TwitterAU

After the impressive diffusion of social media and microblogging websites of the last years, the identification of users having the capability of influencing other users’ choices has becoming an important research topic because of the opportunities it can offer to many business companies. We proposed to model the contents of tweets posted by users to express opinions on items with a three-layer network. Layers represent users, items, and keywords, along with intra-layer interactions among the actors of the same layer. Inter-layer connections are triples (u,i,k) expressing the information that a user u comments on an item i by using a keyword k. By exploiting multilinear algebra, we present a method capable to extract the most active users stating their point of view about dominant items tagged with dominant keywords. Experiments on two real use cases show the ability of the approach to find influential users very active in posting opinions about the topic of interest.

We applied TwitterAU to a real world dataset regarding the tweets posted by people about smartphones. The study stems from the requirement of mobile phone manufacturers to enhance their smartphone branding and competitive positioning in the international market. E-commerce websites apply marketing techniques to recommend their products to customers. So, they need to better understand social network key influencers, brand perception, users preferences by addressing key questions like:
(a) Who is talking about smartphones?
(b) Who are the top influencers and what characterize them?
(c) Which smartphone models are generating most interest?
(d) What aspects or opinion are people associating with every phone brands?

Dataset Smartphone
User: 9783
Smartphone: 51
Keywords: 2706
Edge of tensor: 26673
Download

We considered another real-world dataset related to a different topic: tweets dealing with very popular TV series.

Dataset TV Series:
User: 14207
TV Series: 12
Keywords: 6123
Edge of tensor: 51534
Download

To test the scalability of the approach, we randomly generated a user network composed as follows:

Synthetic networks
User: 4000
Objects: 100
Keywords: 5000
Edge of tensor: from 10K to 100K by step 10K
Download