Defining and identifying stophashtags in instagram

Abstract

Instagram could be considered as a tagged image dataset since it is reach in tags -known as hashtags- accompanying photos and, in addition, the tags are provided by photo owners/creators, thus, express in higher accuracy the meaning/message of the photos. However, as we showed in a previous study, only 30% of Instagram hashtags are related with the visual content of the accompanied photos while the remaining 70% are either related with other meta-communicative functions of the photo owner/creator or they are simply noise and are used mainly to increase photo’s localization and searchability. In this study we call the latter category of Instagram hashtags as ‘stophashtags’, inspired from the term ‘stopwords’ which is used in the field of computational linguistics to refer to common and non-descriptive words found in almost every text document, and we provide a theoretical and empirical framework through which stophashtags can be identified. We show that, in contrary to descriptive hashtags, stophashtags are characterized by high normalized subject (hashtag) frequency on irrelevant subject categories while normalized image frequency is also high.