Abstract: Basketball is one of the most popular sports in the world, and its related events have attracted the attention of many sports fans which brought great social and economic benefits. However, the international top basketball events represented by the national basketball association (NBA), have thousands of games each year, and each game lasted more than two hours, which often deter the audience who are eager to effectively screen interested events or locate wonderful moments. Therefore, a test-based sonification system for basketball is designed. Firstly, keywords were sequentially extracted from the Play-by-Play text data and mapped to the corresponding melody, so as to realize the cross modal representation of data. Then, the music clips are generated from text data, which effectively solves the problem that the user's visual channel is occupied and cannot watch the basketball games, and also makes it possible for users to screen interested games in a short time. Taking the 2021 NBA playoffs as an example, the average music clip generation time for a single game only need 0.094 s, and the audio duration is reduced to 12.27% of the corresponding game video duration. The results of user study with 38 people showed that the proposed system not only reflects the game precisely, but also bring users with good experience.