jieba最早是以Java所撰寫的開源程式[4]。在斷詞時會先建立字典

jieba最早是以Java所撰寫的開源程式[4]。在斷詞時會先建立字典樹(Trie)，以句子中每個字為字首，將所有延伸的詞彙找出，接著對句子生成Directed Acyclic Graph, DAG，根據辭典中詞彙的出現頻率，以動態規劃方法找出最大切分組合，屬於全切分方法，並使用隱藏式馬可夫模型來判斷未知詞，從其他可觀測資料來計算未知資料的機率，也就是說，可以透過字的序列狀態去推測句子的結構，將每種結構的排列組合都做比較，計算出最符合的斷詞位置，找出最佳的斷詞結構和未知詞。

結果 (英文) 1:

jieba is the earliest written Java open-source program [4]. When hyphenation will first establish trie (Trie), each word in the sentence prefix, extending all the vocabulary to find, then the sentence generating Directed Acyclic Graph, DAG, according to the frequency of occurrence of dictionary words to dynamic programming method to find the greatest combination of segmentation, are full segmentation method, and the use of hidden Markov models to determine the unknown word, to calculate the probability of an unknown data from other observations, that is, you can go through the sequence of the status word We speculated that the sentence structure, the permutations and combinations of each structure are compared to calculate the off position best suits the word, to find the best hyphenation structure and unknown words.

結果 (英文) 2:

Jieba was first written as an open source program in Java. A dictionary tree (Trie) is established when word breaks are established, with each word in the sentence prefixed, all extended words are identified, and then directed Acyclic Graph, DAG, based on the frequency of the appearance of words in the dictionary, to find out the maximum slicing combination in a dynamic planning method, which is a full-cut method. The hidden Markov model is used to judge unknown words and calculate the probability of unknown data from other observable data, that is to say, the structure of sentences can be inferred from the sequence state of words, the arrangement combinationofs of each structure can be compared, the most consistent word-breaking position, to find out the best word-breaking structure and unknown words.

結果 (英文) 3:

Jieba is the first open source program written in Java [4]. When breaking words, it will first establish a trie, prefix each word in the sentence, find out all the extended words, and then generate a directed acyclic graph for the sentence, DAG, according to the occurrence frequency of words in dictionaries, uses dynamic programming method to find the maximum segmentation combination, which belongs to the full segmentation method, and uses hidden Markov model to judge unknown words, and calculates the probability of unknown data from other observable data. That is to say, the sentence structure can be inferred through the sequence state of words, and the arrangement and combination of each structure can be compared Find out the most suitable word break position, find out the best word break structure and unknown word.