Forming Dream Teams: A Chemistry-Oriented Approach in Social Networks

Abstract

Scientific collaboration networks are social networks in which vertices represent scientists and edges typically represent co-authorship. Such networks not only permit research into understanding the characteristics of scientific collaboration, but can also provide a basis for building collaborative research platforms to support research groups with functionality such as, information sharing, data repositories, and communication. Collaboration networks are highly clustered, mapping closely to the real world relationships of individual researchers. However, just as big data constitute a well recognised disruptive change to the way basic research is carried out in many fields, there is an equivalent and largely unexplored change in the collaborative relationships between researchers - which are becoming not only larger in scale, but also more distributed and interdisciplinary. One element in this, which we suggest will play a pivotal role in the future, is the formation of teams for big data projects. This paper presents an innovative algorithm for expert team formation called Chemistry Oriented Team Formation (ChemoTF) based on two new metrics; Chemistry Level and Expertise Level. Chemistry Level measures scale of communication required by the task, while Expertise Level measures the overall expertise among potential teams filtered by Chemistry Level. This approach is tested using a large scholarly corpus containing 472,365 individual authors. The ChemoTF algorithm is able to build teams for median average 90 percent of the expected cost, achieving a 99 percent fit while remaining tractable for teams up to 16 individuals - resulting in the formation of more communicative and cost effective teams with higher expertise level.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.