Mining Hypernyms Semantic Relations from Stack Overflow

Abstract

Communication between a software development team and business partners is often a challenging task due to the different context of terms used in the information exchange. The various contexts in which the concepts are defined or used create slightly different semantic fields that can evolve into information and communication silos. Due to the silo effect, the necessary information is often inadequately forwarded to developers resulting in poorly specified software requirements or misinterpreted user feedback. Communication difficulties can be reduced by introducing a mapping between the semantic fields of the parties involved in the communication based on the commonly used terminologies. Our research aims to obtain a suitable semantic database in the form of a semantic network built from the Stack Overflow corpus, which can be considered to encompass the common tacit knowledge of the software development community. Terminologies used in the business world can be assigned to our semantic network, so software developers do not miss features that are not specific to their world but relevant to their clients. We present an initial experiment of mining semantic network from Stack Overflow and provide insights of the newly captured relations compared to WordNet.