PISIT' S THAI NATURAL LANGUAGE PROCESSING LABORATORY
This lab is formed since August 26, 1998
e-mail: [email protected]
For C7 members, please check this C7 address list.
KEYWORDS
Thai Natural Language Processing Lab., words
segmentation, dictionaries, algorithms, Thai text-to-speech.
����ѧ���������§�ٴ����������Ѻ�ӷ��������ѡ
Morphological Derivative for Unknown Words in Thai text-to-speech Synthesis
Pisit Promchan,
Wittaya Wongvachirapanich,
Saanti Chinnakarn
[Full paper in pdf format]
���Ѵ���: �����������ʹͼš���Ԩ����оѲ�����ҧ�к�����ѧ���������§�ٴ
�����·��͡�ҡ�դ�������ö㹡�þٴ������ҹ�ӷ��� � �������� �ѧ�դ�������ö�
��üѹ���§���͡����ҹ�ӷ��������ѡ���ͤӷ������ҡ�㹾��ҹء���� ����дѺ���������
(Precision) �٧�ҡ������÷������ ����ҹ�Ԩ�������ء������ѡ�ǤԴ��Ѫ�ҵ�á��觤���
�������� (Fuzzy Logic) ������ͧ�����ѡ㹡����ѭ��㹡���ѧ������ӷ��������ѡ����ҹ��
㹧ҹ�Ԩ�¹������ʹ��Ƿҧ���ͧ�鹢ͧ��á��觤����������� ʶһѵ¡����ͧ�к������
�͡Ẻ��� ��鹵�Ըա�÷�Ǩ����Ҥ ��û���ҳ˹������§�����§���͢�鹵�ͧ��äӳ
ǹ���Ѵ��ǹ�ͧ�������Ҫԡ��絢ͧ������������ ��мš�÷��ͺ�����Թ����Է��
�Ҿ�ͧ�к� �����㹡óբͧ�����·��� � �к��դ������������� 99.59% ���㹡ó�
�ͧ��ͤ�������Сͺ���¤ӷ��������ѡ�������к��դ������������� 96.69% ���˹觷����
��������������������ǡѹ�դ�����ͺ���� (Recall) �٧�ش��ͻ���ҳ 98% ��� 88%
㹡óբͧ�����·��� � ���㹡óբͧ��ͤ�������Сͺ���¤ӷ��������ѡ����������ӴѺ
Abstract: This paper presents the morphological derivative for unknown words in Thai
text-to-speech synthesis. The research methodology is based on the philosophy of fuzzy
logic theory. The paper contains the basic idea of the fuzzy logic, the system
architecture, parsing algorithm, approximation matching algorithm for Thai or the fuzzy
value calculation. The experimental and performance evaluations are also included. It is
found that the system performs up-to 99.59% of precision in case of the Thai text data
contain both known and unknown words. The system perform up-to 96.69% in term of
precision in case of the Thai text data contains purely unknown. The highest intersection
point between precision and recall are about 98% and 88% for Thai text data contain
both known and unknown words and the Thai text data contains purely unknown
respectively.
Keywords: Thai Text-to-Speech, Synthesis, Parsing, Unknown Word Identification,
Approximate Matching, Algorithms, CTI, Fuzzy Logic
This page hosted by 
�
Get your own Free Home Page