방송이용의 특성을 고려한 어문저작물의 유사도 판단 모델의 법정책적 개선점 연구

In broadcast content, speakers’ utterances may include the literary works of others, and to ensure proper copyright protection, it is necessary to verify such usage and secure supporting evidence. Accordingly, a system for searching works and calculating similarity has been developed. Issues regarding the use of works identified can be resolved through permission from the broadcasting company or agreements with copyright management organizations. In other cases, the use may be considered fair. However, It should be noted that this system could potentially be transformed into an auditing system in broadcasting. Furthermore, when identical or similar expressions appear across multiple sources, the system should clearly identify their origins and provide information on the chronological order of the materials and the attribution of authorship. Additionally, when humans create works using generative AI, it may be necessary—either legally or technically—to identify which parts were generated by AI. It would also be desirable for the system to include features that distinguish non-infringing elements within a work, such as the merger of ideas and expressions, short and simple phrases, essential scenes, and content from the public domain. Although the current system does not yet account for non-verbal similarities or comparisons based on the structure, sequence, and organization (SSO), it could be enhanced to present such information in a structured format. The significance of the similarity calculation system lies in its ability to appropriately identify potentially problematic elements. While it represents an advancement over existing similarity detection programs, its results are based on quantitative sentence-level comparisons, which limits its ability to determine actual copyright infringement. Therefore, a final judgment on infringement must be made by human experts following the system’s analysis.