CN105630770A - Word segmentation phonetic transcription and ligature writing method and device based on SC grammar - Google Patents

Abstract

The invention relates to a method and device for word segmentation, phonetic transcription and continuous writing based on SC grammar, and belongs to the technical field of machine translation in computer science. Firstly, the present invention is based on the word segmentation ambiguity rules of SC grammar, utilizes the adjacency constraints in natural language, and establishes an ambiguity segmentation rule base to eliminate illegal segmentation to improve word segmentation accuracy; secondly, based on the SC grammar's word segmentation continuation rule base and continuation corpus Statistical library, continuation corpus statistical library is used to continuation of those continuation knowledge that cannot be represented as rules; finally, based on the dictionary library of SC grammar, the dictionary is used for forward maximum matching to perform word segmentation, and ambiguous fields call word segmentation ambiguity rules to obtain correct Segment the result, analyze the context of the word to obtain correct part-of-speech tagging and transcription. Compared with the prior art, the invention improves the accuracy of word segmentation, and the word segmentation ambiguity rule base, combination ambiguity lexicon, ligature rule base, dictionary base and ligature corpus statistical base are easy to expand and maintain.

Description

A method and device for word segmentation, phonetic transcription and continuous writing based on SC grammar

technical field

The invention relates to a method and device for word segmentation and phonetic continuation, in particular to a method and device for word segmentation and phonetic continuation based on SC grammar in a Chinese-blind translation system, and belongs to the technical field of machine translation in computer science.

Background technique

Machine translation refers to the process of converting one natural language into another natural language expression by using a computer. The Chinese-blind translation system automatically translates Chinese information into Braille characters, which is of great help to the education and life of blind people. Braille is a special form of phonetic writing. To realize the translation of Chinese characters into Braille, the Chinese word segmentation and ligature should first be converted into Pinyin, and then converted from Pinyin to Braille. Therefore, the accuracy of Chinese word segmentation and phonetic transcription is very high. To a certain extent, it determines the accuracy of Chinese blind translation. Word segmentation is an important rule unique to Chinese Braille. Word segmentation is to write words one by one separately; continuous writing is to write some words together to avoid the loose syllable structure and make it easier to read according to the particularity of Braille. Word segmentation must follow the basic rules of Chinese grammar, language logic, habit and syllable length. In the process of converting Chinese to Pinyin, due to the problem of polyphonic characters in Chinese characters, polyphonic phenomena in words are much less than polyphonic phenomena in words, and words with more than three characters rarely have polyphonic phenomena, so correct word segmentation Can greatly reduce polyphony. However, the problem of individual polyphonic characters still exists. How to correctly transcribe polyphonic characters requires the use of contextual context for natural language analysis and processing. Therefore, there are two difficulties in the process of converting Chinese characters to Braille: 1. Improving the correctness of Chinese word segmentation and ligatures; 2. Combining the context analysis of the context to correctly mark polyphonic characters. Since the domestic translation from Chinese to Braille is still in the manual stage, in order to bring more and better educational materials to the blind, the heavy translation work has brought about a decrease in accuracy, so there is an urgent need for a set of translation from Chinese to Braille The high-accuracy word segmentation and phonetic transcription method lays a solid foundation for Chinese-blind translation.

Contents of the invention

The purpose of the present invention is to solve the problem of realizing Chinese-blind machine translation, and propose a method and device based on SC grammar to realize fast and accurate word segmentation and phonetic ligature.

The idea of the present invention is: 1, based on the word segmentation ambiguity rule of SC grammar, utilize the adjacency constraint condition in the natural language, establish the ambiguity segmentation rule base, improve word segmentation precision with getting rid of illegal segmentation; 2, the word segmentation based on SC grammar The ligature rule base and the ligature corpus statistics base, according to the particularity of Braille, avoid the syllable structure being too loose, which is convenient for the blind to touch and read, and write some words together. The continuous writing corpus statistical database is used to write continuous writing knowledge that cannot be expressed as rules; 3. The dictionary library based on SC grammar uses the dictionary to perform forward maximum matching to perform word segmentation, and the ambiguous fields call the word segmentation ambiguity rules to obtain correct segmentation As a result, correct part-of-speech tagging and transcription are obtained by parsing the context of the word.

The purpose of the present invention is achieved through the following technical solutions:

A method for consecutive writing of word segmentation, phonetic marking based on SC grammar, based on a dictionary database, a combined ambiguous thesaurus, a word segmentation ambiguity rule database, a consecutive writing rule database and a consecutive writing corpus statistical database, comprising the following steps:

(1) receive the Chinese character string and article genre type to be divided into phonetic symbols;

Described character string is pure Chinese character character string, is the character string that does not comprise the special symbols such as numeral, punctuation mark, ASCII code character; If comprise non-Chinese character character in the character string, it is segmented, the non-Chinese character character after segmentation Strings are processed separately. For example, word nodes are directly generated and corresponding types are assigned. For Chinese character strings, step (2) is combined with other processed non-Chinese character substrings and then output.

(2) Segment Chinese character strings based on the dictionary database, and perform part-of-speech tagging and phonetic marking on the word blocks after word segmentation;

(3) According to the article genre type, call the corresponding ligature rule base, and carry out combined ligature writing to the word block of step (2) based on the braille word segmentation ligature rule in the ligature rule base;

(4) Carry out secondary combined ligatures to the combined word chunks based on the ligature corpus statistical database;

(5) Output the generated Chinese character string after word segmentation and phonetic concatenation.

The dictionary database is used for Chinese word segmentation, part-of-speech tagging and phonetic notation, including Chinese word symbols, grammatical and semantic attribute identifiers, context distinguishing functions, and pinyin of words.

The dictionary database is constructed through the following process: a set of grammatical and semantic attribute classification system is defined according to the Chinese dictionary knowledge and included, and language engineers further improve it in the process of debugging the corpus.

The word segmentation based on the dictionary is completed through the following process:

a. Referring to the dictionary library, use the forward maximum matching algorithm to split the sentence to obtain word chunks;

b. Carry out cross ambiguity judgment according to the cross feature of the word block;

c. Carry out ambiguity judgment on word block based on combined ambiguity lexicon;

d. According to the rules of ambiguity, disambiguation is eliminated through reasoning;

e. Output word segmentation results.

The cross ambiguity is in the form of a string AXB, wherein AX constitutes a word, and XB also constitutes a word, and this type of ambiguity is the cross ambiguity. Wherein, the lengths of A, X, and B are greater than or equal to a word length. For example, "have time", "different situations", "big head" and so on all have cross-ambiguity.

Described combination ambiguity lexicon is used to identify the word block that there is combination ambiguity, and what included in the storehouse is the two-character word that there is combination ambiguity, and combination ambiguity word is the word string such as AB, and wherein A, B become words independently respectively, as "Future" in the sentence "he will come to Shanghai." is a combination of ambiguous words.

The combined ambiguous thesaurus is constructed through the following process: language engineers gradually include in the process of debugging a large number of corpus.

The word segmentation ambiguity rule base is used to reason and eliminate ambiguous word blocks to obtain correct word segmentation results, including ambiguous word blocks, conditional functions, and correct word segmentation operations.

The word segmentation and ambiguity rule base is constructed through the following process: language engineers gradually summarize and improve rules in the process of debugging a large number of corpus. The word segmentation ambiguity rule base is subdivided into two types: cross ambiguity rules and combination ambiguity rules. Word chunks with cross ambiguity call cross ambiguity rules for inference disambiguation, and word chunks with combination ambiguity call combination ambiguity rules for inference disambiguation.

The ambiguity judgment of the word block based on the combined ambiguous thesaurus is completed through the following process:

a. For the current word block, use the binary search algorithm to query the combined ambiguous thesaurus;

b. According to the query result, output the combined ambiguity flag.

According to the ambiguity rules, disambiguation through inference is accomplished through the following process:

a. Match the ambiguous part of the ambiguous word in the ambiguity rule for the current lexical block containing the ambiguous flag;

b. If the match is successful, check the conditional function;

c. If the condition check is satisfied, perform the correct word segmentation operation;

d. Output the correct word segmentation results.

The part-of-speech tagging and phonetic marking of the word chunk after the word segmentation is completed through the following process:

a. For the current word block, take out the dictionary information of the word block from the dictionary storehouse;

b. Check the context function item by item;

c. If the context check is satisfied, take out the part of speech and pinyin of the item.

The ligature rule library is used to combine and ligature the word chunks after word segmentation and labeling, including regular word chunk parts, conditional functions, and ligature operations. According to different article genres, the joint writing rule base is subdivided into the classical Chinese rule base and the modern Chinese rule base.

The ligature rule base is constructed through the following process: according to the ligature rules defined in Braille publications, the ligature rules are included one by one, and language engineers further improve during the process of debugging the corpus.

The combined writing of word blocks based on the writing rules is completed through the following process:

a. For several current word blocks, match the word block part in the linking rule;

b. If the match is successful, check the conditional function;

c. If the condition check is satisfied, perform the correct continuous writing operation;

d Output the word segmentation result after ligature.

The continuous writing corpus statistical database is used for performing combined continuous writing on the word blocks combined according to the continuous writing rules, and the library includes word blocks that need to be combined and continuous writing, such as "three major disciplines". The ligature corpus statistical database is subdivided into basic thesaurus and user thesaurus. The basic thesaurus includes some common ligature word blocks, and the user lexicon includes user-defined word blocks that need to be ligatured.

The ligature corpus statistical database is constructed through the following process: collecting some specific ligature word blocks defined in Braille publications, and further perfected by language engineers in the process of debugging the corpus.

The second combined ligature of the combined word block based on the ligature corpus statistical library is completed through the following process:

a. For the current word block, match according to the order of the user thesaurus and the basic thesaurus;

b. If the matching is successful, execute the continuous writing combination;

c. Output the word block result after continuous writing;

A word segmentation and phonetic continuous writing device based on SC grammar, based on a dictionary database, a combined ambiguous thesaurus, a continuous writing corpus statistical database, a continuous writing rule base and a word segmentation ambiguity rule base, including sequentially connected word segmentation modules, part-of-speech tagging and phonetic transcription modules, once The combined continuous writing module and the secondary combined continuous writing module, the word segmentation module, the part-of-speech tagging and the phonetic transcription module are respectively connected to the dictionary database, the word segmentation module is also connected to the combined ambiguous thesaurus and word segmentation ambiguity rule base, and the primary combined continuous writing module is connected to the continuous writing rule base , the secondary combined continuous writing module is connected with the continuous writing corpus statistical database;

The word segmentation module is used to segment the input Chinese character string based on the dictionary database, split it into independent word blocks, and judge whether there is ambiguity in the obtained word blocks based on the cross ambiguity features and the combined ambiguity lexicon during the segmentation process. The ambiguous words are based on the word segmentation ambiguity rule base to eliminate the ambiguity and get the correct word block;

The part-of-speech tagging and phonetic marking module is used to perform correct part-of-speech tagging and transliteration on the word chunks obtained by the word segmentation module based on the dictionary library through the context function check, so as to obtain the correct part of speech and pinyin of the word chunks;

The one-time combined ligature module is used to combine ligatures for part-of-speech tagged word chunks. This module obtains the combined lexical chunks by checking the conditional functions based on the ligature rule library;

The secondary combined ligature module is used to perform the query matching operation on the lexical corpus statistical database of the combined ligature to obtain the combined lexical chunks, and output the lexical chunks with part-of-speech annotation and phonetic transcription.

Preferably, the dictionary database, combined ambiguous thesaurus, consecutive writing corpus statistical database, consecutive writing rule database and word segmentation ambiguity rule database can be continuously modified and improved according to the development of the times, thereby improving the accuracy of word segmentation.

Beneficial effect

Braille is a special form of pinyin, so the accuracy of Chinese word segmentation and phonetic transcription largely determines the accuracy of Chinese-braille translation. The dictionary structure based on the SC grammar designed by the present invention improves the accuracy of the phonetic notation of polyphonic characters, the word segmentation and ligature rules based on the SC grammar improve the accuracy of the word segmentation, and the word segmentation ambiguity rule base, combined ambiguous thesaurus, ligature rule base, The dictionary library and the co-writing corpus statistics library are easy to expand and maintain.

Description of drawings

The present invention is described in detail below in conjunction with accompanying drawing and invention example:

Fig. 1 is a kind of schematic flow chart of the word segmentation phonetic transcription method based on SC grammar of the embodiment of the present invention;

Fig. 2 is a flow chart of word segmentation process;

Fig. 3 is the flowchart of part-of-speech tagging and phonetic transcription process;

Fig. 4 is the flowchart of participle continuous writing process;

FIG. 5 is a schematic diagram of the composition and structure of an SC grammar-based phonetic transcription device according to an embodiment of the present invention.

detailed description

The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

A kind of word segmentation phonetic transcription method based on SC grammar, flow process as shown in Figure 1, comprises the following steps:

⑴ Chinese character strings and article genre types that accept word segmentation and phonetic transcription;

Taking the accepted article genre as modern Chinese and the content of the Chinese character string as "In 2008, Xiao Li was promoted to be the chief engineer of this project" as an example, the implementation process of the method of the present invention will be described.

(2) Segment the Chinese character strings based on the dictionary database, and perform part-of-speech tagging and phonetic marking on the word blocks after word segmentation. As shown in Figure 2, this content is achieved through the following process:

2.1 Based on the dictionary, the Chinese character strings are subjected to the forward maximum matching, and word blocks are segmented.

Combining the maximum word length information in the dictionary and the maximum possible side length in the sentence, determine an optimal maximum side length N, and look it up in the dictionary. For example, the sentence "In 2008, Xiao Li was promoted to be the chief engineer of this project." The maximum word length of "year" in the dictionary is 3, because the longest word starting with year included in the dictionary is 3 characters long. The maximum possible side length of "year" in the sentence is 1, because there are non-Chinese characters behind it, so it is determined that the optimal maximum side length N of "year" in the sentence is 1. If there is such an N word in the dictionary, the match is successful, and the matching field is segmented as a word; if such an N word is not found in the dictionary, the match fails. The last Chinese character is removed from the matching field, and the remaining N-l characters are used as a new matching field for new matching, and so on until the segmentation is successful. That is, a round of matching is completed to segment a word. This goes on and on until all the words are segmented.

2.2 Word block ambiguity judgment

If the segmented word is more than one Chinese character, that is, N>1, then carry out the judgment of cross ambiguity, take the second Chinese character of the word as the beginning of the word, and take the word length >= N as the side length, and execute the above word Segmentation operation, if such a word can be found, it means that the intersection ambiguity exists, and the word segmentation ambiguity rule is called to reason and disambiguate. For example, when the above sentence is segmented into "item", "item" is used as the beginning of the word, and when the word length is 2, it is found that "purpose" is also a word, which shows that "item" has cross ambiguity.

If the current word length is greater than 1 and is 2, then the word may have combination ambiguity, and query the combination ambiguity thesaurus to determine whether it has combination ambiguity. For the example string, since "item" is not in the combined ambiguity corpus, "item" only has cross ambiguity. If "item" is in the combination ambiguity thesaurus, then "item" has both cross ambiguity and combination ambiguity.

2.3 Inference Disambiguation

According to the ambiguity flag type of the current word, the corresponding word segmentation ambiguity rules are invoked to disambiguate. The ambiguity rule base contains ambiguity segmentation rules in some specific words, parts of speech or attributes, such as combination ambiguity rules: "NP (future), NP (PLA) → DWD (A)", wherein, " NP (future), NP (PLA)" is the first part of the ambiguity rule, that is, the part of the ambiguous word block. "DWD(A)" is the third part of the ambiguity rule, that is, the correct word segmentation operation part, and the second part of the ambiguity rule in this rule, that is, the conditional function part is empty; this rule indicates that when the A word block "future" When following a B word block, namely representing the noun (NP) of place (PLA), this A word block will be cut off " DWD (A) ". For example, the sentence "he will come to Shanghai." After steps 2.1 and 2.2, it is found that "future" has combination ambiguity, the matching rule "NP(future), NP(PLA)→DWD(A)" is successful, and the correct segmentation of "future" is "future". The expression form of intersection ambiguity rules and combination ambiguity rules is the same, but the content is different. For the above sentence "item" has cross ambiguity, the cross ambiguity rule is invoked for inference disambiguation. There is no corresponding matching rule in the ambiguity rule base, but the word segmentation algorithm in the present invention is based on the maximum matching in the forward direction, so according to the principle of the longest forward direction, the correct word is segmented into "items".

Follow the above steps for the following Chinese character strings until all words are segmented. The chunks of the above sentence after segmentation are:

In 2008/year/, /Xiao/Li/promoted/for/general/engineer/of/this/a/project/. /

2.4 Part-of-speech tagging and phonetic transcription

As shown in Figure 3, it is the process of part-of-speech tagging and phonetic marking for a word block, specifically:

Query the dictionary for each Chinese character block, and retrieve the dictionary information of the word. For example, the first Chinese character block "year" in the current sentence is represented in the dictionary as follows:

$ year

TIM:(NCGEN,nian)S(L,(1,1),[AP;Q;WH;R]) "nian2"

AP: (AGEN) "nian2"

Among them, "$year" is the first part of the Chinese word, that is, the symbol part of the Chinese word. "TIM:(NCGEN,nian)" is the second part of Chinese words, that is, the grammatical semantic attribute identifier part; it indicates that "year" can be used as a time word (TIM) in a sentence. "S(L,(1,1),[AP;Q;WH;R])" is the third part of Chinese words, that is, the context discrimination function part. It says that if "year" is used as a time word (TIM) in a sentence, the first word to its left must be an adjective (AP) or a numeral (Q) or a question word (WH) or a pronoun (R). "nian2" is the fourth part of the Chinese word, which is the pinyin part of the word.

In the above sentence, "2008" is a numeral (Q), which satisfies the first clause of "year", and takes out the part of speech TIM and the pinyin "nian2". Proceeding in this way, the result of part-of-speech tagging and transcription of the sentence above is:

2008/Q/2008 /TIM/nian2, /BD/, small /AP/xiao3 Li /R/li3 promoted /VP/jin4sheng1 to /SV/wei2 this /R/zhe4 /L/ge4 project /NP/xiang4mu4 /DEF/de0 chief/AP/zong3 engineer/NP/gong1cheng2shi1. /BD/.

After part-of-speech tagging and phonetic marking are performed on the word block, word segmentation and co-writing will be carried out through the process as shown in Figure 4, as follows:

(3) According to the article genre type, call the corresponding ligature rule base, and carry out combined ligature based on the word block of step (2) based on the braille word segmentation ligature rule in the ligature rule base;

This is a modern text article, call the modern text continuous writing rules, and take out the word chunks after word segmentation from left to right. When the current word chunk is "2008/Q/2008", the matching rule is successful

S1{label:Q}S2{label:NP/L/TIM,length:1}||S1,S2

Among them, "S1{label:Q}S2{label:NP/L/TIM,length:1}" is the first part of the rule, that is, the part of the rule word block. It indicates that the first word block in the rule is a numeral (Q), and the second word block is a noun (NP) or a quantifier (L) or a time word (TIM) with a word length (length) of 1. The current rule does not have a conditional function. "S1, S2" is the third part of the rule, that is, the linking operation part, which indicates that word blocks S1 and S2 need to be linked together. Therefore, the word block "2008/Q/2008/TIM/nian2" needs to be written consecutively. The new word block after the continuous writing is expressed as "2008/QCH/2008nian2", and the QCH mark indicates that the word block is a word block after the continuous writing. Take out the next possible ligature word block "小/AP/xiao3", match the ligature rules, and execute the above steps in order to get a combined ligature word block:

In 2008 /QCH/2008nian2, /BD/, Xiaoli /QCH/xiao3li3 was promoted from /VP/jin4sheng1 to /DEF/de0 chief engineer /QCH/zong3gong1cheng2shi1 of /SV/wei2 /QCH/zhe4ge4 project /NP/xiang4mu4. /BD/.

(4) Carry out secondary combination and continuous writing of the combined word blocks based on the continuous writing corpus statistical database;

From left to right, take out the word blocks that have been combined once and then match the word blocks in the user lexicon and the basic lexicon according to the longest optimal principle. .

In 2008 /QCH/nian2, /BD/, Xiao Li /QCH/xiao3li3 was promoted from /VP/jin4sheng1 to the /DEF/de0 chief engineer /QCH/zong3gong1cheng2shi1 of the /SV/wei2 /QCH/zhe4ge4 project /NP/xiang4mu4. /BD/.

⑸ Output the generated Chinese character string after word segmentation and phonetic concatenation.

Based on the above-mentioned SC grammar-based word segmentation and phonetic continuation method, a device for word segmentation and phonetic continuation based on SC grammar has been realized, as shown in Figure 5. It can be seen from the figure that the device is based on dictionary database and continuation corpus statistics Literacy rule library, combination ambiguity thesaurus and word segmentation ambiguity rule library, including word segmentation module, part-of-speech tagging and phonetic transcription module, first-time combined ligature module and secondary combination ligature module, word segmentation module, part-of-speech tagging and phonetic notation module respectively with the dictionary The word segmentation module is also connected to the combined ambiguous thesaurus and the word segmentation ambiguous rule base respectively, the first combined continuation module is connected to the continuation rule base, and the secondary combination continuation module is connected to the continuation corpus statistical database;

The word segmentation module is used to segment the input Chinese character string based on the dictionary database, split it into independent word blocks, and judge whether there is ambiguity in the obtained word blocks based on the cross ambiguity features and the combined ambiguity lexicon during the segmentation process. The ambiguous words are based on the word segmentation ambiguity rule base to eliminate the ambiguity and get the correct word block;

The part-of-speech tagging and phonetic marking module is used to perform correct part-of-speech tagging and transliteration on the word chunks obtained by the word segmentation module based on the dictionary library through the context function check, so as to obtain the correct part of speech and pinyin of the word chunks;

The one-time combined ligature module is used to combine ligatures for part-of-speech tagged word chunks. This module obtains the combined lexical chunks by checking the conditional functions based on the ligature rule library;

The secondary combined ligature module is used to perform the query matching operation on the lexical corpus statistical database of the combined ligature to obtain the combined lexical chunks, and output the lexical chunks with part-of-speech annotation and phonetic transcription.

As time goes by, people will constantly change the existing usage of this term and create new words, so the dictionary database, combined ambiguity thesaurus, ligature corpus statistical database, ligature rule base and word segmentation ambiguity rule base can all be maintained , so that it can be constantly changed and improved according to the development of the times, so as to improve the accuracy of word segmentation.

Experimental results

The word segmentation, phonetic and continuous writing method based on SC grammar effectively solves the problems of Chinese word segmentation ambiguity, continuous writing and correct phonetic marking of polyphonic characters in the process of Chinese-blind conversion, and realizes the efficient and intelligent translation conversion from Chinese to Braille. The translation accuracy rate is higher than 90%.

The invention adopts artificial intelligence technology, organically integrates various analysis and processing strategies such as rules and examples, efficiently and accurately performs word segmentation, phonetic transcription and continuous writing on Chinese sentences, and improves the correctness of Chinese-blind translation. The present invention designs a rule expression language based on SC grammar, good expansibility, high expression efficiency, and humanization. The rule expression has universal applicability and can be extended to solve other natural language processing problems.

Claims (10)

1. A method for continuous writing of word segmentation and phonetic marking based on SC grammar, characterized in that: based on dictionary database, combined ambiguous thesaurus, word segmentation ambiguity rule base, continuous writing rule base and continuous writing corpus statistics base, comprising the following steps: Step 1, receiving the Chinese character string and article genre type to be divided into phonetic symbols; Step 2. Segment the Chinese character string based on the dictionary database, and perform part-of-speech tagging and phonetic marking on the word blocks after the word segmentation; Step 3. According to the type of article genre, call the corresponding ligature rule base, and combine the lexical blocks in step (2) based on the braille word segmentation ligature rules in the ligature rule base; Step 4, based on the continuation corpus statistical library, perform secondary combination continuation on the combined word blocks; Step 5, output the generated Chinese character string after word segmentation and phonetic concatenation. 2. a kind of word segmentation phonetic marking method based on SC grammar according to claim 1, is characterized in that, described dictionary storehouse is used for Chinese word segmentation, part-of-speech tagging and phonetic marking, comprises Chinese word symbol, grammatical semantic attribute identifier , context discrimination function, pinyin of words. 3. a kind of word segmentation phonetic transcription method based on SC grammar according to claim 1, is characterized in that, described word segmentation based on dictionary storehouse is finished by following process:: a. Referring to the dictionary library, use the forward maximum matching algorithm to split the sentence to obtain word chunks; b. Carry out cross ambiguity judgment according to the cross feature of the word block; c. Carry out ambiguity judgment on word block based on combined ambiguity lexicon; d. According to the rules of ambiguity, disambiguation is eliminated through reasoning; e. Output word segmentation results. 4. a kind of word segmentation phonetic transcription method based on SC grammar according to claim 3, it is characterized in that, described combination ambiguity lexicon is used for identifying the word block that there is combination ambiguity, what included in the storehouse is that there is combination ambiguity word. 5. according to claim 3-4 arbitrary described a kind of word segmentation phonetic continuation method based on SC grammar, it is characterized in that, described word segmentation ambiguity rule base is used for reasoning and eliminating ambiguity word block, obtains correct word segmentation result, comprises Ambiguity word blocks, conditional functions, and correct word segmentation operations. According to the ambiguity rules, disambiguation through reasoning is completed through the following process: a. Match the ambiguous part of the ambiguous word in the ambiguity rule for the current lexical block containing the ambiguous flag; b. If the match is successful, check the conditional function; c. If the condition check is satisfied, perform the correct word segmentation operation; d. Output the correct word segmentation results. 6. a kind of word segmentation phonetic marking method based on SC grammar according to claim 1, is characterized in that, described part of speech tagging and phonetic marking are carried out to the word block after word segmentation and finish by following process: a. For the current word block, take out the dictionary information of the word block from the dictionary storehouse; b. Check the context function item by item; c. If the context check is satisfied, take out the part of speech and pinyin of the item. 7. a kind of word segmentation phonetic transcription method based on SC grammar according to claim 1, is characterized in that, described continuation rule storehouse is used for word segmentation and the lexical block after labeling carries out combination continuation, comprises regular word block part, Conditional function, continuous writing operation; according to different article genres, the continuous writing rule base is subdivided into a classical Chinese rule base and a modern Chinese rule base; the combined joint writing of word blocks based on the continuous writing rules is completed through the following process: a. For several current word blocks, match the word block part in the linking rule; b. If the match is successful, check the conditional function; c. If the condition check is satisfied, perform the correct continuous writing operation; d. Output the word segmentation results after ligature writing. 8. a kind of method for continuous writing based on SC grammar according to claim 1, it is characterized in that, described continuous writing corpus statistics storehouse is used to carry out secondary combination continuous writing to the word block after combining according to continuous writing rule, in library The words that need to be combined are collected; the ligature corpus is subdivided into basic thesaurus and user thesaurus, in which the basic thesaurus contains some general ligature words, and the user lexicon includes user-defined word blocks that need to be ligatured ; The combined lexical block is carried out through the following process to carry out the secondary combined continuation based on the continuation corpus statistical library: a. For the current word block, match according to the order of the user thesaurus and the basic thesaurus; b. If the matching is successful, execute the continuous writing combination; c. Output the word block result after writing. 9. A device for continuous writing of word segmentation, transcription and transcription based on SC grammar, characterized in that, based on a dictionary database, a combined ambiguous thesaurus, a continuous writing corpus statistical database, a continuous writing rule base and a word segmentation ambiguity rule base, including sequentially connected word segmentation modules and part-of-speech tags And the transcription module, the first-time combination continuation module and the second-time combination continuation module, the word segmentation module, the part-of-speech tagging and the transcription module are connected to the dictionary database respectively, and the word segmentation module is also connected to the combination ambiguity lexicon and word segmentation ambiguity rule library respectively, and the first-time combination continuation The module is connected with the ligature rule library, and the secondary combined continuation module is connected with the continuation corpus statistical database; The word segmentation module is used to segment the input Chinese character string based on the dictionary database, split it into independent word blocks, and judge whether there is ambiguity in the obtained word blocks based on the cross ambiguity features and the combined ambiguity lexicon during the segmentation process. The ambiguous words are based on the word segmentation ambiguity rule base to eliminate the ambiguity and get the correct word block; The part-of-speech tagging and phonetic marking module is used to perform correct part-of-speech tagging and transliteration on the word chunks obtained by the word segmentation module based on the dictionary library through the context function check, so as to obtain the correct part of speech and pinyin of the word chunks; The one-time combined ligature module is used to combine ligatures for part-of-speech tagged word chunks. This module obtains the combined lexical chunks by checking the conditional functions based on the ligature rule library; The secondary combined ligature module is used to perform the query matching operation on the lexical corpus statistical database of the combined ligature to obtain the combined lexical chunks, and output the lexical chunks with part-of-speech annotation and phonetic transcription. 10. a kind of word segmentation and phonetic continuous writing device based on SC grammar according to claim 9, is characterized in that, described dictionary storehouse, combination ambiguity lexicon, continuous writing corpus statistics base, continuous writing rule base and word segmentation ambiguity rule base all can be Maintenance, so that it is constantly changed and improved according to the development of the times, so as to improve the accuracy of word segmentation.