i am working on an analytical framework which i intend to classify responses into either beliefs, attitudes, values and emotions. the problem is coming up with clear distinction on the sub-domains is problematic.
You can refer to intensity, duration, having a known stimuli, and other scales. There is a plenty of research in this field, especially for emotions, feelings, and moods. Emotions, for example, are distinct from feelings since they are usually short-term, intensive, have an object or a reason as a stimuli. Fear (short-term, intensive, caused by specific (known) threat), being a basic emotion, is distinct from anxiety (long-term feeling, less intensive, initial stimuli is not clear or cannot be easily remebered).
I don't know how you want to combine emotions and values or beliefs. I assume you have a different understanding of these concepts. It would help, if you could define them here. I refer above to basic emotions in the sense of Damacio or Ekman.
I think the best way to go about this would be to prepare a large inventory of candidate items covering all of your hypothesised domains, and then run a pilot study to examine the way that people respond on these scales. You should then perform a Principal Components Analysis (PCA) on the data. A PCA is a type of exploratory factor analysis which should identify the factor structure of your data, and this will allow you to confirm if the domains you expect are present. Examination of the loadings of your items on these factors should allow you to identify whatever factors are extracted, and give you a sound foundation for deciding which domains to use, and which items to use for each domain in a refined version of your inventory.