CN1038365A - Pinyin sequence is to the conversion method of Chinese character string (phrase) - Google Patents

Abstract

Pinyin sequence is to the conversion method of Chinese character string (phrase), based on Chinese-character pronunciation database and Chinese-character words database, set up thereon and finished automatic word segmentation, homonym is debated the converter of the pinyin sequence of knowledge to Chinese character string, be aided with spell check, automatically memory, making code for composing word, technology such as conversion repeatedly, the user can select to use the four tones of standard Chinese pronunciation of Chinese character, with further raising input speed.Debate at homonym and to have considered the collocation of word frequency, speech speech in the knowledge, close on factors such as environment, prefix and suffix, proper noun, (name, place name), the repetition rate of coding, the user that have reduced homonym significantly import continuously with the Chinese phonetic alphabet of standard in use, need not word of every input and promptly stop and select.

Description

Pinyin sequence is to the conversion method of Chinese character string (phrase)

The invention belongs to computer Chinese information input treatment technology.

Existing Chinese phonetic code keyboard input technology mainly is based upon in the input of single Chinese character, though there are some Chinese character processing systems to introduce the phrase input technology, phrase only is used as a kind of special circumstances of single Chinese character and goes to handle; And,, used " the special simplification " phonetic mnemonic symbol in the prior art in order to improve input speed relatively, as ang h, eng g, ah a, ch replaces or the like with i.The result who does like this be keyboard input technology more by specialization, common, do not use through the user of specialized training and feel very inconvenience; And because Chinese character syllable is few, number of words is many, the repetition rate of coding is high, input speed is difficult to be improved.

In the retrieval of before the present patent application, carrying out, obtained the 87102586th and No. 87108006 application that Nippon Electric Company, Ltd. proposes in China, her rattan of inventor has proposed the continuous input of the Chinese phonetic alphabet and the cutting techniques of Chinese phrase in these two application cases, and her rattan has proposed to be listed as the trigger pip (being the phrase list separator) that the Chinese character string converter is started working with the very high syllable of special partition, punctuation mark, occurrence frequency (Chinese character) etc. as starting phonetic.Deviser of the present invention thinks, use special division mechanism requirement user to understand the defined whole phrases of imput process system, memory capacitance is too big, and use the very high syllable of occurrence frequency as the phrase separator, do not consider the making words rule of Chinese phrase, artificial destruction a large amount of existing Chinese vocabularies.As " " the separation rate be 99.8, but under " come " and " settling " two kinds of situations " " group speech direction different fully.

In view of the above-mentioned deficiency of prior art, deviser of the present invention attempts to finish the conversion method design of a kind of pinyin sequence to Chinese character string (phrase), at first the input of single Chinese character is converted into the input based on Chinese character string (phrase); Minimizing user's as much as possible recall info is until reducing to zero; Simultaneously improving input speed as an important purpose of design; And the standard Chinese phonetic plan that when the user imports, adopts country to issue; The user can select to use the four tones of standard Chinese pronunciation.

Deviser of the present invention is after having carried out a large amount of fundamental researchs to Chinese character and Chinese phrase, proposed based on Chinese-character pronunciation database and Chinese terms database, set up thereon and finish automatic word segmentation, homonym and debate pinyin sequence that knowledge etc. handles converter to Chinese character string (phrase), and comprise spell check, automatic memory, making code for composing word, repeatedly the pinyin sequence of technology such as conversion is to the conversion method of Chinese character string (phrase).

1. Chinese-character pronunciation database

Comprise 6763 Chinese characters among the GB GB2312-80 in the Chinese-character pronunciation database, left the ability of handling unsimplified Hanzi simultaneously.

Contain in the Chinese-character pronunciation database: the usage frequency of the initial consonant of the Chinese phonetic alphabet, simple or compound vowel of a Chinese syllable and tone and Chinese character.

The data structure of Chinese-character pronunciation:

That is:

Chinese character::＜initial consonant＞＜simple or compound vowel of a Chinese syllable＞＜four tones of standard Chinese pronunciation＞＜usage frequency＞

As:

Existing: :=＜x＞＜ian＞＜4＞＜0115＞

2. term database

Having collected in the term database with " modern Chinese dictionary " is 60,000 speech of source, has added the word of more than 1,000 custom simultaneously.

Contain in the term database: entry, part of speech, semantic item and usage frequency.

The data structure of word:

That is:

Entry: :=＜the speech sound＞＜part of speech＞＜semantic item＞＜usage frequency＞

As:

Call: :=＜diaoyong＞＜V＞＜HJ28＞＜10191＞

Part of speech V verb

Semantic item H character 1 activity class

J character 2 life aspects

28 character, 3 meaning of a word are for using, be suitable for, use for reference

(1) entry: surplus in the of totally six ten thousand.Be divided into two parts: 30,000 of general term; 30,000 of general speech.

(2) the part of speech item comprises:

Notional word: noun, verb, adjective, number, measure word, pronoun.

Function word: adverbial word, preposition, conjunction, auxiliary word, modal particle, onomatopoeia.

Wherein each also is divided into some subclasses, as:

The subclass of noun is:

Termini generales, proper noun, time noun, term position

The subclass of verb is:

General verb, there are verb, directional verb, link-verb

(3) comprise under the semantic item:

Character 1, character 2,

Character as noun is:

Show abstract, the table material, the table material, the table characteristic

(4) usage frequency:

Usage frequency has characterized entry characteristic in use.When homonym occurs, arrange the sequencing of its appearance to use frequency.

3. automatic word segmentation

The user imports the standard pinyin sequence continuously by keyboard, when inputting punctuation mark or after importing the pinyin sequence of 8 Chinese characters continuously, carries out participle to realize correct retrieval by automatic word segmentation technology.Maximum matching algorithm is adopted in automatic word segmentation, from reverse pinyin sequence is divided, and is mated with term database, provides selectable various result.

Segmenting method: the speech syllable stream of keying in from keyboard is

S＝S1S2…Si

Wherein: Si is certain Chinese character syllable

The participle flow process is:

4. homonym is debated knowledge

The subject matter of phonetic transformation system software design is how to improve the accuracy of phonetic to the Chinese character conversion.The characteristic voice of Chinese is that syllable is few, has a large vocabulary, and that is to say that unisonance words amount is big.The key that improves the conversion accuracy is exactly the knowledge of debating that improves homonym, and this just is dependent on the knowledge base that word is described.

(1) word frequency:

Described the usage frequency of entry in aforesaid term database, this frequency promptly is the fixedly word frequency of entry; The user is when the manuscript of a certain professional domain of input, and entry shows dynamic temporary transient word frequency again, and this temporary transient word frequency of this technical notes when meeting homonym, can provide the priority rank of homonym according to the word frequency data of transient state on screen.

(2) speech speech collocation:

Has certain collocating rule between speech and the speech.This technology is set up numercal description to this morphology.Speech A and speech B have the collocation relation, and then A and B just have the R of relation.When homonym occurs, judge the entry of homonym appearance before and the entry that homonym occurs afterwards, do not have the unisonance entry that concerns R to get rid of.

Definition: can constitute word as two speech A and B, say that then they have collocation and concern and be described as R:

Phrase: :=＜A＞＜B＞

The collocation of a speech and other speech has the certain grammar semantic relation.Example:

If: the NP noun phrase

AP adjective phrase

The DP adverbial phrase

The VP verb phrase

The PP prepositional phrase

N noun D adverbial word

V verb P preposition

A adjective C conjunction

M number H auxiliary word

Q measure word T modal particle

R pronoun L onomatopoeia

A noun phrase (NP)

Version:

NP：：＝＜N＞

NP：：＝＜NP＞＜NP＞

NP：：＝＜AP＞＜NP＞

NP：：＝＜MP＞＜NP＞

NP：：＝＜NP＞＜C＞＜NP＞

NP::=＜NP＞" "＜NP＞

NP::=＜VP＞" "＜NP＞

NP::=＜PP＞" "＜NP＞

B verb phrase (VP)

Version:

VP：：＝＜V＞

VP：：＝＜DP＞＜VP＞

VP：：＝＜VP＞＜C＞＜VP＞

Example:

C adjective phrase (AP)

Version:

AP：：＝＜A＞｜＜M＞｜＜Q＞｜＜R＞

AP：：＝＜AP＞＜AP＞

AP：：＝＜DP＞＜AP＞

AP：：＝＜AP＞＜C＞＜AP＞

Example:

D adverbial phrase (DP)

Version:

DP::=＜D＞" "

DP：：＝＜DP＞＜DP＞

Example:

E prepositional phrase (PP)

Version:

PP：：＝＜P＞

PP：：＝＜P＞＜NP＞

Example:

Example 1:

Import from keyboard

jiaojiandebufa

Through after the word segmentation processing

jiaojian de bufa

Retrieve homonym

jiaojian de bufa

Strong and vigorous paces

Tiptoe is no lack of

Constitute rule according to phrase and carry out reasoning, the version of judging this phrase is:

NP::=＜AP＞" "＜NP＞

Then get rid of the possibility that " tiptoe " and " being no lack of " exists., and provide correct or selectable result.

Example 2:

Import from keyboard

feichangmeili

Through after the word segmentation processing

feichang meili

Retrieve homonym

feichang meili

Very beautiful

Glamour

According to conditional inference, adjective connects after the adverbial word; Adverbial word generally can not modification noun, then can get rid of the possibility that " glamour " exists.

Constitute the rule analysis by phrase, can get rid of a large amount of homonyms, but still not be whole.This is because in Chinese, though different parts of speech has different taxemes, has a few words can have the taxeme of two or more part of speech, and tangible contact is arranged again on meaning, the concurrence phenomenon of Here it is speech.

(3) close on environment:

In the sentence of Chinese, some word always appears at the beginning or the end of sentence, gives specific sign these words, when conversion sign is carried out conversion and provides correct result.As:

zhe na mon

Syllable is when in short beginning occurs, and system changeover goes out:

This certain

The pronoun of giving instruction is used.

zhe le guo

Syllable is during attached to the back of speech, phrase or sentence, and system changeover goes out:

Mistake

Doing auxiliary word uses.

(4) preceding add composition and after add composition:

In Modern Chinese, primary word adds go forward addition part or the loyal ㄓ of back addition as root and grows the high extra large ┡ commonplace thumb of  triangular bream of copying of beating with a stick of soul and look an ancient spear 4 and think borders and walk lamely and add composition after adding composition and 97 before meaningful 3 of Xin's a tree, used in making timber for boats, to preceding add composition and after add composition and provide sign, provide correct result during conversion.As

Bandaoli-→ half+conductor-→ semiconductor

Before add composition (prefix)

Little one is not good on not having very much

High the old and new do not have many down very much

Each is outer do not have just entirely can in

Bai Changzuo very is subjected to most again

Redder half black appliances hang down iron

Total this anti-soil woman wood is ancient

Main old former its pair of weak point should

Anti-certain secondary non-cotton Dan Chao

The bad man in ocean makes a noise sub accurate pure

After add composition (suffix)

People's the state's head of having gone up exists

In Nian Jiaxin hand water

Member's power under the ground property machine

Person's sound car is in the backization face

Wind be preceding method army

Party outside portion's book day eye limit

The district's opinion of shutting up some amount already

In team's thing place degree chamber device

Stage body system circle office of skin factory

The worker sends product class power gold ribbon

Station goods institute of society sheet class table

Family, rate village is fallen newspaper and is seen woods

The other group of the long section's of expense sense powder

Volume shape fan instrument group of bundle

(5) proper noun capitalization:

The knowledge description that has comprised proper noun in the words knowledge base, comprising surname, place name, organization names etc., if at first letter capitalization of importing words pinyin, system then can identify correct result effectively ".

5. spell check device

The phonetic converter technique shows stronger intelligent characteristic in design.System can follow the tracks of each letter that the user keys in real time, checks whether the input keystroke is legal.In time provide alarm at user keystroke when being illegal, avoid invalid keystroke, to improve input efficiency.

The present invention sets up database with all syllables of standard Chinese, forms the voice chain, and each the syllable string by the keyboard input all enters this voice chain and carries out real-time coupling.If it fails to match, provide alarm and rejection word string subsequently by the spell check device.

In input process, illegal keystroke has two kinds of situations: the one, and user's words pinyin is wrong; The 2nd, wrong keystroke.If do not carry out real-time spell check, in a string pinyin syllable of user's input as mistake occurs, then system searching is less than corresponding words or provide non-correct words, and just make that the code of input is invalid this moment.If carry out the spell check in real time, system informs the user and rejects error code when wrong spelling or mistake keystroke occurring, has just improved the accuracy of input, thereby has improved input efficiency.

For example:

Import word " correction ", if " repairing xiu " has been combined into " repairing xiou ", then when keying in, the user is when being input to " o ", and system then provides the code string of sound prompting customer inspection input.

6. many transducers

General keyboard input technology promptly no longer keeps the code string of just having imported behind the code of having imported word or speech.The phonetic converter technique is then kept the code string of firm end input, and the user can reproduce the code string of input just by functional key, for conversion use once more.

Use " repeatedly transducer " that two kinds of situations are generally arranged:

The one, " repetition " be the word or the speech of input just.If think input " glad ", can import " happiness " as: user earlier, use " conversion once more " to reappear just now code string then, promptly remake linear transformation, occur second " happiness ", thereby finish the input of this reduplication by space bar.

The 2nd, " correction " be the word or the speech of input just.If the code string of input is wrong, the spell check also fails to find under the situation of this mistake, reproduces this code string with regard to available " conversion once more ", carries out conversion after revising again, has so just improved the efficient of input.

7. memory and making code for composing word device automatically

Because the word in the language is constantly development, add the specific demand in each specialism, field, so the huge again demand that can not satisfy reality fully in word storehouse, so system offers the ability that the user generates neologisms.

(1) automatic memory

In the process of user operation, the word that runs into user's input when system is not that thumb is waitd upon Jia and protected the low occasion  far away of  and wash the San libation at an ancient wedding ceremony and look an ancient spear in the system word storehouse

Automatically remember the max cap. of entry: 5000

The maximum number of words of each entry: 8 words

Give an example 1: memory " gray system "

Key in: huisexitong

After pressing space bar and finishing, system provide " grey " earlier and homonym selective; Provide " system " then, at this moment, the printed words of " memory automatically " appear in the screen lower right corner.

Give an example 2: memory name " Rigen "

Key in: ligen

After pressing space bar and finishing, system provide " lining " earlier and phonetically similar word selective; The phonetically similar word that provides " root " then is selective.After selection finished, this speech was promptly remembered by system.

Automatically the word in the word of memory and system word storehouse is according to using with quadrat method.

The inconsistent situation of transformation results that the user wants that the speech imported and system provide may appear in automatic memory process.

As: the user wants to import certain name and " often levies ", and after the key entry " changzheng ", system changeover goes out " Long March ", and this does not just meet user's requirement.

At this moment, can use " conversion once more " to recover the code string " changzhen " of input just, or re-enter this code string, press enter key and finish input process, system will provide the conversion of each syllable one by one, select back memory automatically for the user.

Automatically the speech of memory has following feature:

(1) speech of just having been remembered does not deposit in user's the phrase library for temporary memory immediately, but in temporary transient dictionary, just keeps for a long time with (more than the three times) systems that increases of access times, and promptly deposits user thesaurus in.

(2) word of new memory has preferential usage degree, and as homonym, then the speech of new memory makes number one, and is easy to use.

Automatically the memory process is that system manages, be to carry out in extremely natural mode, it is always effective that the sensation of giving the user is exactly the code of importing, code that can " waste " do not imported because system does not have the desired speech of user fully shows " close friend " of this system to the user in this point.

(2) making code for composing word device

Making code for composing word provides not only " group neologisms " but also can " make new sign indicating number " to the user.Different with " memory automatically ", making code for composing word is that neologisms are remembered by " pressure " system, and gives speech new code.

The entry capacity of making code for composing word: 1000

The maximum number of words of each entry: 16 words

The longest code of each entry: 8 characters

The making code for composing word process is given an example:

" China Science and Technology Association "

(1) presses ALT+6, screen display [neologisms] printed words

(2) according to the input method participle of words or not participle input;

(3) press space bar;

(4) press enter key;

(5) input " kx " is as code;

(6) press enter key after, making code for composing word finishes;

Input " n " earlier when (7) using, input code " kx " can demonstrate " China Science and Technology Association " then.

Automatically memory and making code for composing word all are the abilities that produces neologisms in order to provide to the user, and two functions complement each other, and use all very convenient.All remember in user's phrase library for temporary memory by the neologisms that these two kinds of approach produce, different user both can have been shared also and can independently separately use.

8. extended operation rule

Using the extended operation rule is in order to improve input efficiency and system changeover accuracy.Use the user of this system can grasp these rules for the first time.

1. phonetic and tone indication

Phonetic and tone indication is the important component part of the Chinese phonetic alphabet, adopts mark to transfer the system that is more convenient for to break up phonetically similar word and homonym.This system can be added to tone mark on the corresponding vowel.What tone mark adopted is following numerical key:

Numerical key tone circumflex

1 high and level tone

2 rising tones

Sound on 3

4 falling tones

0 softly, noiseless

Phonetic and tone indication can use also and can not use, because of the people suitable.But adopt mark to transfer the accuracy that can improve system changeover.

2. one character word of simple spelling

In order to improve input efficiency, system has stipulated that the highest monosyllabic word of a small amount of usage degree can be with the input of simplicity form.One character word of simple spelling has:

Q goes his y one p mistake of my r people t of w

S is that g h in f ground and the j of d seen l with regard to k

Z is not at the big b of the little c of x

Zh ch and goes out on the sh

One character word of simple spelling can together be imported with the speech of front when input.

For example:

The word enter key

Huijiaq goes home

Beautiful meilid

3. two character word of simple spelling

In the system the highest a part of disyllabic word of usage degree is provided the input format of simplicity, promptly imported first initial consonant of each word of two-character word.

For example:

The two-character word enter key

We are wm

They are tm

Know zd

Two character word of simple spelling

The own zj of we they tm of the wm ql that gets up

There is not my work gz people comrade rm tz

Can ky know the zd cl yj that comes out

But the present xz revolution of this zg gm ks

The certain yd of these zx of problem wt is because yw

How but your nm of yh after the zm ds

The many xd of thing dx produce sc people rm

See kj world sj dj development fz

So thought sx is for wl workman gr sy

If local some yx jt today rg of df

The xl zl gj of country science kx here gets off

Economic jj and eq carry out jx struggle dz

Masses qz should yg study xx or hs

Situation qk thinks that all yq of jd need xy

Way bf begins the ks aspect fm xq that goes down

Concern that the special tb of the historical ls of gx may kn

Thing sq is the bg complete wq of nm so only

Though the young qn of sr must bx politics zz

The general yb understanding rs of ny is because yy like that

Different bt or hz be fc bd not only very

Rh and bq think rw so yc then

So ys

Claims (5)

1, pinyin sequence is characterized in that to the conversion method (abbreviation conversion method) of Chinese character string (phrase),

A. Chinese-character pronunciation database, the data structure of Chinese-character pronunciation is

Chinese character initial consonant simple or compound vowel of a Chinese syllable four tones of standard Chinese pronunciation usage frequency

B. Chinese-character words database, the data structure of Chinese-character words is

Entry-word sound part of speech semantic item usage frequency

C. be based upon pinyin sequence on above-mentioned two Basis of Database to the converter of Chinese character string, converter comprises automatic word segmentation and homonym identification, and the decision rule of homonym identification comprises word frequency, the collocation of speech speech, closes on environment, prefix and suffix, proper noun (name and place name);

2, conversion method as claimed in claim 1 is characterized in that, the information of user's keyboard input needs through a spell check device;

As claim 1 and 2 described conversion methods, it is characterized in that 3, comprise a repeatedly transducer, when the device of spell check, repeatedly converter design is in its back;

4, as claim 1 and 2 described conversion methods, it is characterized in that, comprise an automatic memory and a making code for composing word device, memory is subjected to the support of a temporary transient dictionary automatically, and this dictionary and Chinese-character words database adopt with a kind of data structure;

5, conversion method as claimed in claim 3 is characterized in that, comprises an automatic memory and a making code for composing word device, and memory is subjected to the support of a temporary transient dictionary automatically, and this dictionary and Chinese-character words database adopt with a kind of data structure;