In this report, we had access to a limited number of variables related to each review. Therefore, we could not account for various possible confounders such as the time of review, the nationality of the reviewer, the difference between every drink of a single kind/style of beer and the differences in perceptions between people's experience of taste. We did attempt to look at the variables we did have access to such as the location of the review and the inherent quality of the beer (as estimated by overall rating). We also did not interpret the context of words. While this is not an issue for most descriptive words ('strawberry' will only ever refer to strawberry taste), it relevant for negations ('does not taste of strawberry'). However, we believe that this is not a major issue as the majority of descriptors are positive (taste like...).
We did not use any additional datasets, only the data provided to us for the project, therefore there are not ethical considerations related to data collection. We also believe that the ethical implications of this work are negligible, other than perhaps some disgruntled beer experts would be displeased with a scientific analysis of their precious art.