2000 Volume 14 Issue 1
Published: 15 February 2000
-
Select all|
-
Review2000, 14(1): 1-6.The dictionary mechanism serves as one of the basic components in Chinese word segmentation systems. Its performance influences the segmentation speed significantly. In this paper , we design and implement three typical dictionary mechanisms ,i. e. binary-seek-by-word ,TRIE indexing tree and binary-seek-by-characters ,from word segmentation point of view , and compare their space and time complexity experimentally. It can be seen that the binary-seek-by-characters model is the most appropriate one being capable of fulfilling the need for speed of practical Chinese word segmenters to the maximum extent .
-
Review2000, 14(1): 7-12.This paper dynamically builds parameter table and threshold by extracting and analyzing usage frequency of characters of Chinese names based on large-scale corpus and researches evaluation function for Chinese name recognition. And it presents the method of Chinese name recognition without text segmentation. After open test , the recall rate and precision rate are respectively 95.23% and 87.31%.
-
Review2000, 14(1): 13-21.This paper presents a kind of Chinese parsing method which takes the DOP technique as the basic frame and utilizes the similarity-based probability estimate technique. In the implementation , every input sentence must by preprocessed through the initial selection in word level and part-of-speech level. Then the fragment-combination forms of the input sentence are acquired based on the constructed knowledge source which includes treebank , fragment-bank and fragment-combination-bank. Finally , the similarity estimate between the input sentence and the initial selection result is proceeded by using the similarity-based probability estimate technique. So the combination parsing process of the input sentence can be completed successfully. To prove the efficiency of the proposed method , the knowledge source is constructed based on the real-world Chinese corpus which involves 1,000 Chinese sentences , and the other real-world Chinese corpus which includes 100 Chinese sentences is used as the test set . The experiment result shows that every test parameter is satisfactory and the parsing process can be implemented efficiently.
-
Review2000, 14(1): 22-26.This paper puts forward a binarizing method named double-threshhold methodology which can eliminate the frame line effectively. And also a 3-lines-first-single-line-later method to segment the charaters accurately.
-
Review2000, 14(1): 27-31.A general parallel algorithm is proposed in this article. The algorithm can be used in many multi-datablock processing system. It can speed up the system doubled and redoubled. The core of the algorithm is fake MMX technology and it has no restriction to the machine ,which is illustrated in detail. In this paper ,the application of the algorithm in pattern transformation method with cosine function is also discussed.
-
Review2000, 14(1): 32-38.Association is one of the most important methods for chinese character input ,it can improve the input speed. We brought forward the intelligent association concept in this paper ,analysed the principle and problems encountered for intelligent association. We constructed 3 kinds of intelligent association schemes ,make use of the relationship between chinese words or characters. The average correct association rate is up to 40%.
-
Review2000, 14(1): 39-47.This paper addressed the construction of a dynamic-alphabet order-0 model of Chinese text for arithmetic coding and provided an algorithm of Chinese text compression. The model had shown to be of high performance because the algorithm driven by it could compress Chinese texts more efficiently than those that combined both LZ and Huffman coding. Because order-0 model laid the foundation of order-n models ,what the paper discussed was important to the text compression of any large-alphabet natural language ,such as Chinese ,Japanese and Korean.
-
Review2000, 14(1): 48-55.In this paper the encoding technique for Java is analyzed and the Chinese problems which lie in the current Java developing tools are discussed. Then a new method in which the Chinese code recognition and conversion techniques is adopted is presented to solve the Chinese incompatibility for different Sun JDK version. At last the experiments prove the method is feasible.
-
Review2000, 14(1): 56-64.On the basis of coded character sets for Tibetan information processing ,we need to discuss the sequence of Tibetan words and the method of making sequence. First , we put forward a concept of construction sequence for Tibetan words , which is much different from the sequence of Tibetan transliteration. There are two sequences in Tibetan , one is word order in dictionary , the other is character order in words as well as in a dictionary , and both of them make the structure inconceivable complex with multi-hierarchies. Secondly , we design a wit programme to assign distinctive values to all possible characters and coding characters , which make the sequence of Tibetan words to be realized in computer. Furthermore , we announce our diagram of technological process. Academic press limited.
