Table 1.
Measures for base graphs depicting growth of probiotics communities on Twitter from 2009–17.
Table 2.
The top ten models and their hyperparameters based on coherence scores.
Fig 1.
A series of word clouds depicting the top 10 words in each topic.
The size of each word increases as a function of it’s relative frequency in the model. The model shows us the most prevalent words (and their weights) it has grouped together.
Fig 2.
t-SNE plot illustrating the distribution of each tweet, and it’s dominant topic (colour).
The plot denotes each topic using a colour, with each point being a document (tweet). The distances between each topic indicate the inter-topic distance. For instance, Topic 0 (Functional Food) is by far the largest and most dominant topic with the most tweets. It is also quite distinct from other topics in the array as most of the tweets are clustered together without other colours (topics) mixed in.
Table 3.
Community themes with descriptions and exemplars.
Fig 3.
Alluvial diagram showing the variations in online probiotic community themes on Twitter from 2009–2017.
Each coloured block represents a theme with the stream fields showing how the respective themes varied from one year to the next. There are total of five themes across the nine year period. All the five themes were noticed for the years 2013 an 2014. Health Benefits, COM1 Advertising and Multibrand Advertising are the most popular themes.
Table 4.
Ranking of top Twitter accounts by out-degree and in-degree frequencies used as surrogates to identify hubs and authorities respectively within the UK probiotcs network.
Fig 4.
Bump charts demonstrating longitudinal patterns in top hubs and authorities of Twitter probiotics chatter from 2009–2017.
Lower numbers on the y-axis indicate a higher rank. COM1 is the only account that is consistently present in both the top hubs amd authorities charts. COM (Commerical Organizations) are more prevalent as top authorities while having a minimal presence as top hubs.
Fig 5.
Plots of in-degree values (posting) against out-degree values (tagging) demonstrating online behaviour for key influencers are plotted in scatter plots for the years 2009–2017.
The top two accounts with highest degree values are labelled in these plots. In addition, the percentage of accounts which have out-degree values more than in-degree values are displayed alongside the year in the plots. COM (Commerical Organizations) consistently appear as top accounts across the years. The tagging behavior is dominant until 2012 while posting behavior takes precedence in the last five years of the analysis period.