| Summary: | 碩士 === 東吳大學 === 資訊管理學系 === 103 === With the advancement of Internet technologies and prevalence of mobile communication devices, diverse types of information can now be obtained more easily and efficiently than before. However, processing a large amount of information is time consuming. This study proposes using Domain Word to produce a summary of financial news that are often read by investors. This approach extracts important sentences in each news article and presents them in simple and concise key points. Thus, readers can read news quickly, thereby reducing their reading time.
In this study, study and compare three methods. The domain word approach proposed in this study was adapted from two methods, namely, examining sentences and titles that overlapping with each other, and TF-IDF vocabulary scoring. The proposed method was subsequently assessed by comparing a summary manually produced by finance specialists with a summary automatically generated through the proposed method. The results indicated that the method proposed in this study yielded an F-measure of more than 79%, demonstrating the best performance compared with the other two methods. This study verified that the domain word approach can effectively produce a highly precise news summary.
|
|---|
