Towards a Cross-Country Analysis of Software-Related Tweets

[Context and motivation] Twitter is one of the most widely used micro-blogging platforms. Globally distributed developers and software companies use Twitter to communicate about software updates, bugs and other type of information related to the software. End-users from diverse geographical regions also use Twitter to give feedback about the software they use. Previous research has shown that this feedback is valuable for requirements engineering, containing information such as feature requests and usage scenarios. However, the effect of the country of origin on software-related tweets has not been studied so far. [Question] In this paper, we investigate to what extent people from various countries provide distinct feedback regarding certain characteristics on Twitter. [Principal ideas/results] We collected 70,759 tweets (Original: 17,940, Replies: 52,819) from popular Twitter support accounts of ten software applications for two months. In the subsequent analysis, we selected the tweets originating from the eight most popular countries and analyzed a sample of 1,813 tweets with the help of automatic and manual content analysis. Results show that out of three characteristics (content, sentiment and text length); content, and sentiment differ significantly at the country level in some cases. These characteristics are used in algorithms automatically processing user feedback. Such algorithms are commonly used for requirements engineering tasks. [Contributions] Our findings show the importance of considering software-related user feedback on Twitter from a diverse audience during the design, testing, and validation of feedback processing algorithms to minimize bias concerning different countries of origin.