EP1711936A2

EP1711936A2 - Method for automatic correspondence between graphical and phonetic elements

Info

Publication number: EP1711936A2
Application number: EP04816413A
Authority: EP
Inventors: Edmond Lassalle
Original assignee: France Telecom SA
Current assignee: Orange SA
Priority date: 2003-12-18
Filing date: 2004-12-17
Publication date: 2006-10-18
Also published as: US20070055515A1; WO2005062292A3; WO2005062292A2; FR2864281A1

Abstract

The invention relates to a method for automatically segmenting any graphical chain into graphemes and any phonetic chain into phonemes by rewriting graphical chains (words) into phonetic chains. The inventive method consists in assessing first probabilities of the rewrite of graphical elements into phonetic elements (E2), in determining (E3-E9) second probabilities (p(g1, gm/p1, ,pn)) of MN rewrites of M graphical chains successively concatenating M graphical elements into N phonetic chains successively concatenating N graphical elements for each rewrite of a given graphical chain having M graphical elements into a corresponding phonetic chain having N graphical elements and in establishing links between the last elements (gm, pn) of the graphical and phonetic chains of the second rewrites in order to constitute, a path segmenting the given graphical chain into graphemes corresponding to phonemes segmenting the corresponding phonetic chain, respectively in the matrix MN of the second probabilities.

Description

Method for automatic correspondence between graphic elements and phonetic elements

The present invention generally relates to the automatic extraction of linguistic knowledge from a corpus of transcriptions of graphic chains into phonetic chains. More particularly, it relates to the transcription of typographic elements such as characters in a predetermined language into phonetic elements.

Currently, each word of a language constitutes a graphic chain which is transcribed phonetically into a chain of phonemes by a phonetician. For any new word to add to a learning corpus, the phonetician must intervene to transcribe this new word phonetically. The learning corpus thus provides only global grapheme / phoneme transcriptions. For example in the global transcription "lane" / [ryε 1], the corpus indicates that overall, the graphic chain "lane" translates into phonetic chain. However, it is not explained that in any way, unitarily, the typographic element "r" is transcribed phonetically. The global transcription does not also indicate the syllables or graphemes making up the graphic chain and the phonetic elements making up the phonetic chain. Now knowing the elementary transcription of each typographic element makes it possible thereafter, by character-by-character analysis of any graphic chain, to determine one or more phonetic chains associated with the graphic chain. Phonetic transcriptions are useful error correction systems to recognize lexical errors when entering text on a keyboard. There is therefore a need from a raw transcription to extract finer elementary transcriptions.

The invention aims to automatically deduce raw transcriptions of graphic strings, such as words and surnames, for example, into phonetic strings, transcriptions of graphic elements, such as characters, into phonetic elements composing the phonetic strings in order to automatically segment any graphic chain into graphemes and any phonetic chain into phonemes. The elementary transcriptions graphic element by graphic element, that is to say character by character, then facilitate the automatic global transcription of any additional graphic chain brought to the corpus of graphic chains, on the basis in particular of a concatenation of phonetic elements unequivocally corresponding to the characters of the additional graphic string. To this end, a method according to the invention automatically corresponds graphic elements making up given graphic strings to phonetic elements making up corresponding phonetic chains, after having initially entered global transcriptions of graphic chains into phonetic chains in a database accessible by the computer and having estimated and recorded in the base of the first probabilities of elementary transcriptions of the graphic elements respectively into the phonetic elements. The method is characterized by the following stages: for each transcription of a graphic chain given to M graphic elements into a phonetic chain corresponding to N phonetic elements, determine second probabilities of MN second transcriptions of M graphic chains successively concatenating the M graphic elements in N phonetic chains successively concatenating the N phonetic elements, each according to a respective first probability and the greatest of three respective second probabilities determined previously, and establishing and memorizing a link between the last elements of the graphic and phonetic chains of each second transcription and the last elements of the graphic and phonetic chains of the transcription relating to the largest of the three respective second probabilities so that links established in a matrix of size MN relating to the second probability és constitutes a unique path between the last and first pairs of graphical and phonetic elements of the matrix to segment the given graphic chain into graphemes corresponding respectively to phonemes segmenting the corresponding phonetic chain and to record the correspondences between the graphemes and phonemes in basically, the number of graphic elements in a grapheme being identical to the number of phonetic elements in the corresponding phoneme, so that any new graphic chain is automatically transcribed into a phonetic chain segmented into phonemes by means of recorded correspondences. According to other features of the invention, the respective first probability for determining a second probability relating to a second transcription of a graphic chain concatenating m graphic elements into a phonetic chain concatenating n phonetic elements, with 1 <m < M and l ≤ n ≤ N, is relative to the last elements in the graphic chain with m graphic elements and the phonetic chain with n phonetic elements. The three respective second probabilities previously determined for the second transcription of the graphic chain with m graphic elements in the phonetic chain with n phonetic elements preferably relate respectively to a second transcription of a graphic chain with m-1 graphic elements in the chain phonetic with n phonetic elements, a second transcription of the graphic chain with m graphic elements into a phonetic chain with n-1 phonetic elements and a second transcription of the graphic chain with m-1 graphic elements into the phonetic chain with n-1 elements phonetic. For example, the invention transcribes phonetically from the corpus of global transcriptions such as "lane" | [ryεl] the graphic elements "r", "u", "e", "lie" respectively in the phonetic elements [r], [y], [ε], [1]. The invention can be likened to a syllabation which makes it possible by analysis to decompose a global transcription into elementary transcriptions, and to locally match grapheme / phoneme sub-transcriptions. The division into initial graphemes and phonemes and the one-to-one correspondence of each graphic element to each phonetic element of the cut phonemes is called grapheme | phoneme alignment. According to the previous example, the invention produces the following alignment: M -. ^H "u""e""binds" [r] [y] [ε] [1 **]. The symbol * indicates a silent and meaningless phonetic element.

Other characteristics and advantages of the present invention will appear more clearly on reading the following description of several preferred embodiments of the invention, by way of nonlimiting examples, with reference to the corresponding appended drawings in which: FIG. 1 is an algorithm of main steps of the automatic correspondence method according to the invention; and - Figure 2 is a sub-step algorithm of a step of determining first individual probabilities included in the automatic matching method.

As shown in FIG. 1, the method of automatic correspondence of graphic elements and phonetic elements according to the invention comprises main steps El to Eli. These steps are for the most part implemented for example in the form of software i implemented in a terminal, such as a personal computer or a mobile in a cellular radiocommunication network, and linked in particular to a software system for spell checking. of lexical errors which can be integrated into a word processing system or a system of linguistic exercise. The terminal contains or can access a database of the type used in artificial intelligence. The database stores a corpus C of initial global transcriptions. Initially in step E1, the global transcriptions (CG | CP) are formed by pairs each matching a graphic chain CG, such as a word in a predetermined language or a patronymic name, with a phonetic chain CP. These transcriptions were determined and entered by a phonetician using an appropriate form displayed by the computer. Corpus C makes GC graphic chains each composed of one or more typographical elements (characters), called graphic elements g ^ of an alphabet G = {STi / ••• / 9ι) correspond to I elements in the predetermined language, with 1 <i <M, respectively to phonetic chains CP each composed of one or more phonetic elements p of an alphabet P = {Pi, ..., P _j } to J phonetic elements with 1 <j <J and I ≠ J a priori. However, at this stage, the segmentation of the chain CG into syllables or graphemes each comprising one or more graphic elements is ignored, and the segmentation of the chain CP into phonemes each comprising one or more phonetic elements. Typically, the alphabets G and P have around thirty elements. They thus present a possibility of 30 × 30,900 possible pairs of graphic element and of phonetic element. In practice, the corpus C contains at least 100,000 global transcriptions of typographic strings CG into phonetic strings CP, which preserves the invention from gross errors in probability estimates, as will be seen below. In step E2, the first elementary transcription probabilities P (gjjp- _j ) so that an element graph g ^ corresponds to the phonetic element p are first estimated and recorded in the database with the corpus of global transcriptions C. The estimated values of the first probabilities are as close as possible respectively to the maximum probability values sought so that the method of the invention operating by iterations converges quickly while avoiding retaining local maxima. The concatenative nature of the global transcriptions of the chains leads to the hypothesis of a correlation between the rank r _q of the graphic elements in a graphic chain CG and the rank r _p of the phonetic elements in the corresponding phonetic chain CP. For example in the global transcription (beau | bo), it is more likely that the graphic element b, by its position at the beginning of the CG chain, translates into a phonetic element [b] rather than it translates into [ o] phonetics positioned at the end of the corresponding chain CP. In this example, the correlation of ranks brings the graphic elements [b] and [e] closer to the phonetic element [b], and the graphic elements [a] and [u] to the phonetic element [o]. The initial estimation algorithm E2 of the first probabilities P (g _j jpj) comprises the following sub-steps E21 to E27. In sub-step E21, IJ contingency numbers ^κ gi _p j 'respectively associated with the elementary transcriptions (jjp _j ) of a graphic element of the alphabet G and of a phonetic element of the alphabet P are set to zero . The number of contingencies K _qj _ _p is equal at the end of step E2 to the number of times estimated where the graphic element gj is transcribed in the phonetic element pj in the various global transcriptions of typographic chains CG into phonetic chains CP included in the corpus C. For each chain transcription (CG | CP), as indicated in substep E22, the ranks of the graphic elements in the chain CG and the ranks of the phonetic elements in the chain CP are normalized as a function of the respective lengths l _g and 1 „of the chains CG and CP which may be different. In sub-step E23, the rank r of a phonetic element in the chain CP is deduced from the rank r _{g; j} _ of a graphic element g ^ in the chain CG with which the phonetic element of rank r will be associated, according to the following relation: r = whole part (r _gj _. l _p / l _q ). The number of contingencies K _g ^ _pj associated with the elementary transcription of the graphic element g. _j _ in the phonetic element pj is then incremented by 1 only if the phonetic element p-4 is located at the deduced rank r in the chain CP, as indicated in substeps E24 and E25. Sub-steps E22 to E25 are repeated for each global transcription (CG | CP) of the corpus C, as indicated in sub-step E26. When all the global transcriptions of the corpus have been traversed, the following sub-step 26 estimates all the first probabilities P (g- p) of elementary transcription between the graphic elements and the phonetic elements, according to the following relationships for each graphic element g _j _: j = JP (9ilPj) = K _gip j / ∑ Kgipj j = ι after calculating the term sum in the denominator for the graphic element g- _j .. Returning to FIG. 1, the correspondence process is continued by steps E3 to E10 which segment each graphic chain CG read in the corpus of the database in order to automatically correspond in a one-to-one manner each segment of the chain CG , called grapheme, comprising one or more one-segment graphic elements, called phoneme, comprising one or more phonetic elements resulting from a segmentation of the corresponding phonetic chain CP. A graphic chain CG comprises M consecutive graphic elements g ^ to g _M and the phonetic chain CP corresponding to the chain CG comprises N consecutive phonetic elements p ₁ to p _N with the integer N different, or possibly equal to the integer M. The probability P (g _{1 #} ... g _m , ... g _M \ p _lr ... p _n , ... p _N ) so that the chain CG corresponds to the chain CP, with 1 <m <M and l ≤ n ≤ N, is determined as a function of the first elementary transcription probabilities P (SfjJpj) estimated and recorded previously in step E2, and of a similarity between the chains CG and CP. Similarity is based on the editing distance of Damerau-Levenshtein DLM (Damerau-Levenshtein Metric), but by maximizing and not minimizing. The probability P (CG | CP) is determined by dynamic programming, using the following iteration formula for any couple m, n such that 1 <n ≤ N and l ≤ m ≤ M:

P ( _gι g ₂ ... g _m | p _! P _2. • .p _n ) = P (g _m | p _n ) max [P (g _ιg2 ... g _m _ι | p ₁ p _2. • - P _n ) _•

P (9ig ₂ - ^{• •} 9mlPlP2- ^• -Pn-l) / P (9l92- ^• ^• -Pn-1> 1 ^• The concatenative nature of global chain transcriptions and grapheme / phoneme transcriptions makes it possible to apply effective Markov models. For the given probability of a transcription of a chain g ^, g ₂ • • • _m ^into a chain p ₁ p ₂ ... p _n , the extension of the graphic chain, respectively phonetic, by a new graphic element g _{m +} ι, respectively P _{n +} i gives rise either to the same phonetic chain, respectively graphic, or to the addition of a new phonetic element, respectively graphic. Expressed in terms of probability, P (χg ₂ - • • g _m + ιlPιP2 • • - n _{+ l} ) ^only depends on the probabilities of three possible transcriptions: either P (g _x g ₂ ... g _m | pχP ₂ . ..p _{n + 1} ) either P ⁽ g ₁ g ₂ .-. g _m + ιlPιP2- ^• -Pn> or P (g _x g ₂ ... g ^ • • • P _n > • This dependence is expressed by the editing distance equal to the greater of the three probabilities indicated above After having set the indices m and n to zero for a global transcription (CG | CP) in step E3 and incrementing the indices m and n by 1 in steps E4 and E5, iterations begin in steps E6 and E7 by determining the probabilities so that the M successive concatenations of the graphic elements g ₁ to g _M of the chain CG correspond to the first phonetic element p ^ _ of the chain CP, either: P (9lz ••• g _m IPl> = ^p (gmlPl) max [P (g ₁ , ... g _m -ιlPι) 3 with 1 <m <M, starting with the elementary probability P (gιlPι) Then as illustrated by Step E8, the method is continued by iterations to determine the probability that M concatenations graphics g- _j _ to g _M GC chain correspond to the first two phonetic elements -p ^ and p ₂ of the chain PC, using the probabilities previously determined for the first graphic element P _] _, that is: P (9 _l , • • - g _m IP _l '2) = P (g _m I 2) ma [P (g!,... G _m _ι | p ₂ ), p (gι, • •. G _m | Pι). (ι ^• • - gm-ιl ι> l ^• Then the process is continued by adding a phonetic element p _n to determine the M probabilities P (gi i,. • -p _n ) to P (g- ^ ..., g _M | Pι, • - -P _n ) up to M probabilities relating to the chain CP = (p- _] _, ... u) The computer constructs and memorizes gradually by iterations of steps E4 to E8, a matrix of second probabilities P (g ₁ , ... g _m | p- _j _, ... p _n ) with M columns for successive concatenations of the M graphic elements and with N lines for successive concatenations of the N phonetic elements, by operating line by line according to the example above and starting with the probability (gιl ι) and ending with the probability P (g ₁ , ... g _M | pι, ... p _N ). Each iteration relating to the (mn) th transcription [(g _lΛ ... g _m ) | (pi, ... p _n )] establishes a link between the pair (g _m , P _n ) and the pair with the greatest probability of the three probabilities previously determined p among the three couples (9m- _l 'n)'

(9π nl) ^and (gm-l'Pn-1 ^ • ^The li ^is stored in the computer. When the torque _(m g _/ P _n) is connected to the pair _(Sm-l 'n) ^ s' acts of an elementary transcription of ( ^• gm-i'9m) ⁱⁿ gm 'when the couple (g _m , p _n ) is connected to the couple (9m'Pn-l)' ^ it is an elementary transcription of g _m in (p _n -i / n) ^; and when the couple (g _m / P _n ) is linked to the couple (g _m _ 1 / Pn-ι) 'ϋ it is an elementary transcription of g _m in p _n . Thus, at each probability determination P (g _1; ... g _m ) | ( _] _, ... p _n ) a link is memorized in the computer. The links trace a unique path also gradually memorized in the computer and connecting the first couple (g- _| _, p _x ) to the last couple (g _M , p _N ) in the matrix with M columns and N rows. The single path topology in the size matrix MN segments the graphic chains CG into graphemes and the phonetic chains CP into phonemes and aligns the graphic elements and the phonetic elements in one-to-one correspondence. If a segment of the path follows a portion of a line between two graphic elements, the concatenation of the graphic elements of the line portion corresponds to the phonetic element of the line supplemented by one or more silent and meaningless phonetic elements in order to form a pair of graphemes and phonemes having the same number of elements, which pair is stored in the computer. If a segment of the path follows a portion of column between two phonetic elements, the graphic element of the column supplemented by one or more graphic elements without meaning corresponds to the concatenation of the phonetic elements of the portion of column in order to form a couple of graphemes and of phoneme having the same number of elements, which couple is memorized in the computer. A change of direction of the path towards the horizontal, the vertical or the diagonal in the matrix indicates a segmentation of the chains CG and CP. As a simple example, we seek to segment the global transcription of the word CG = "beautiful" into the phonetic chain CP = [bo] assuming that step E2 has estimated the following first individual probabilities in the corpus C:

P (b | b) = 0.9; P (e | b) = 0.1; P (a | b) = 0.1; P (u | b) = 0.1, P (e | o) = 0.2; P (a | o) = 0, l; P (u | o) = 0.2; P (b | o) = 0, l. For the transcription (beau | bo) of the corpus, the M = 4 iterations of steps E5, E6 and E7 for each M = 2 rows of the size matrix (4.2] produce the following table:

The symbol <- indicates that the couple (g _m , p _n ) is linked to the couple (g _m -i Pn _. ) '^ - ^e symbol indicates that the couple (g _m , p _n ) is linked to the couple (g _m , p _n _ - _[ _); and the symbol indicates that the couple (g _m , p _n ) is related to the couple (g _m -i / Pn- ₁ ^ • ^The symbol t \ associated with the transcription (be | bo) indicates that the latter is deduced and therefore linked to the transcription (b | b) which precedes it. The symbol indicates a segmentation border between pairs of graphemes and phonemes. The following alignment is deduced from this table: b eau bo **. The symbol * designates a silent phonetic element In order to perfect the correspondences between the graphemes and the phonemes and the correspondences between the graphic elements and the phonetic elements, preferably as indicated by the step Eli, the first probabilities P (gι | pι) to (P ( g _j | p _j ) transcriptions of each of the graphic elements respectively into the J phonetic elements (step E2) and in particular the contingency numbers ^κ g _l p _l to K _gI pj (sub-step E25) are again estimated in function in particular of the ranks of elements p honetic placed in the given phonetic chains CG which were segmented into phonemes in the previous step E10. Again second probabilities P (g _x , ... g pi • • - _n ) ^{of MN} second transcriptions of each global transcription of a graphic chain given to M graphic elements (CG) into a corresponding phonetic chain (CP) to N phonetic elements are determined by the execution of steps E3 to E10 so that in the next step E10 links are established between couples (g _m / P _n ) of a new matrix with M columns and N rows and therefore a corrected path connecting the last couple (9M / PN) ^{to the} First couple (gχ, Pι) in the new matrix of second probabilities of size MN. Possibly, thanks to the high processing capacity and speed of the computer, other iterative loops of steps E2 to Eli can be executed in the computer until the convergence of the correspondence process, that is to say - say until the established path becomes constant from one loop to the next. After the segmentation of all the graphic and phonetic chains of the corpus G into graphemes and phonemes, the base recorded all the correspondences between the graphic and phonetic elements and the correspondences between the graphemes and phonemes for all the corpus C traversed.

Any new graphic chain added to the corpus can then be automatically transcribed into a phonetic chain segmented into phonemes using in particular the correspondences previously established and recorded according to the invention, which progressively enriches the corpus in the database and increases the accuracy of transcriptions. As already said, phonetic transcriptions are useful for correcting software systems misspelling to recognize lexical mistakes when entering text on a terminal keyboard. Thus when the new graphic chain added to the corpus is entered on a keyboard of a terminal, the phonetic chain segmented into phonemes by means of the recorded correspondences is used for a spelling correction of the new graphic chain entered. The method of the invention can also be used as a tool for automatically generating short SMS messages from text written in everyday language. To do this, it requires a learning corpus C, the transcriptions of which are adapted to the automatic generation of short messages and which correspond respectively to graphic chains CG, such as words and phrases, to phonetic chains CP including "phonemes". are phonetically readable by any non-phonetic person. For example, the corpus establishes the following correspondences in French between graphic and phonetic chains: air R busy OQP case K. Thus a new graphic chain entered in a terminal is automatically transcribed by the method of the invention into a phonetic chain segmented into phonemes readable by any non-phonetic person by means of the correspondences recorded to be included in a message. short. According to the previous example, the sentence in French "I look busy" entered in the terminal is automatically transcribed into a short message according to Gl 'ROQP to be transmitted by the terminal, "phonetic strings" [G], [1 '], [R] and [OQP] being phonetically readable by any non-phonetic user. As a variant, the "phonetic chains" [G], [1 '], [R] and [OQP] can be assimilated to phonetic elements to constitute a phonetic chain [Gl'ROQP].

According to a preferred implementation of the method of the invention, the steps of the method of the invention are determined by the instructions of a computer program incorporated in a computer such as a terminal, a personal computer, a server or any other computer system. The program automatically matches graphic elements composing given graphic chains to phonetic elements making up corresponding phonetic chains, after having initially entered global transcriptions of graphic chains into phonetic chains in a database accessible by the computer and having estimated and recorded in the base of the first probabilities of elementary transcriptions of the graphic elements respectively into the phonetic elements. The program includes program instructions which, when said program is loaded and executed in the computer whose operation is then controlled by the execution of the program, carry out the steps of the method according to the invention. Consequently, the invention also applies to a computer program, in particular a computer program on or in an information medium, suitable for implementing the invention. This program can use any programming language, and be in the form of source code, object code, or intermediate code between code source and object code such as in a partially compiled form, or in any other form desirable for implementing the method according to the invention.

Claims

1 - Process implemented in a computer to automatically correspond graphic elements (g- _j _) composing given graphic strings to phonetic elements (p _j ) composing corresponding phonetic chains, after having entered (El) transcriptions initially global (CG | CP) of the graphic chains into the phonetic chains in a base accessible by the computer and having estimated and recorded in the base (E2) of the first probabilities (P (gjjp _j )) of elementary transcriptions of the graphic elements respectively phonetic elements, characterized by the following steps: for each transcription of a given graphic chain (CG) with M graphic elements into a corresponding phonetic chain (CP) with N phonetic elements, determine (E3 - E9) of the second probabilities (P (g _l7 ... g _m | p ₁ , ... p _n )) of MN second transcriptions of M graphic chains concatenating successively the M graphic elements in N phonetic chains successively concatenating the N phonetic elements, each according to a respective first probability and the greatest of three respective second probabilities determined previously, and establishing and memorizing (E10) a link between the last elements ('n) graphic and phonetic chains of each second transcription and the last elements of the graphic and phonetic chains of the transcription relating to the largest of the three respective second probabilities so that links established in a matrix of size MN relating to the second probabilities constitutes a unique path between last and first pairs of graphic and phonetic elements of the matrix to segment the given graphic chain into graphemes corresponding respectively to phonemes segmenting the corresponding phonetic chain and to record the correspondences between the graphemes and phonemes in the base, the number of graphic elements in a grapheme being identical to the number of phonetic elements in the corresponding phoneme, so that any new graphic chain is automatically transcribed into a phonetic chain segmented into phonemes by means of the recorded correspondences. 2 - Method according to claim 1, according to which the first respective probability for the determination (E3 - E9) of a second probability (P (gL, .. - _m | Pi / • ..p _n )) relating to a second transcription of a graphic chain concatenating m graphic elements into a phonetic chain concatenating n phonetic elements, with 1 <m <M and 1 ≤ n <N, relates to the last elements in the graphic chain with m graphic elements and the phonetic chain with n phonetic elements.

3 - Process according to claim 1 or 2, according to which the three respective second probabilities determined previously for the second transcription of the graphic chain with m graphic elements into the phonetic chain with n phonetic elements respectively relate to a second transcription of a graphic chain with m-1 graphic elements in the phonetic chain with n phonetic elements, a second transcription of the graphic chain with m graphic elements in one phonetic chain with n-1 phonetic elements and a second transcription of the graphic chain with m-1 graphic elements into the phonetic chain with n-1 phonetic elements.

4 - Method according to any one of claims 1 to 3, comprising an estimation of other first probabilities (P (gjJPj)) of transcriptions of each of the graphic elements respectively into the phonetic elements depending in particular on the ranks of the phonetic elements placed in the given phonetic chains (CG) which have been segmented into phonemes in order to again determine (E6) second probabilities (P (g _l7 ... g _m | i • • -Pn) of MN second transcriptions of each transcription d a graphic string given to M graphics (CG) into a corresponding phonetic string (CP) N phonetic elements and establish a fixed path from the last couple (9M'PN) ^to ^Prof. Emier couple (g_ Pι) in a new matrix of second probabilities of size MN.

5 - Method according to any one of claims 1 to 4, according to which the new graphic chain is entered on a keyboard of a terminal and the phonetic chain segmented into phonemes by means of the recorded correspondences is used for a spelling correction of the new graphic chain entered.

6 - Process according to any one of claims 1 to 4, according to which the phonetic strings are phonetically readable by any non-phonetic person, and the new graphic chain is automatically transcribed into a phonetic chain segmented into phonemes readable by any non-phonetic person by means of recorded correspondence to be included in a short message.

7 - Computer program able to be implemented in a computer to automatically correspond graphic elements (g ^) making up given graphic strings to phonetic elements (p- _j ) making up corresponding phonetic strings, after entering ( El) initially global transcriptions (CG | CP) of graphic strings into phonetic chains in a base accessible by the computer and having estimated and recorded in the base (E2) of the first probabilities (P (gil -)) of elementary transcriptions graphic elements respectively into phonetic elements, said program comprising instructions which, when the program is loaded and executed in the computer, carry out the following steps: for each transcription of a given graphic chain (CG) with M graphic elements in a corresponding phonetic chain (CP) with N phonetic elements, determine (E3 - E9) of the second pro babilities (P (g- ^ ... g _m | Pι, .. -P _n )) of MN second transcriptions of M graphic chains successively concatenating the M graphic elements into N phonetic chains successively concatenating the N phonetic elements, each in function of a respective first probability and the greatest of three respective second probabilities determined previously, and establishing and memorizing (E10) a link between the last elements (g _m / P _n ) d The graphic chains and phonetics of each second transcription and the last elements of the graphic and phonetic chains of the transcription relating to the largest of the three respective second probabilities so that links established in a matrix of size MN relating to the second probabilities constitutes a unique path between last and first pairs of graphical and phonetic elements of the matrix to segment the given graphic chain into graphemes corresponding respectively to phonemes segmenting the corresponding phonetic chain and to record the correspondences between graphemes and phonemes in the base, the number of graphic elements in a grapheme being identical to the number of phonetic elements in the corresponding phoneme, so that any new graphic chain is automatically transcribed into a phonetic chain segmented into phonemes by means of recorded correspondences.