Đánh giá tập nhãn và xác định lỗi tự động trong kho ngữ liệu đã gán nhãn

• Main Author: Đỗ, Thị Thanh Tâm
• Format: Theses and Dissertations
• Language: other
• Published: Đại học Quốc gia Hà Nội 2016
• Subjects: Công nghệ thông tin; Khoa học máy tính; Tin học
• Online Access: http://repository.vnu.edu.vn/handle/VNU_123/8263
• Institution: Vietnam National University, Hanoi

The first part evaluates properties of tagset and possibility convertible of tagsets in Vietnamese. In the part, main goal is to optimize which tagset is better and whether small tagset can convert into large one and reverse. Thesis achieves this goal by using internal, external criteria and statistic of lost ambiguous token. Internal criterion tests whether token assigns POS accurately. External criterion checks quantity of linguistic information is retained. In particular, internal criterion relates to frame and purity notion. To investigate retained information, we carried out merging some tags based on certain classification factor. Each different tagset, we had different parameter. As the result, classification based on syntax has better result but ambiguous words are large. Besides, in Vietnamese, it is hard to convert between tagsets.