Objective. The study aims to identify the features as well as the strengths and weaknesses of works that have produced semantic or thematic classifications of Hafez’s poetry.

Methodology. By examining the full text of fourteen printed works and five computer- and AI-generated classifications, the study evaluates the methods used to extract themes, categorize verses, and provide access pathways to the verses and topics.



Findings: Among the printed sources, only three works have classified all the verses. Naturally, the five computer sources faced no limitations in covering the entire corpus of verses, and their access pathways are fully available. However, in printed sources, due to the large amounts of data and space limitations, only three works included all the verses. A major shortcoming of most printed works is the lack of indexes, which reduces the quality of user access. Another weakness—seen in all five computer-generated works and in most printed ones—is the compilers’ reliance solely on the explicit meanings of the verses and their neglect of the implicit or hidden meanings. A third shortcoming is that the compilers have categorized the verses based on personal taste and prior assumptions, drawing little on the diverse—and sometimes conflicting—perspectives of Hafez scholars. Incorporating such perspectives could have strengthened the credibility of their classifications. Forming thematic categories in the five modern studies on Hafez’s poetry—based on the vocabulary present in the verses—is highly debatable. Moreover, the fact that at least four of the studies with modern approaches rely solely on Hafez, Hooman (1979) raises the suspicion that their researchers were unaware of other thematic classifications of Hafez’s poetry.



Conclusion: The role of both human and software-based factors (especially Artificial Intelligence) is significant in accomplishing the task of classifying poetry. However, the importance of each factor may shift at different stages. Undoubtedly, limitations in selecting verses—given current technological capabilities—are no longer relevant, and covering all verses has become simple. Regarding the assignment of meaning to verses, it appears that human thought and reasoning still outweigh software programs and even artificial intelligence, at least for now. Using authoritative commentaries that consider both levels of meaning (literal and implicit) offers an optimal solution, as combining the views of various interpreters provides audiences with diverse needs access to multiple perspectives on Hafez's ideas. Documentation and labeling of concepts, as well as forming thematic categories through the collaboration of human expertise and artificial intelligence, yield higher-quality results. Certainly, determining categories and their subcategories requires ongoing interaction between Hafez scholars and information science specialists. The extraction and classification of concepts and the construction of corpora—using the latest technological achievements—highlight the crucial importance of software and AI specialists in ensuring precision throughout the process. Drawing on the experience of existing classifications, one may hope that the way for creating of an ontology of Hafez’s poetry has grown shorter.