Log in / Register
Home arrow Computer Science arrow Social Computing, Behavioral-Cultural Modeling and Prediction
< Prev   CONTENTS   Next >

2 Influence Factors for Research Topic Adoption

This work concerns the adoption of research topics as evidenced by authors' publications. A “topic” is defined as a popular term that represents a specific scientific concept. Adopting a topic means an author has published at least one paper that contains the term in either the title or the abstract.[1] An author X “follows” another author Y if X adopts a topic after Y adopts the same topic. This following behavior is also interpreted as an infection process [2]. The following lists the factors that potentially play a role for the influence one author has on the others to adopt a research topic.

2.1 Direct Observation

The direct observation of the adoption is the time when an author adopts the topic, and a cascade can be generated based on the chronological ordered observations. Each topic has its own cascade. According to Gomez-Rodriguez et al. [2], recovering an influence network with K edges will require 2K~5K cascades. This inferred influence network could be used to predict new topic adoption based on the assumption that if author X followed Y in many topics before, it is also likely X will follow Y again for a new topic. Generally speaking, the closer in time X followed Y , the more influence Y has on X based on this direct observation [2].

2.2 Indirect Observation

The direct observation helps to recover the influence network based on past topic adoption cascades, but it requires sufficient number of relevant cascades to learn the network. This limitation motivates the consideration of other factors that indirectly implies the relationship, hence the influence one author may have on another to adopt a new topic.

Social Connections. The basic idea is that authors are more likely to receive influence from their colleagues, co-authors and other socially connected peers, than unknown persons. Sun et al. [6] [7] suggested that the social connections could be the main reason of people establishing a new relation. Sun et al. [6] use meta-path to define the social connections. A meta-path is defined on the network schema, where nodes are object types and edges are relations between object types. For example, two authors are coauthored in a paper is defined as: Author-Paper-Author. To define the social relationships among authors, eight common social connections are defined: CI: cite peer's paper; CA: coauthor; CV: publish in the same venue; CSA: are co-authors of same authors; CT: write about the same topic; CICI: cites papers cite peer's; CIS: cite the same papers; SCI: cited by the same papers.

Topic Popularity. Besides the social connections affecting the adoption probability, the research topic itself is also a factor. Not only the match between the topic and the author's own interest, but also the popularity of the topic can affect how fast and easily it is being adopted. The topic popularity is defined as the average number of papers adopting the topic since the first year it appeared. Fig. 1 shows the popularity of four selected machine learning topics extracted from 20 conferences from 1989 to 2010.

Fig. 1. Individual Popularity

As the figure shows, different terms have different popularity curves. Given the same social connections and the same initial authors who adopted the topic earliest, different topics could result in different adoption cascades. There also exists the case where an author adopts the topic without receiving any influence from her social peers. Based on these observations, this work hypothesizes that topic popularity can be a factor that affects the adoption process. To include the topic popularity in the model, a virtual author is added where the connection between any author to this virtual author represents the influence comes from the topic popularity.

  • [1] Topic adoption is a complex process. Nevertheless, to some extent, we believe that publishing papers with topic terms are reflective of the nature of the adoption
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science