If the content to be verbalized is available in an unambiguous formal representation, standard output templates can often be used for generation in sufficiently narrow domains

Despite this inverse relationship, language generation has traditionally received less attention from computational linguists than language understanding, because if the content to be verbalized is available in an unambiguous, formal representation, standard output templates can often be used for generation, at least in sufficiently narrow domains. Even for unrestricted domains, the transduction from an explicit, unambiguous internal semantic representation to a word sequence is much less proble