CN105404621B - A kind of method and system that Chinese character is read for blind person - Google Patents
A kind of method and system that Chinese character is read for blind person Download PDFInfo
- Publication number
- CN105404621B CN105404621B CN201510623525.5A CN201510623525A CN105404621B CN 105404621 B CN105404621 B CN 105404621B CN 201510623525 A CN201510623525 A CN 201510623525A CN 105404621 B CN105404621 B CN 105404621B
- Authority
- CN
- China
- Prior art keywords
- braille
- participle
- word
- chinese character
- polyphone
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Document Processing Apparatus (AREA)
Abstract
The present invention proposes a kind of method and system that Chinese character is read for blind person, it is related to natural language processing technique field and the human-computer interaction technique field towards disabled person, this method includes obtaining Chinese language text, participle operation is carried out to the Chinese language text, generate Chinese character string, by pronunciation dictionary, polyphone dictionary and word frequency information, with reference to the part-of-speech tagging that participle obtains, each word in the Chinese character string is converted into corresponding phonetic and is connected as pinyin string;By searching for phonetic and the control dictionary of blind symbol, the pinyin string is converted into blind symbol string, braille participle is carried out to the blind symbol string by participle model, generate initial braille participle, the Chinese character string is merged with the initial braille participle, new braille participle is generated, the new braille participle is adjusted according to braille word link writing rule;To carrying out braille mark tune according to the new braille participle after braille word link writing rule adjustment, final braille participle is generated, the final braille participle is shown.
Description
Technical field
Human-computer interaction technique field the present invention relates to natural language processing technique field and towards disabled person, particularly
It is related to a kind of method and system that Chinese character is read for blind person.
Background technology
In current information-intensive society, the level of IT application is continuously improved, and information technology is in the work of people, studying and living
It is widely applied, and internet also becomes an important component in people's daily life, network is with a kind of convenient
Mode the information resources of magnanimity are provided for people.In China, various digitlizations, network text resource are mostly with Chinese language
This form storage, and these resources are difficult to be used by existing 12,000,000 blind person in China.Which prevent blind person as normal person
The information resources of magnanimity are equally enjoyed, the telecoms gap between blind person and normal person are made constantly to expand, blind person is in informationized society
In survival and development ability further restricted.It is a large amount of on network although existing speech synthesis technique reaches its maturity
Textual resources can be converted to audio file blind person by phonetic synthesis can obtain these information, but language by the sense of hearing
The storage of sound resource, which is compared, expends space, and in carrying, inquiry etc. and inconvenience, moreover, voice channel obtains information
Less efficient, therefore, for blind person, it is still to obtain the most important mode of information to read textual resources.
The word that China blind person uses in reading writing is Chinese braille, and Chinese braille is based on Blair (Braille)
Braille system, each blind symbol arrange 6 totally o'clock as one basic structure using two, and the protrusion that this 6 points have, some is not raised,
64 kinds of variations are formed, can represent 64 kinds of different characters.In Chinese braille, each character is represented in the Chinese phonetic alphabet respectively
An initial consonant, simple or compound vowel of a Chinese syllable or tone, different characters according to Chinese phonetic alphabet rule form legal syllables to represent Chinese character, therefore,
Chinese braille is substantially a kind of alphabetic writing.Braille is generally printed and is written on special thicker braille paper, in braille
The point position of protrusion is extruded on paper for blind person's touching reading.To enable blind person's touching reading braille on computers, currently it has been equipped with
An aobvious device is put with blind use has been produced, this equipment can be connect with computer, receive the blind symbol string in computer, and by it in point
The point position of corresponding protrusion is shown as on aobvious machine side plate, after new blind symbol string is received, original point position can be removed on panel
Again new point position is shown.
Device is a little shown although having, blind person is still difficult to read Chinese language text on computers, and reason is to also need to
Chinese language text is converted into braille.Phenomena such as due to a sound multiword of Chinese generally existing, a word multitone so that Chinese is to blind
The not simple rule of the conversion of text is corresponding, and needs to consider grammer, semanteme etc..What is more important, braille, which has, to be divided
Word combination handwriting rule, it is desirable that the word or phrase that will be provided with certain semanteme are separated with one " short side ", in order to which blind person understands.
Current existing method is generally basede on braille word link writing rule and Chinese word segmenting result is adjusted with blind after being segmented
Text, but since braille word link writing rule is generally related to semanteme and has certain subjectivity, it is automatically complete by computer
Into when participle accuracy rate it is relatively low, after these methods is used to be converted, it is also necessary to do the work of a large amount of artificial corrections, cause
Inefficiency, but also the time of the acquisition of braille text resource is longer and cost is higher.Therefore, the blind conversion of the Chinese is improved
Accuracy rate reduces the operation of artificial correction, accelerates the efficiency of the blind conversion of the Chinese, for improving Chinese information resource in blind community
In popularity rate, blind community is allowed, which to better blend into mainstream society, important realistic meaning.
Invention content
In view of the deficiencies of the prior art, the present invention proposes a kind of method and system that Chinese character is read for blind person.
The present invention proposes a kind of method that Chinese character is read for blind person, including:
Step 1, Chinese language text is obtained, participle operation is carried out to the Chinese language text, Chinese character string is generated, passes through the word that pronounces
With reference to the part-of-speech tagging that participle obtains, each word in the Chinese character string is converted to for allusion quotation, polyphone dictionary and word frequency information
Corresponding phonetic is simultaneously connected as pinyin string;
Step 2, by searching for phonetic and the control dictionary of blind symbol, the pinyin string is converted into blind symbol and is gone here and there, passes through participle
Model carries out braille participle to the blind symbol string, generates initial braille participle, the Chinese character string and the initial braille are segmented
It is merged, generates new braille participle, the new braille participle is adjusted according to braille word link writing rule;
Step 3, to carrying out braille mark tune, generation according to the new braille participle after braille word link writing rule adjustment
Final braille participle shows the final braille participle.
The Chinese character string is converted into the tool of pinyin string in the step 1 by the described method that Chinese character is read for blind person
Body step is:
Step 2.1 judges whether each word is multi-character words for each word in the Chinese character string, if multi-character words, and
The corresponding phonetic of the multi-character words can be found in pronunciation dictionary, then directly returns to the corresponding phonetic of the multi-character words, otherwise
Perform step 2.2;
Step 2.2 by the multiword word segmentation be Chinese character sequence, take Chinese character all in the multi-character words successively, it is right
Each Chinese character performs step 2.3 to 2.4;
Step 2.3 judges whether the current Chinese character is polyphone for current Chinese character, lookup polyphone dictionary, if not
Polyphone searches the phonetic of the current Chinese character in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;
Step 2.4 then performs following steps if polyphone, the specific steps are:
If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;If multiword
Word then performs following step:
For the polyphone w in multi-character wordsk, a) step, the word W with follow-up n word one n+1 words of compositionk,n=wkwk+1…
wk+n, W is searched in polyphone phrase dictionaryk,n, such as find, then with Wk,nIn be searched the pronunciation of word as polyphone wk
Pronunciation and return;If do not found, then b) step is performed, the word W of a n+1 words is formed with the word of front nn-k,k=wn- kwn-kk+1…wn, W is searched in polyphone phrase dictionaryn-k,k, such as find, then with Wk,nIn be searched the pronunciation conduct of word
The pronunciation of polyphone and return are not searched such as, then form the word W of a n words with the follow-up and word of front n-1 respectivelyk,n-1、
Wn-k+1,k, the multi-character words are performed respectively a), b) step, until determining the polyphone wkPronunciation;
Step 2.4.2 assumes that the polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are
Ppos, weights λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, it is
It unites and calculates a score Score for each pronunciation of the polyphonei, wherein Scorei=λ1·Ppos(tonei)+λ2·
Plm(tonei)+λ3·Pfreq(tonei), take out the pronunciation of highest scoringFinal phonetic as polyphone is simultaneously
It returns.
The described method that Chinese character is read for blind person, the step of being merged in the step 2, are, for the Chinese
Word string C=c1c2…cmWith the initial braille participle B=b1b2…bn, wherein ci,bjThe Chinese character string and described is represented respectively
A participle in initial braille participle segments B for the initial braille, B is mapped to the corresponding Chinese character string B'=
b1'b'2…b'n, wherein b'jB is segmented for the initial braillejIt is mapped as the participle after Chinese.
The described method that Chinese character is read for blind person, braille word link writing rule is as follows in the step 2:
Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+n
POSkFor activation condition, m and n expressions need to check the preceding m word and n word of current new braille participle respectively, such as
Fruit m and n are 0, then it represents that this is a word segmentation regulation, and what is represented after second colon is the part of speech combination of participle, if full
The foot combination, then perform the operation after right arrow.
The described method that Chinese character is read for blind person, braille mark tune described in the step 3 the specific steps are:
The phonetic of the corresponding word of new braille participle after each adjustment is checked successively, and the rule in being assembled with braille mark
It is then compared, if meeting condition, current new braille is segmented into rower tune, the form that the braille mark is assembled is as follows:
Mark adjusts rule:tonek:[n]:tonek…tonek+n
Wherein tonekFor the phonetic that current new braille segments, n is to need to check that the rear n of current new braille participle is a new blind
The phonetic of text participle, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower
It adjusts.
The present invention also proposes a kind of system that Chinese character is read for blind person, including:
Pinyin string module is obtained, for obtaining Chinese language text, participle operation is carried out to the Chinese language text, generates Chinese character
String,, will be in the Chinese character string with reference to the part-of-speech tagging that participle obtains by pronunciation dictionary, polyphone dictionary and word frequency information
Each word is converted to corresponding phonetic and is connected as pinyin string;
It obtains new braille to segment and adjust module, for the control dictionary by searching for phonetic and blind symbol, by the phonetic
String is converted to blind symbol and goes here and there, and carries out braille participle to the blind symbol string by participle model, initial braille participle is generated, by described in
Chinese character string is merged with the initial braille participle, new braille participle is generated, according to braille word link writing rule to described new
Braille participle is adjusted;
Braille display module, for blind to being carried out according to the new braille participle after braille word link writing rule adjustment
Text mark is adjusted, and generates final braille participle, and the final braille participle is shown.
The Chinese character string is converted by the system that Chinese character is read for blind person, described obtain in pinyin string module
Pinyin string the specific steps are:
Step 2.1 judges whether each word is multi-character words for each word in the Chinese character string, if multi-character words, and
The corresponding phonetic of the multi-character words can be found in pronunciation dictionary, then directly returns to the corresponding phonetic of the multi-character words, otherwise
Perform step 2.2;
Step 2.2 by the multiword word segmentation be Chinese character sequence, take Chinese character all in the multi-character words successively, it is right
Each Chinese character performs step 2.3 to 2.4;
Step 2.3 judges whether the current Chinese character is polyphone for current Chinese character, lookup polyphone dictionary, if not
Polyphone searches the phonetic of the current Chinese character in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;
Step 2.4 then performs following steps if polyphone, the specific steps are:
If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;If multiword
Word then performs following step:
For the polyphone w in multi-character wordsk, a) step, the word W with follow-up n word one n+1 words of compositionk,n=wkwk+1…
wk+n, W is searched in polyphone phrase dictionaryk,n, such as find, then with Wk,nIn be searched the pronunciation of word as polyphone wk
Pronunciation and return;If do not found, then b) step is performed, the word W of a n+1 words is formed with the word of front nn-k,k=wn- kwn-kk+1…wn, W is searched in polyphone phrase dictionaryn-k,k, such as find, then with Wk,nIn be searched the pronunciation conduct of word
The pronunciation of polyphone and return are not searched such as, then form the word W of a n words with the follow-up and word of front n-1 respectivelyk,n-1、
Wn-k+1,k, the multi-character words are performed respectively a), b) step, until determining the polyphone wkPronunciation;
Step 2.4.2 assumes that the polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are
Ppos, weights λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, it is
It unites and calculates a score Score for each pronunciation of the polyphonei, wherein Scorei=λ1·Ppos(tonei)+λ2·
Plm(tonei)+λ3·Pfreq(tonei), take out the pronunciation of highest scoringFinal phonetic as polyphone is simultaneously
It returns.
The system that Chinese character is read for blind person, the new braille of acquisition, which is segmented and adjusted in module, to be merged
The step of be, for the Chinese character string C=c1c2…cmWith the initial braille participle B=b1b2…bn, wherein ci,bjTable respectively
Show a participle in the Chinese character string and initial braille participle, B is segmented for the initial braille, B is mapped to pair
The Chinese character string B'=b answered1'b'2…b'n, wherein b'jB is segmented for the initial braillejIt is mapped as the participle after Chinese.
The system that Chinese character is read for blind person, the new braille of acquisition segment and adjust braille in module and segment
Combination handwriting rule is as follows:
Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+n
POSkFor activation condition, m and n expressions need to check the preceding m word and n word of current new braille participle respectively, such as
Fruit m and n are 0, then it represents that this is a word segmentation regulation, and what is represented after second colon is the part of speech combination of participle, if full
The foot combination, then perform the operation after right arrow.
The system that Chinese character is read for blind person, the specific step of braille mark tune described in the braille display module
Suddenly it is:
The phonetic of the corresponding word of new braille participle after each adjustment is checked successively, and the rule in being assembled with braille mark
It is then compared, if meeting condition, current new braille is segmented into rower tune, the form that the braille mark is assembled is as follows:
Mark adjusts rule:tonek:[n]:tonek…tonek+n
Wherein tonekFor the phonetic that current new braille segments, n is to need to check that the rear n of current new braille participle is a new blind
The phonetic of text participle, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower
It adjusts.
By above scheme it is found that the advantage of the invention is that:
The present invention is different from the blind switch technology of the existing Chinese, first carries out Chinese word segmenting to Chinese character string, then in word segmentation result
On with a series of complex word link writing rule carry out after-treatment way, the present invention using structure based on statistical machine
The braille participle model of device learning art directly carries out single step participle to blind symbol string, and word segmentation result substantially conforms to braille participle
Combination handwriting rule need to only carry out finely tuning can be used as braille output on a small quantity, compared with prior art, avoid and answered with computer disposal
It is miscellaneous, be involved in the problems, such as semantic word link writing rule caused by accuracy rate it is not high, participle accuracy rate and the blind conversion of the whole Chinese are accurate
True rate has larger promotion.
Description of the drawings
Fig. 1 is the method flow diagram that Chinese character is read for blind person;
Fig. 2 is that the Chinese character string after participle is converted to the flow chart of pinyin string.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, to this
The method for reading Chinese character for blind person of invention is further elaborated, it should be understood that specific implementation described herein
Example is not intended to limit the present invention only to explain the present invention.
The method main flow for reading Chinese character for blind person of the present invention as shown in Figure 1, is inputted as a Chinese sentence
Son, i.e. a Chinese character string are exported as corresponding braille, and are shown in blind on the aobvious device of point.
Step 1. Chinese word segmenting.Chinese word segmentation system is used for the sequence of Chinese word, to obtain the Chinese character string cutting of input
Chinese character string after to participle, while part of speech is marked for each word, Chinese word segmenting can be used current existing various methods and be
System, such as maximum based on dictionary or smallest match method, based on the method for Hidden Markov Model (HMM), based on maximum entropy mould
Method of type etc.;
Chinese character string after participle is converted to pinyin string by step 2., i.e., is believed using pronunciation dictionary, polyphone dictionary and word frequency
Each word in Chinese character string after participle with reference to the part-of-speech tagging that participle obtains, is converted to corresponding phonetic and is connected as by breath
Pinyin string, mapping table of the pronunciation dictionary for Chinese words (including monosyllabic word and multi-character words) and phonetic.In one embodiment
In, the scale of pronunciation dictionary is 70,000 words or so, and all polyphones and each of which multitone are listed in the polyphone dictionary
The corresponding multiple phonetics of word, the word frequency information are the frequency of occurrences in Chinese language text of each Chinese character, which adopts in advance
It counts to obtain with a large amount of Chinese language texts.In one embodiment, the scale of word is 7000 words or so in word frequency information.
Specific steps for participle below, as shown in Figure 2:
Step 2.1 for each word in the Chinese character string after participle, judge the word whether be multi-character words (comprising two or
Above Chinese character), if multi-character words, and the corresponding phonetic of the word can be found in pronunciation dictionary, then directly return to the spelling
Otherwise sound performs step 2.2;
Step 2.2, by the sequence that word segmentation is Chinese character, takes its institute successively for the word (monosyllabic word or multi-character words) of input
Some Chinese characters to each Chinese character, perform step 2.3 to 2.4;
Step 2.3 judges whether the word is polyphone for current Chinese character, lookup polyphone dictionary, if not polyphone,
The phonetic of the word is searched in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;
Step 2.4 need to integrate the phonetic that much information determines polyphone for polyphone.The specific steps are:
If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;Otherwise it first holds
Row following step:
For the polyphone w in multi-character wordsk, a) with follow-up n word composition one n+1 words word Wk,n=wkwk+1…wk+n,
W is searched in polyphone phrase dictionaryk,n, such as find, then using in the phrase word pronunciation as polyphone pronunciation and return
It returns;If do not found, then the word W of a n+1 words b) is formed with the word of front nn-k,k=wn-kwn-kk+1…wn, in polyphone phrase word
W is searched in allusion quotationn-k,k, such as finding, then the pronunciation and return using the pronunciation of the word in the phrase as polyphone, are not searched such as,
The word W of a n words is then formed with the follow-up and word of front n-1 respectivelyk,n-1、Wn-k+1,k, which is performed respectively a), b) step, directly
Pronounce to the determining polyphone.If during n=1, Wk,k+1、Wk-1,kStill can not in polyphone phrase dictionary lookup to pronunciation,
Then return to sky;
Step 2.4.2 assumes that polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are Ppos,
Weights are λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, system is
Each pronunciation of polyphone calculates a score Scorei, wherein Scorei=λ1·Ppos(tonei)+λ2·Plm(tonei)+
λ3·Pfreq(tonei), take out the pronunciation of highest scoringAs polyphone final phonetic and return.It needs
Illustrate, for part of speech, word frequency, all types of each pronunciations of language model probability, need to be normalized, it is all kinds of
The weights of type can be set based on experience value.
Pinyin string is converted to blind symbol and gone here and there by step 3..By searching for phonetic and the control dictionary of blind symbol, step 2 is obtained
Pinyin string be converted to it is blind symbol string, blind symbol string at this time is the blind symbol string not segmented.The control dictionary of the phonetic and blind symbol is
The mapping table of phonetic and corresponding blind symbol.
Step 4. with the trained participle model of statistical machine learning method using braille participle is carried out in advance, and generation is just
Beginning braille segments.Using the common perceptron model in currently associated field, using the braille language for having divided word during model training
Material, the feature used is unitary feature, binary feature and attributive character.Each gone here and there during participle to blind symbol can be with cutting
Whether position extracts feature and calculates probability using trained model, needed to carry out word in the position according to probabilistic determination
Cutting.
Training pattern uses perceptron algorithm, learns to from the discriminate mapping model for being input to output, input is trained
Sentence in language material, output are corresponding annotation results.
Word disaggregated model is used to the participle of braille sentence.Give a sentence being made of n word, the process of participle
It is that this sentence is divided into m (m≤n) block, each piece is a significant word.One, which is distributed, to each word represents it in word
Participle problem is converted to word classification problem by the category of middle position.Using b, m, e, boundary categories of the s as word, b, m, e points
Starting position, centre position, end position that the word is located at word are not represented, and the behalf word is monosyllabic word.Decoding process is to seek
Looking for makes the highest annotated sequence y of goals for evaluation function f (x).
Wherein, f (x) scores are accumulated each word and the score of category pair, (i, t) ∈ y (s.t.1≤i≤n, t ∈ b,
M, e, s }), Φ (x, y) is feature extraction function,It is parameter vector.Participle uses Viterbi decoding algorithm.
Step 5. Chinese and initial braille participle are merged, i.e., braille are segmented using Chinese braille word segmentation result and tied
Fruit is finely adjusted, to further improve the accuracy rate of participle.
For Chinese word segmentation C=c1c2…cmB=b is segmented with braille1b2…bn, wherein ci,bjIt represents respectively Chinese and blind
A participle in text, segments B for braille, B can be mapped to corresponding Chinese word segmentation B'=b1'b'2…b'n, wherein
b'jB is segmented for braillejIt is mapped as the participle after Chinese.B' is segmented into edlin to Chinese word segmentation C and the braille for being mapped as Chinese
Range alignment can obtain segment different in C and B', with above-mentioned fusion rule, determine the final result of different fragments
It is using Chinese word segmentation result or braille word segmentation result.Assuming that segment different in C and B' is respectively defined as CH=
ch1ch2…chmAnd BR=br1br2…brn, it is as follows:
Step 5.1 assumes chiFor i-th of participle, br in CHjFor j-th of participle in BR, initial value i, j are both configured to 1
Step 5.2 is respectively compared chiAnd brjIfIllustrate in first participle, during braille participle includes
Text participle, then for first participle, the result br segmented using braillej;Opposite, ifThen using Chinese
The result ch of participlei
Step 5.3 initial setting up k=1
5.3.1 forSituation, define chi,i+k=chi…chi+k, compare chi,i+kAnd brj:
If a) chi,i+1=brj, i=i+2, j=j+1 are set, if i>M or j>N jumps to step 5.4, otherwise,
Jump to step 5.2
If b)K=k+1 jumps to 5.3.1
If c)Illustrate chi+kIn include brjIn the last character, define the position of the word as pos,
Then using pos as boundary, by chi+kIt is divided into chi+k,posAnd chi+k,after_pos, wherein chi+k=chi+k,poschi+k,after_pos,
chi+k,posRepresent chi+kIn the 1st phrase formed to os word of pth, chi+k,after_posRepresent chi+kMiddle pth os+1 words arrive
The phrase of the last character composition.By the i-th+k in Chinese word segmentation participle chi+k,after_posIt replaces, that is, updates CH=ch1…
chi+k-1chi+k,after_poschi+k+1…chm, i=i+k, j=j+1 jump to step 5.2
5.3.2 forSituation, define brj,j+k=brj…brj+k, compare brj,j+kAnd chi:
If a) brj,j+1=chi, then i=i+1, j=j+2, jump to step 5.2
If b)K=k+1 jumps to 5.3.2
If c)Illustrate brj+kIn include chiIn the last character, define the position of the word as pos,
Then using pos as boundary, by brj+kIt is divided into brj+k,posAnd brj+k,after_pos, wherein brj+k=brj+k,posbrj+k,after_pos,
brj+k,posRepresent brj+kIn the 1st phrase formed to os word of pth, brj+k,after_posRepresent brj+kMiddle pth os+1 words arrive
The phrase of the last character composition.Jth+k participle br during braille is segmentedj+k,after_posIt replaces, that is, updates BR=br1…
brj+k-1brj+k,after_posbrj+k+1…brn, i=i+1, j=j+k jump to step 5.2
Step 5.4 terminates integration algorithm
Step 6. is according to braille word link writing rule adjustment word segmentation result.Check the corresponding part of speech of participle successively, and with it is blind
Activation condition in literary word link writing rule set is compared, if met, the Conditions On The Results that applying rules are concentrated carry out
Participle or write the two or more syllables of a word together.Braille word link writing rule set form is as follows:
Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+n
For the rule in rule set, the part of speech POS before first colonkIt is activation condition, it can be with including in one after rule
Number, m and the n expression of the inside need to check the preceding m word currently segmented and n word respectively, if m and n are 0, then it represents that this
It is a word segmentation regulation.What is represented after second colon is the part of speech combination of participle, if meeting the combination, performs right arrow
Operation after head.
Step 7. braille mark tune.Check the phonetic of each corresponding word of participle successively, and the rule in being assembled with braille mark into
Row compares, if meeting condition, to current word into rower tune.The form that braille mark is assembled is as follows:
Mark adjusts rule:tonek:[n]:tonek…tonek+n
Wherein tonekFor the phonetic of current word, the n expressions in square brackets need to check the spelling of the rear n word of current word
Sound, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower tune
Step 8. braille is shown, i.e., is output to braille blind on the aobvious device of point.Current existing various points can be used and show device
Product, and call its corresponding output interface.
The present invention also proposes a kind of system that Chinese character is read for blind person, including:
Pinyin string module is obtained, for obtaining Chinese language text, participle operation is carried out to the Chinese language text, generates Chinese character
String,, will be in the Chinese character string with reference to the part-of-speech tagging that participle obtains by pronunciation dictionary, polyphone dictionary and word frequency information
Each word is converted to corresponding phonetic and is connected as pinyin string;
It obtains new braille to segment and adjust module, for the control dictionary by searching for phonetic and blind symbol, by the phonetic
String is converted to blind symbol and goes here and there, and carries out braille participle to the blind symbol string by participle model, initial braille participle is generated, by described in
Chinese character string is merged with the initial braille participle, new braille participle is generated, according to braille word link writing rule to described new
Braille participle is adjusted;
Braille display module, for blind to being carried out according to the new braille participle after braille word link writing rule adjustment
Text mark is adjusted, and generates final braille participle, and the final braille participle is shown.
It is described obtain pinyin string module in by the Chinese character string be converted into pinyin string the specific steps are:
Step 2.1 judges whether each word is multi-character words for each word in the Chinese character string, if multi-character words, and
The corresponding phonetic of the multi-character words can be found in pronunciation dictionary, then directly returns to the corresponding phonetic of the multi-character words, otherwise
Perform step 2.2;
Step 2.2 by the multiword word segmentation be Chinese character sequence, take Chinese character all in the multi-character words successively, it is right
Each Chinese character performs step 2.3 to 2.4;
Step 2.3 judges whether the current Chinese character is polyphone for current Chinese character, lookup polyphone dictionary, if not
Polyphone searches the phonetic of the current Chinese character in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;
Step 2.4 then performs following steps if polyphone, the specific steps are:
If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;If multiword
Word then performs following step:
For the polyphone w in multi-character wordsk, a) step, the word W with follow-up n word one n+1 words of compositionk,n=wkwk+1…
wk+n, W is searched in polyphone phrase dictionaryk,n, such as find, then with Wk,nIn be searched the pronunciation of word as polyphone wk
Pronunciation and return;If do not found, then b) step is performed, the word W of a n+1 words is formed with the word of front nn-k,k=wn- kwn-kk+1…wn, W is searched in polyphone phrase dictionaryn-k,k, such as find, then with Wk,nIn be searched the pronunciation conduct of word
The pronunciation of polyphone and return are not searched such as, then form the word W of a n words with the follow-up and word of front n-1 respectivelyk,n-1、
Wn-k+1,k, the multi-character words are performed respectively a), b) step, until determining the polyphone wkPronunciation;
Step 2.4.2 assumes that the polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are
Ppos, weights λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, it is
It unites and calculates a score Score for each pronunciation of the polyphonei, wherein Scorei=λ1·Ppos(tonei)+λ2·
Plm(tonei)+λ3·Pfreq(tonei), take out the pronunciation of highest scoringFinal phonetic as polyphone is simultaneously
It returns.
The new braille of acquisition, which segments and adjusts the step of being merged in module, is, for the Chinese character string C=
c1c2…cmWith the initial braille participle B=b1b2…bn, wherein ci,bjThe Chinese character string and the initial braille are represented respectively
A participle in participle segments B for the initial braille, B is mapped to the corresponding Chinese character string B'=b1'b'2…
b'n, wherein b'jB is segmented for the initial braillejIt is mapped as the participle after Chinese.
It is described to obtain that new braille segment and to adjust braille word link writing in module regular as follows:
Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+n
POSkFor activation condition, m and n expressions need to check the preceding m word and n word of current new braille participle respectively, such as
Fruit m and n are 0, then it represents that this is a word segmentation regulation, and what is represented after second colon is the part of speech combination of participle, if full
The foot combination, then perform the operation after right arrow.
Braille mark tune described in the braille display module the specific steps are:
The phonetic of the corresponding word of new braille participle after each adjustment is checked successively, and the rule in being assembled with braille mark
It is then compared, if meeting condition, current new braille is segmented into rower tune, the form that the braille mark is assembled is as follows:
Mark adjusts rule:tonek:[n]:tonek…tonek+n
Wherein tonekFor the phonetic that current new braille segments, n is to need to check that the rear n of current new braille participle is a new blind
The phonetic of text participle, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower
It adjusts.
Below by Chinese is carried out to a Chinese sentence to the conversion of braille and display as example, this is discussed in detail
Invention for blind person read Chinese character method and system implementation process, it is clear that the example be only intended to for example,
Without being intended to limit the scope of the invention.
Assuming that the Chinese sentence that need to be converted to braille is:" Beijing is their destination ", using Chinese word segmenting module into
Row Chinese word segmenting simultaneously carries out part-of-speech tagging, and obtained result is:" Beijing/NR is /VC they/PN/DEG purposes/NN/
NN”。
Call Chinese character string that word segmentation result is converted to pinyin string to pinyin string modular converter, for " Beijing ", "Yes", " she
", " purpose " this five words, can directly confirm pronunciation by searching for Pronounceable dictionary;For " " and " " the two words, by
In being all polyphone, algorithm need to be called to determine that polyphone pronounces.
By " " for word, by part-of-speech tagging " " part of speech of word is " DEG ", which can be confirmed by " DEG "
Pronunciation for " de ", due to can uniquely be confirmed by part of speech " " word pronunciation, so:
Ppos(de)=1,
Ppos(di)=0
Under conditions of previous word is " they ", by searching for probabilistic language model, pronunciation can be obtained as " de "
Probability is 0.45, and it is 0.05 to pronounce for the probability of " di ":
Plm(de)=P (de | tamen)=0.45
Plm(di)=P (di | tamen)=0.05
After being normalized, it can obtain:Plm(de)=0.9, Plm(di)=0.1
In word frequency dictionary search " " individual character word frequency, it is 185 times to pronounce for the number of " de ", is pronounced for " di "
Number is 75 times, and by calculating it is found that the probability that pronunciation is " de " is 0.71, it is 0.29 to pronounce for the probability of " di "
Based on experience value, set part of speech, language model, word frequency three's probability weight all for 1/3, then:
Compared by score, it may be determined that polyphone " " final pronunciation be " de ".
It is similar, it may be determined that " " pronunciation of word is " di ".The corresponding pinyin string of Chinese sentence is finally obtained as " bei
jing shi ta men de mu di di”。
Calling pinyin string, it is " B to obtain the corresponding blind symbol string of pinyin string to blind symbol string modular converter!G*:T9 M0 D MU
DI DI”.(ASCII character that the braille used in this specification is expressed as blind symbol encodes, and the point position form of non-blind symbol.Hereafter
In it is identical.)
Braille word-dividing mode is called to segment blind symbol string, the blind symbol string after being segmented is " B!G*:|T9 M0|D|
MU DI DI”。
Chinese and braille word segmentation result Fusion Module is called to merge Chinese word segmentation result and braille word segmentation result.
Will after participle braille string correspondence to Chinese string, can obtain use the Chinese character string of braille participle for " Beijing is/they// purpose
The Chinese character string of braille participle is carried out editing distance with the Chinese character string of Chinese word segmenting and is aligned, can obtain subordinate list 1 by ground ":
Subordinate list 1:Chinese, the braille participle table of comparisons
Chinese and braille participle in subordinate list 1 are compared, there are two different segments, segment 1 " Beijing is " and 2 " purposes of segment
Ground ".
Segment 1 is handled, the Chinese word segmenting of segment 1 is " Beijing/be ", and braille participle is " Beijing is ", takes Chinese
First participle " Beijing " of participle and first participle " Beijing is " of braille participle are compared, due to the in braille participle
One word " Beijing is " contains first word " Beijing " in Chinese word segmenting, continues to check second word "Yes" of Chinese word segmenting,
And be combined to form first word " Beijing is " that " Beijing is " segments with braille and compared with first word " Beijing ",
Because the two is identical and segment 1 in there is no other untreated words, according to choosing the more word of number of words as finally segmenting
Rule, it is thus determined that the participle of segment 1 is " Beijing is ".
Similar, it may be determined that the participle of segment 2 is " destination ".Finally, it may be determined that the word segmentation result after fusion is
" Beijing is/they// destination ".
Word segmentation result adjustment module is called, according to Chinese word segmenting annotation results, Pekinese's part of speech is " NR ", i.e. proprietary name
Word just carries out write the two or more syllables of a word together in braille standard for proper noun followed by single syllable generic noun, " Beijing " followed by "Yes" in example,
Part of speech is " VC ", i.e., " link-verb ", is unsatisfactory for the condition of braille standard, should not carry out write the two or more syllables of a word together, copes with the participle " north of fusion
Capital is " split, obtain " Beijing/be ", after adjusted, obtained word segmentation result for " Beijing/be/they// purpose
Ground ", corresponding braille participle representation is " B!G*:T9M0 D MUDIDI”.
Call braille mark mode transfer block to word segmentation result into rower tune.It is provided in braille standard, " he ", " she ", " word " need to make
With special representation method, mark is had to for " she " word and is adjusted.The blind symbol of " she " is " T9 ", and tone is the first sound, in blind symbol
It is expressed as " A ", the representation of braille string is " B after mark is adjusted!G*:T9AM0 D MUDIDI”.
Braille display module is called to include braille string blind on the aobvious device of point.
Claims (10)
- A kind of 1. method that Chinese character is read for blind person, which is characterized in that including:Step 1, Chinese language text is obtained, carries out participle operation to the Chinese language text, generates Chinese character string, by pronunciation dictionary, more Sound word dictionary and word frequency information with reference to the part-of-speech tagging that participle obtains, each word in the Chinese character string are converted to corresponding Phonetic is simultaneously connected as pinyin string;Step 2, by searching for phonetic and the control dictionary of blind symbol, the pinyin string is converted to the blind symbol string not segmented, is passed through Using braille participle is carried out to the blind symbol string with the trained participle model of statistical machine learning method in advance, generation is initial blind The Chinese character string with the initial braille participle is merged, new braille participle is generated, according to braille word link writing by text participle Rule is adjusted the new braille participle;Step 3, to carrying out braille mark tune according to the new braille participle after braille word link writing rule adjustment, generation is final blind Text participle shows the final braille participle.
- 2. the method for Chinese character is read for blind person as described in claim 1, which is characterized in that by the Chinese in the step 1 Word string be converted into pinyin string the specific steps are:Step 2.1 judges whether each word is multi-character words, if multi-character words, and is sending out for each word in the Chinese character string The corresponding phonetic of the multi-character words can be found in sound dictionary, then directly returns to the corresponding phonetic of the multi-character words, otherwise performs Step 2.2;Step 2.2 by the multiword word segmentation be Chinese character sequence, Chinese character all in the multi-character words is taken successively, to each Chinese character performs step 2.3 to 2.4;Step 2.3 judges whether the current Chinese character is polyphone for current Chinese character, lookup polyphone dictionary, if not multitone Word searches the phonetic of the current Chinese character in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;Step 2.4 then performs following steps if polyphone, the specific steps are:If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;If multi-character words, Then perform following step:For the polyphone w in multi-character wordsk, a) step, the word W with follow-up n word one n+1 words of compositionk,n=wkwk+1…wk+n, W is searched in polyphone phrase dictionaryk,n, such as find, then with Wk,nIn be searched the pronunciation of word as polyphone wkPronunciation And it returns;If do not found, then b) step is performed, the word W of a n+1 words is formed with the word of front nn-k,k=wn-kwn-kk+1…wn, W is searched in polyphone phrase dictionaryn-k,k, such as find, then with Wk,nIn be searched word pronunciation as polyphone pronunciation simultaneously It returns, does not search such as, then form the word W of a n words with the follow-up and word of front n-1 respectivelyk,n-1、Wn-k+1,k, to the multi-character words Perform respectively a), b) step, until determining the polyphone wkPronunciation;Step 2.4.2 assumes that the polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are Ppos, Weights are λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, system is Each pronunciation of the polyphone calculates a score Scorei, wherein Scorei=λ1·Ppos(tonei)+λ2·Plm (tonei)+λ3·Pfreq(tonei), take out the pronunciation of highest scoringAs polyphone final phonetic and return It returns.
- 3. the method for Chinese character is read for blind person as described in claim 1, which is characterized in that merged in the step 2 The step of be, for the Chinese character string C=c1c2…cmWith the initial braille participle B=b1b2…bn, wherein ci,bjTable respectively Show a participle in the Chinese character string and initial braille participle, B is segmented for the initial braille, B is mapped to pair The Chinese character string B '=b ' answered1b′2…b′n, wherein b 'jB is segmented for the initial braillejIt is mapped as the participle after Chinese.
- 4. the method for Chinese character is read for blind person as described in claim 1, which is characterized in that braille segments in the step 2 Combination handwriting rule is as follows:Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+nWord segmentation regulation:POSkFor activation condition, m and n expressions need to check the preceding m word and n word of current new braille participle respectively, if m and n All it is 0, then it represents that this is a word segmentation regulation, and what is represented after second colon is the part of speech combination of participle, if meeting the group It closes, then performs the operation after right arrow.
- 5. the method for Chinese character is read for blind person as described in claim 1, which is characterized in that braille described in the step 3 Mark adjust the specific steps are:The phonetic of the corresponding word of new braille participle after each adjustment is checked successively, and the rule in being assembled with braille mark carries out It compares, if meeting condition, current new braille is segmented into rower tune, the form that the braille mark is assembled is as follows:Mark adjusts rule:tonek:[n]:tonek…tonek+nWherein tonekFor the phonetic of current new braille participle, n is to need to check rear n new braille participles of current new braille participle Phonetic, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower tune.
- 6. a kind of system that Chinese character is read for blind person, which is characterized in that including:Pinyin string module is obtained, for obtaining Chinese language text, participle operation is carried out to the Chinese language text, generates Chinese character string, is led to Pronunciation dictionary, polyphone dictionary and word frequency information are crossed, with reference to the part-of-speech tagging that participle obtains, by each word in the Chinese character string It is converted to corresponding phonetic and is connected as pinyin string;It obtains new braille to segment and adjust module, for the control dictionary by searching for phonetic and blind symbol, the pinyin string is turned The blind symbol string not segmented is changed to, by using in advance with the trained participle model of statistical machine learning method to the blind symbol string Braille participle is carried out, generates initial braille participle, the Chinese character string is merged with the initial braille participle, generation is new blind Text participle is adjusted the new braille participle according to braille word link writing rule;Braille display module, for carrying out braille mark according to the new braille participle after braille word link writing rule adjustment It adjusts, generates final braille participle, the final braille participle is shown.
- 7. the system of Chinese character is read for blind person as claimed in claim 6, which is characterized in that in the acquisition pinyin string module By the Chinese character string be converted into pinyin string the specific steps are:Step 2.1 judges whether each word is multi-character words, if multi-character words, and is sending out for each word in the Chinese character string The corresponding phonetic of the multi-character words can be found in sound dictionary, then directly returns to the corresponding phonetic of the multi-character words, otherwise performs Step 2.2;Step 2.2 by the multiword word segmentation be Chinese character sequence, Chinese character all in the multi-character words is taken successively, to each Chinese character performs step 2.3 to 2.4;Step 2.3 judges whether the current Chinese character is polyphone for current Chinese character, lookup polyphone dictionary, if not multitone Word searches the phonetic of the current Chinese character in pronunciation dictionary and returns to the phonetic;Otherwise step 2.4 is performed;Step 2.4 then performs following steps if polyphone, the specific steps are:If the current polyphones of step 2.4.1 come from a monosyllabic word, step 2.4.2 is directly performed;If multi-character words, Then perform following step:For the polyphone w in multi-character wordsk, a) step, the word W with follow-up n word one n+1 words of compositionk,n=wkwk+1…wk+n, W is searched in polyphone phrase dictionaryk,n, such as find, then with Wk,nIn be searched the pronunciation of word as polyphone wkPronunciation And it returns;If do not found, then b) step is performed, the word W of a n+1 words is formed with the word of front nn-k,k=wn-kwn-kk+1…wn, W is searched in polyphone phrase dictionaryn-k,k, such as find, then with Wk,nIn be searched word pronunciation as polyphone pronunciation simultaneously It returns, does not search such as, then form the word W of a n words with the follow-up and word of front n-1 respectivelyk,n-1、Wn-k+1,k, to the multi-character words Perform respectively a), b) step, until determining the polyphone wkPronunciation;Step 2.4.2 assumes that the polyphone has tone1,...,tonenCommon n pronunciation, participle part of speech definition of probability are Ppos, Weights are λ1, probabilistic language model is defined as Plm, weights λ2, participle word frequency definition of probability is Pfreq, weights λ3, system is Each pronunciation of the polyphone calculates a score Scorei, wherein Scorei=λ1·Ppos(tonei)+λ2·Plm (tonei)+λ3·Pfreq(tonei), take out the pronunciation of highest scoringAs polyphone final phonetic and return It returns.
- 8. the system of Chinese character is read for blind person as claimed in claim 6, which is characterized in that described to obtain new braille participle simultaneously The step of being merged in adjustment module is, for the Chinese character string C=c1c2…cmWith the initial braille participle B=b1b2… bn, wherein ci,bjA participle in the Chinese character string and the initial braille participle is represented respectively, for the initial braille B is segmented, B is mapped into the corresponding Chinese character string B '=b '1b′2…b′n, wherein b 'jB is segmented for the initial braillejMapping For the participle after Chinese.
- 9. the system of Chinese character is read for blind person as claimed in claim 6, which is characterized in that described to obtain new braille participle simultaneously It is as follows to adjust braille word link writing rule in module:Combination handwriting rule:POSk:[m,n]:POSk-m+…+POSk+…+POSk+n→POSk-m…POSk+nWord segmentation regulation:POSkFor activation condition, m and n expressions need to check the preceding m word and n word of current new braille participle respectively, if m and n All it is 0, then it represents that this is a word segmentation regulation, and what is represented after second colon is the part of speech combination of participle, if meeting the group It closes, then performs the operation after right arrow.
- 10. the system of Chinese character is read for blind person as claimed in claim 6, which is characterized in that in the braille display module The braille mark tune the specific steps are:The phonetic of the corresponding word of new braille participle after each adjustment is checked successively, and the rule in being assembled with braille mark carries out It compares, if meeting condition, current new braille is segmented into rower tune, the form that the braille mark is assembled is as follows:Mark adjusts rule:tonek:[n]:tonek…tonek+nWherein tonekFor the phonetic of current new braille participle, n is to need to check rear n new braille participles of current new braille participle Phonetic, tonek…tonek+nTo mark tune condition, if pinyin sequence meets mark tune condition, to tonekInto rower tune.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510623525.5A CN105404621B (en) | 2015-09-25 | 2015-09-25 | A kind of method and system that Chinese character is read for blind person |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510623525.5A CN105404621B (en) | 2015-09-25 | 2015-09-25 | A kind of method and system that Chinese character is read for blind person |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105404621A CN105404621A (en) | 2016-03-16 |
CN105404621B true CN105404621B (en) | 2018-07-10 |
Family
ID=55470115
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510623525.5A Active CN105404621B (en) | 2015-09-25 | 2015-09-25 | A kind of method and system that Chinese character is read for blind person |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105404621B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107203508A (en) * | 2016-03-17 | 2017-09-26 | 富士施乐实业发展(中国)有限公司 | Braille document generating method and system |
CN107273357B (en) * | 2017-06-14 | 2020-11-10 | 北京百度网讯科技有限公司 | Artificial intelligence-based word segmentation model correction method, device, equipment and medium |
CN107368474B (en) * | 2017-07-07 | 2020-08-04 | 浙江理工大学 | Automatic efficient translation and conversion method from Chinese to braille |
CN107886808B (en) * | 2017-11-03 | 2021-03-09 | 中国科学院计算技术研究所 | Braille square auxiliary labeling method and system |
CN108062886A (en) * | 2017-11-03 | 2018-05-22 | 中国科学院计算技术研究所 | Braille point interactive mode mask method and system |
CN108052936B (en) * | 2017-11-03 | 2021-06-29 | 中国科学院计算技术研究所 | Automatic inclination correction method and system for Braille image |
CN108491441B (en) * | 2018-02-12 | 2022-02-01 | 北京联合大学 | Braille information statistical system |
CN108461111A (en) * | 2018-03-16 | 2018-08-28 | 重庆医科大学 | Chinese medical treatment text duplicate checking method and device, electronic equipment, computer read/write memory medium |
CN110920268B (en) * | 2019-11-19 | 2021-05-28 | 西安交通大学 | Braille inscription method and system |
CN111078898B (en) * | 2019-12-27 | 2023-08-08 | 出门问问创新科技有限公司 | Multi-tone word annotation method, device and computer readable storage medium |
CN113035026B (en) * | 2021-03-10 | 2022-06-17 | 之江实验室 | Audio-visual tactile perception matching method without barriers for braille information |
CN116432603B (en) * | 2023-03-27 | 2023-10-13 | 之江实验室 | Memory and calculation integrated Chinese braille chip |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1323003A (en) * | 2001-06-22 | 2001-11-21 | 清华大学 | Intelligent Chinese computer system for the blind |
CN1323004A (en) * | 2001-06-08 | 2001-11-21 | 清华大学 | Automatic conversion method from Chinese braille to Chinese character |
WO2002006916A3 (en) * | 2000-07-18 | 2003-10-30 | Yishay Langenthal | Reading aid for the blind |
CN1591414A (en) * | 2004-06-03 | 2005-03-09 | 华建电子有限责任公司 | Automatic translating converting method for Chinese language to braille |
CN102184172A (en) * | 2011-05-10 | 2011-09-14 | 中国科学院计算技术研究所 | Chinese character reading system and method for blind people |
-
2015
- 2015-09-25 CN CN201510623525.5A patent/CN105404621B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002006916A3 (en) * | 2000-07-18 | 2003-10-30 | Yishay Langenthal | Reading aid for the blind |
CN1323004A (en) * | 2001-06-08 | 2001-11-21 | 清华大学 | Automatic conversion method from Chinese braille to Chinese character |
CN1323003A (en) * | 2001-06-22 | 2001-11-21 | 清华大学 | Intelligent Chinese computer system for the blind |
CN1591414A (en) * | 2004-06-03 | 2005-03-09 | 华建电子有限责任公司 | Automatic translating converting method for Chinese language to braille |
CN102184172A (en) * | 2011-05-10 | 2011-09-14 | 中国科学院计算技术研究所 | Chinese character reading system and method for blind people |
Non-Patent Citations (4)
Title |
---|
EasyBraille:中文汉语盲文自动转换系统;朱小燕,包塔;《自然语言理解与机器翻译——全国第六届计算语言学联合学术会议论文集》;20010801;326-331 * |
汉字—盲文转换系统的设计;杨潮,车磊;《北京印刷学院学报》;20111231;第19卷(第6期);第4节,图4 * |
汉语—盲文机器翻译系统的研究与实现;李宏乔 等;《计算机应用》;20021110;第22卷(第11期);第2.3节,第3.2节,第3.4节 * |
面向统计机器翻译的领域自适应方法研究;苏晨;《中国优秀硕士学位论文全文数据库 信息科技辑》;20150815;第I138-765页正文第22页第3.3节 * |
Also Published As
Publication number | Publication date |
---|---|
CN105404621A (en) | 2016-03-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105404621B (en) | A kind of method and system that Chinese character is read for blind person | |
CN107741928B (en) | Method for correcting error of text after voice recognition based on domain recognition | |
CN106598939B (en) | A kind of text error correction method and device, server, storage medium | |
CN105957518B (en) | A kind of method of Mongol large vocabulary continuous speech recognition | |
US8131539B2 (en) | Search-based word segmentation method and device for language without word boundary tag | |
CN109637537B (en) | Method for automatically acquiring annotated data to optimize user-defined awakening model | |
US9613621B2 (en) | Speech recognition method and electronic apparatus | |
WO2018153213A1 (en) | Multi-language hybrid speech recognition method | |
CN104166462A (en) | Input method and system for characters | |
CN109241540A (en) | A kind of blind automatic switching method of Chinese based on deep neural network and system | |
CN110083711A (en) | A kind of phonetic transcriptions of Chinese characters conversion method and converting system | |
CN103810993B (en) | Text phonetic notation method and device | |
US20180089176A1 (en) | Method of translating speech signal and electronic device employing the same | |
Stein et al. | Hand in hand: automatic sign language to English translation | |
CN109754791A (en) | Acoustic-controlled method and system | |
CN107229611B (en) | Word alignment-based historical book classical word segmentation method | |
JP2011008784A (en) | System and method for automatically recommending japanese word by using roman alphabet conversion | |
CN111429886B (en) | Voice recognition method and system | |
JP2006235916A (en) | Text analysis device, text analysis method and speech synthesizer | |
Dinarelli et al. | Concept segmentation and labeling for conversational speech | |
Saychum et al. | Efficient Thai Grapheme-to-Phoneme Conversion Using CRF-Based Joint Sequence Modeling. | |
KR101777141B1 (en) | Apparatus and method for inputting chinese and foreign languages based on hun min jeong eum using korean input keyboard | |
JP2001229162A (en) | Method and device for automatically proofreading chinese document | |
Scherbakov et al. | VectorWeavers at SemEval-2016 Task 10: From incremental meaning to semantic unit (phrase by phrase) | |
CN112786002B (en) | Voice synthesis method, device, equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |