Human perceptions of nature, once the domain of the social sciences, are now an important part of environmental research. In ecology, cultural values have become a key component of ecosystem services, and in conservation, people’s perceptions can influence which species are traded, protected, and persecuted. This transdisciplinary shift has brought the human dimensions of nature into focus. However, the data and tools to tackle this research are lacking or are difficult to apply. For example, currently available approaches like sentiment analysis could view text describing the beauty of sharks as negative, simply because the word ‘shark’ has a negative connotation in these methods. Here, we present a collection of text classification models to measure public opinions on nature and hunting that were trained using an extensive dataset of social media messages from Twitter. These models allow us to identify text relevant to the broad topics of hunting and nature, describing whether opinions are pro- or against-hunting, or show interest, concern or fear of nature. The methods also include a biographical classification – describing whether the author of the text is a person, nature expert, nature organisation, or ‘Other’. The models are designed to support qualitative analysis of big data, but can be applied to smaller data problems. The models accurately classified biographies, text related to hunting and nature, and the stance towards hunting and nature (weighted F-scores: 0.79 - 0.99; 1 indicates perfect accuracy). All tested sentiment analysis methods failed to distinguish between hunting (e.g. pro- vs. against-hunting) and nature stances. These models are presented in the form of an R package classecol.