be used to answer research questions in history. In particular, we addressed the gender and ethnic bias problem in computer science research by looking at different collaboration patterns in the temporal co-authorship network. Finally, we present another use case, based on collaboration data collected at the National Electronics and Computer Technology Center (NECTEC) in Thailand. We build a temporal collaboration network where researchers are connected if they worked together on one or more artifacts, focusing on measuring productivity and quality of research and development, while linking these metrics to the structure of the collaboration network. [less ▲]

DeXTER (DeepTextMiner): A deep learning, critical workflow to contextually enrich digital collections and visualise them

Software (2021)

From Digitized Sources to Digital Data, Behind the Scenes of (Critically) Enriching a Digital Heritage Collection

in Weber, Andreas; Heerlien, Maarten; Gassó Miracle, Eulàlia (Eds.) et al Proceedings of the International Conference Collect and Connect: Archives and Collections in a Digital Age (2020)

Digitally available repositories are becoming not only more and more widespread but also larger and larger. Although there are both digitally-born collections and digitised material, the digital heritage ... [more ▼]

Digitally available repositories are becoming not only more and more widespread but also larger and larger. Although there are both digitally-born collections and digitised material, the digital heritage scholar is typically confronted with the latter. This immediately presents new challenges, one of the most urgent being how to find the meaningful elements that are hidden underneath such unprecedented mass of digital data. One way to respond to this challenge is to contextually enrich the digital material, for example through deep learning. Using the enrichment of the digital heritage collection ChroniclItaly 3.0 [10] as a concrete example, this article discusses the complexities of this process. Specifically, combining statistical and critical evaluation, it describes the gains and losses resulting from the decisions made by the researcher at each step and it shows how in the passage from digitised sources to enriched material, most is gained (e.g., preservation, wider and enhanced access, more material) but some is also lost (e.g., original layout and composition, loss of information due to pre-processing steps). The article concludes that it is only through a critical approach that the digital heritage scholar can successfully meet the interpretive challenges presented by the digital and the digital heritage sector fulfil the second most important purpose of digitisation, that is to enhance access. [less ▲]

A Memory-Based Label Propagation Algorithm for Community Detection

in Aiello, Luca Maria; Cherifi, Chantal; Cherifi, Hocine (Eds.) et al Complex Networks and Their Applications VII (2018, December 02)

The objective of a community detection algorithm is to group similar nodes in a network into communities, while increasing the dis- similarity between them. Several methods have been proposed but many of ... [more ▼]

The objective of a community detection algorithm is to group similar nodes in a network into communities, while increasing the dis- similarity between them. Several methods have been proposed but many of them are not suitable for large-scale networks because they have high complexity and use global knowledge. The Label Propagation Algorithm (LPA) assigns a unique label to every node and propagates the labels locally, while applying the majority rule to reach a consensus. Nodes which share the same label are then grouped into communities. Although LPA excels with near linear execution time, it gets easily stuck in local optima and often returns a single giant community. To overcome these problems we propose MemLPA, a novel LPA where each node imple- ments memory and the decision rule takes past states of the network into account. We demonstrate through extensive experiments on the Lancichinetti-Fortunato-Radicchi benchmark and a set of real-world net- works that MemLPA outperforms most of state-of-the-art community detection algorithms. [less ▲]

A Degenerate Agglomerative Hierarchical Clustering Algorithm for Community Detection

in Nguyen, Ngoc Thanh; Hoang, Duong Hung; Hong, Tzung-Pei (Eds.) et al Intelligent Information and Database Systems (2018)

Community detection consists of grouping related vertices that usually show high intra-cluster connectivity and low inter-cluster connectivity. This is an important feature that many networks exhibit and ... [more ▼]

Community detection consists of grouping related vertices that usually show high intra-cluster connectivity and low inter-cluster connectivity. This is an important feature that many networks exhibit and detecting such communities can be challenging, especially when they are densely connected. The method we propose is a degenerate agglomerative hierarchical clustering algorithm (DAHCA) that aims at finding a community structure in networks. We tested this method using common classes of graph benchmarks and compared it to some state-of-the-art community detection algorithms. [less ▲]