A network approach to topic models

Gerlach, Martin; Peixoto, Tiago P.; Altmann, Eduardo G.

doi:10.1126/sciadv.aaq1360

Cited by 194 publications

(191 citation statements)

References 57 publications

Supporting

Mentioning

169

Contrasting

Unclassified

Order By: Relevance

“…We used the Force Atlas 2 algorithm to structure the semantic network (Jacomy, Venturini, Heymann, & Bastian, ). To detect topics, we employed the Louvian method for community detection (Blondel, Guillaume, Lambiotte, & Lefebvre, ; Gerlach, Peixoto, & Altmann, ). We found that this approach detected topics better than the more common Latent Dirichlet Allocation topic modeling approaches (Hong & Davison, ) or the structural topic modeling approach (Roberts, Stewart, Tingley, & Airoldi, ; see Supplementary Information S1.3).…”

Section: Methodsmentioning

confidence: 99%

Do Parliamentary Debates of e‐Petitions Enhance Public Engagement With Parliament? An Analysis of Twitter Conversations

Asher¹,

Leston‐Bandeira²,

Spaiser³

2019

Policy & Internet

View full text Add to dashboard Cite

The U.K. Parliament introduced an e‐petitions system in 2015 with the aim of significantly enhancing its relationship with the public. We explore whether this aim is being met through the analysis of Twitter data from conversations on e‐petitions debated in parliament. We use natural language processing, machine learning, and social network analysis of Twitter data to explore what it shows about the extent of people's engagement, the contents of Twitter e‐petition conversations, who is taking part, and how they interact. Our findings provide interesting insights into how people perceive the e‐petition procedures in terms of fairness and responsiveness, suggesting that petition parliamentary debates should be more inclusive of the original petitions’ aims. The results also point to homophily tendencies present in the Twitter e‐petition discussions.

show abstract

Section: Methodsmentioning

confidence: 99%

Do Parliamentary Debates of e‐Petitions Enhance Public Engagement With Parliament? An Analysis of Twitter Conversations

Asher¹,

Leston‐Bandeira²,

Spaiser³

2019

Policy & Internet

View full text Add to dashboard Cite

show abstract

“…A pictorial representation of the activity of the user and the relationship between posts, pages and topics in reported in Figure 1. [22] . Single icons created using the free gallery of OpenOffice.…”

Section: A Users' News Consumptionmentioning

confidence: 99%

“…By representing the relationship between words and documents (posts in our case) as a bipartite network, the algorithm proposed by [22] detects communities (i.e. cluster of densely interconnected nodes) using a hierarchical Stochastic Block Modeling (hSBM) algorithm [33][34][35].…”

Section: A Topic Modeling Algorithmmentioning

confidence: 99%

“…In [22] a comparison between topic modeling and community detection algorithms, namely pLSA and SBM and LDA and hSBM, is carried on in order to demonstrate the suitability of hSBM for topic modeling problems. In particular, a mixed membership version of the SBM is formally proven to be equivalent to pLSA while the hSBM is shown to be conceptually similar to LDA.…”

Section: A Topic Modeling Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Selective exposure shapes the Facebook news diet

Cinelli¹,

Brugnoli²,

Schmidt

et al. 2020

PLoS ONE

View full text Add to dashboard Cite

The social brain hypothesis fixes to 150 the number of social relationships we are able to maintain. Similar cognitive constraints emerge in several aspects of our daily life, from our mobility up to the way we communicate, and might even affect the way we consume information online. Indeed, despite the unprecedented amount of information we can access online, our attention span still remains limited. Furthermore, recent studies showed the tendency of users to ignore dissenting information but to interact with information adhering to their point of view. In this paper, we quantitatively analyze users' attention economy in news consumption on social media by analyzing 14M users interacting with 583 news outlets (pages) on Facebook over a time span of 6 years. In particular, we explore how users distribute their activity across news pages and topics. We find that, independently of their activity, users show the tendency to follow a very limited number of pages. On the other hand, users tend to interact with almost all the topics presented by their favored pages. Finally, we introduce a taxonomy accounting for users behavior to distinguish between patterns of selective exposure and interest.Our findings suggest that segregation of users in echo chambers might be an emerging effect of users' activity on social media and that selective exposure -i.e. the tendency of users to consume information interest coherent with their preferences -could be a major driver in their consumption patterns. *

show abstract

“…in Section 1.Before looking at SBMs with topic modelling for graph and textual data, it is useful to look at a recently proposed SBM for textual data byGerlach et al (2018). Instead of probabilitistically factorising the document-word frequency matrix M (Section 10), they treated the relations between the words and the documents as a bipartite graph, in which both the words and the documents are the nodes.…”

mentioning

confidence: 99%

A review of stochastic block models and extensions for graph clustering

Lee¹,

Wilkinson²

2019

Appl Netw Sci

145

View full text Add to dashboard Cite

There have been rapid developments in model-based clustering of graphs, also known as block modelling, over the last ten years or so. We review different approaches and extensions proposed for different aspects in this area, such as the type of the graph, the clustering approach, the inference approach, and whether the number of groups is selected or estimated. We also review models that combine block modelling with topic modelling and/or longitudinal modelling, regarding how these models deal with multiple types of data. How different approaches cope with various issues will be summarised and compared, to facilitate the demand of practitioners for a concise overview of the current status of these areas of literature.

show abstract

A network approach to topic models

Abstract: A new approach to topic models finds topics through community detection in word-document networks.

Cited by 194 publications

References 57 publications

Do Parliamentary Debates of e‐Petitions Enhance Public Engagement With Parliament? An Analysis of Twitter Conversations

Do Parliamentary Debates of e‐Petitions Enhance Public Engagement With Parliament? An Analysis of Twitter Conversations

Selective exposure shapes the Facebook news diet

A review of stochastic block models and extensions for graph clustering

Contact Info

Product

Resources

About