Many studies have presented computational models of musical structure, as an important aspect of musicological analysis. However, the use of grammar-based compressors to automatically recover such information is a relatively new and promising technique. We investigate their performance extensively using a collection of nearly 8000 scores, on tasks including error detection, classification, and segmentation, and compare this with a range of more traditional compressors. Further, we detail a novel method for locating transcription errors based on grammar compression. Despite its lack of domain knowledge, we conclude that grammar-based compression offers competitive performance when solving a variety of musicological tasks.

许多研究提出了音乐结构的计算模型，作为音乐学分析的一个重要方面。然而，使用基于语法的压缩器来自动恢复此类信息是一种相对较新且有前途的技术。我们使用近 8000 个分数的集合广泛调查了它们的性能，包括错误检测、分类和分割等任务，并将其与一系列更传统的压缩器进行比较。此外，我们详细介绍了一种基于语法压缩定位转录错误的新方法。尽管缺乏领域知识，但我们得出结论，基于语法的压缩在解决各种音乐学任务时提供了有竞争力的性能。