A Method for Identifying a Series of Programs in Japanese Television Broadcasting

Abstract

　This research focuses on a series of broadcast programs, compares and examines a wide range of elements such as URLs included in program information, and shows effective methods for identification. The proposed method considers the matching method for each field and the combinations between them. A series identification method was proposed on condition that either of two fold was satisfied, one condition sets the delimiter and matches the master title extracted from the title exactly. The other condition extracts the URL from the program details and matches it with considering the Jaro-Winkler distance to the title. Then, We created an evaluation dataset for 35 days of program information from September to October 2021, and the method was evaluated using the number of series, the precision and recall of the broadcast program pairs as metrics. The results showed that the method using the master title extracted from the title is effective. In addition, the results indicated that adding the Jaro-Winkler distance is useful for improving the accuracy of the identification condition by URL.