Tagging conversations following the semantic approach
To address the limitations of the keyword-based approach to tag conversations, many NLP-backed solutions offer a more robust approach based on the meaning of the ticket content rather than the actual words that compose it. Such solutions use high dimensional vectors to represent words and sentences and can easily identify similarity and proximity of meaning or context. In more practical terms, two cases containing the utterances “I want my money back” and “I want an immediate refund” will both get tagged as a refund request.
This kind of processing can grasp the meaning of a ticket and do tag it according. It is naturally more reliable than the Zendesk automatic tagging, which can only tag a single word.