On-demand Influencer Discovery on Social Media

Abstract

Identifying influencers on social media, such as Twitter, has played a central role in many applications, including online marketing and political campaigns. Compared with social media celebrities, domain-specific influencers are less expensive to hire and more engaged in spreading messages such as new treatment or timely prevention for HIV. However, most of the existing topic modeling based approaches fail to identify influencers who are dedicated to the rare yet important topics such as HIV and suicide. To alleviate this limitation, we investigate an on-Demand Influencer Discovery (DID) framework that is able to identify influencers on any subject depicted by a few user-specified keywords, regardless of its popularity on social media. The DID model employs an iterative learning process that integrates the language attention network as a subject filter and the influence convolution network built on user interactions. Comprehensive evaluations on Twitter datasets show that the DID model can reliably identify influencers even on rare subjects such as HIV and suicide, outperforming existing topic-specific influencer detection models.

Topic-based influential user detection: a survey

“…While very few approaches have used manually handcrafted keywords to define multiple topics, [14,131], various other techniques, such as topic modeling [7,9,13,20,23,28,37,39,47,60,64,67,82,84,96,98,104,108,109,111,115,117,117,121,125,130,132], machine learning [85,108], and platform structures [50,126] are used to infer multiple topics in the literature. The way content is organized on a platform can be considered as a representation of a topic based on platform structures; e.g., a board on the Pinterest platform discusses a single topic.…”

“…In the literature, different techniques, such as statistical measures [18,26,29,53,63,81,93,99,106,107], random-walk approaches [2, 7, 7, 14, 16, 23, 24, 37, 39, 65, 67-69, 75, 76, 84, 88, 96, 108, 111, 115, 124, 126, 128], propagation algorithms [28,85,108,115], machine learning [25,50,[129][130][131], topic modeling [9,13,20,64,82,98,104,109,117,121], similarity measures [47,117,125], and optimization algorithms [113] are used for topic-based influential user detection. The Pagerank algorithm, i.e.…”

“…Manually hand-crafting precise keywords sets denoting multiple topics is very challenging and rarely explored [14,131] as a multiple topic detection techniques in the literature. In this section, we discuss multiple topic-based IUD approaches that use a set of keywords to denote each topic.…”

“…Zheng et al [131] use neural networks to formulate the problem of topic-based IUD as supervised learning. First, words in the posts published by a user are converted to a vector representation using Glove [80] word embedding, and then the vector representation is used as input to the proposed convolutional neural network to predict the influence score of the users.…”

“…First, words in the posts published by a user are converted to a vector representation using Glove [80] word embedding, and then the vector representation is used as input to the proposed convolutional neural network to predict the influence score of the users. In order to generate the ground truth influential users, [131] proposes another neural architecture that can take the vector representation of users' posts and a set of keywords representing the topic, and predict the relatedness of the post with respect to the keywords. The user who posts content related to the keywords and has the highest number of followers are considered as ground truth influential users for training the convolutional neural network.…”