I have read Xiao's paper, thanks very much for introducing the paper to me. Could we explain it like this: semantic prosody is the affective meaning or connotational meaning produced when a node occurs together with its collacates.
I would entirely agree with you in your definition if the statement "or connotational meaning" is removed.
True, semantic prossodies are often conflated with connotations in the literature, but to my understanding, connotational meanings are inherent in lexcial items whereas semantic prosodies are essentially collocational.
"As with idioms, we can describe a complex lexical item holistically as a semantic unit whose meaning cannot be inferred from decomposing it into the smaller lexical items it consists of. It is, however, a matter of degree to what extent the meaning of a complex lexical item is independent of the meaning of the parts it is composed of. Sometimes it can be useful to describe a complex lexical item by assuming that its node is imbued with certain semantic (usually connotative) features inherent in the other elements that the complex lexical item consists of. This approach turns our attention to the phenomenon called semantic prosody (for connotative features) or semantic preference (for denotation features.)" (摘自wolfgang, My version of corpus linguistics)
A really good question indeed. The corpus-based approach is good at showing you what you see - or even what you never think you will see, but never tells you why that is so. You will need to use other methodolody and knowledge - maybe psychology and siciology in this case - to find out answers to such questions.
This is what Louw (2000: 52) says -
The fact that negative prosodies are more frequent than positive ones ought not to surprise us greatly. In the same way that unrequited love forms most of the subject matter for the greatest love poetry in English and not requited love (with the superb exception of John Donne's The Good Morrow), we ought not to be surprised to find that contented human beings utter much less than discontented ones. Besides, requited love is a fairly busy and time consuming state and one which is as inimical to the act of writing. However, if fracture and discord are the basis of unhappy contexts of situation, then the over-provision of intimacy, comfort, food, drink and possessions are likely to form the basis of contented situations. John Sinclair long ago indicated to the author (personal communication) that the author might find the positive prosodies clustering around forms such as warm.
The concordance for warmth and bears this out and creates scope for the investigation of positive prosodies with objectivity rather than schmaltz. However, it should be noted that this concordance for warmth and (Appendix 3) does not single out a semantic prosody, because its authorial function hardly rises above the banal and it fails too on the related criterion of pragmatism. It is a good example, however, of the strong provision of connotation.