Rhythm text generation method and device applied to dialects

Abstract

This invention discloses a method and apparatus for generating prosodic text in dialects, comprising: segmenting the text to be annotated in the target dialect to obtain original units; determining a target dialect pinyin code matching the target dialect type; encoding any original unit based on the target dialect pinyin code to obtain a corresponding target unit; and sorting all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated to obtain the dialect prosodic text. It can be seen that this invention uses the target dialect pinyin code to sequentially encode the text to be annotated in units of original units, and converts the text to be annotated into dialect prosodic text that conforms to the pronunciation characteristics of the dialect through the dialect prosodic features carried by the target dialect pinyin code. This can capture its accurate prosodic characteristics, thereby improving the accuracy of dialect prosodic text generation, and further contributing to improving the accuracy of dialect prosodic expression during speech synthesis.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and in particular to a method and apparatus for generating prosodic text for dialects. Background Technology

[0002] Prosodic text is a text representation that adds prosodic information to ordinary text. The prosodic information embedded in prosodic text includes the pronunciation of Chinese characters and numbers. This prosodic information is an important component of language, significantly influencing its rhythm and melody. Prosodic text has a wide range of applications, including but not limited to text-to-speech (TTS), speech recognition, and speech sentiment analysis. In speech synthesis, prosodic text can help generate more natural and expressive speech. In speech recognition, prosodic information can help improve the accuracy of sentence structure and punctuation recognition. In speech sentiment analysis, prosodic information can serve as an important clue to emotional state.

[0003] However, existing prosodic texts are often generated based on the pronunciation rules of Mandarin. Since the prosodic characteristics of dialects differ greatly from those of standard languages ​​such as Mandarin, when generating prosodic texts of dialects based on Mandarin data, it may be impossible to capture their accurate prosodic characteristics. As a result, when speech is synthesized based on these prosodic texts, the generated speech cannot accurately express the prosodic features of the dialect. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and apparatus for generating prosodic text for dialects, which can generate prosodic text that conforms to the prosodic characteristics of dialects and improve the accuracy of dialect prosodic expression during speech synthesis.

[0005] To address the aforementioned technical problems, the first aspect of this invention discloses a method for generating prosodic text applicable to dialects, the method comprising:

[0006] The text to be labeled in the target dialect is segmented according to a preset text segmentation method to obtain all the original units of the text to be labeled. All the original units include one or more of Chinese characters, Pinyin strings and number strings.

[0007] Based on the type of the target dialect, a target dialect pinyin code that matches the target dialect is determined from a plurality of pre-determined dialect pinyin codes. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect.

[0008] For any of the original units, the original unit is encoded based on the target dialect pinyin encoding to obtain the target unit of the original unit;

[0009] Based on the sorting order of the original units corresponding to each target unit in the text to be annotated, all target units are sorted to obtain the dialect prosodic text corresponding to the text to be annotated.

[0010] As an optional implementation, in the first aspect of the present invention, the method for generating prosodic text applied to dialects further includes:

[0011] For any one of the sample dialects in the multiple sample dialects, perform phonetic analysis on all phonemes in the sample dialect to extract basic phonetic features, wherein the multiple sample dialects include the target dialect;

[0012] Based on the basic phonetic features, a set of dialect initial consonant strings and a set of dialect final vowel strings are determined. The set of dialect initial consonant strings includes all dialect initial consonant strings representing the pronunciation of the beginning part of all syllables in the sample dialect. The set of dialect final vowel strings includes all dialect final vowel strings representing the pronunciation of the non-beginning part of all syllables in the sample dialect.

[0013] Based on the basic speech features, a set of dialect tone characters is determined, which includes all dialect tone characters representing all pronunciation tones in the sample dialect.

[0014] For any syllable of the sample dialect, the syllable pinyin code is determined according to the dialect initial consonant string, dialect final vowel string and dialect tone character corresponding to the syllable in a predetermined dialect pinyin encoding order. The syllable pinyin codes corresponding to all syllables of the sample dialect are used as the dialect pinyin codes corresponding to the sample dialect, wherein the dialect initial consonant string is a string that can be omitted.

[0015] Determine the dialect pinyin codes corresponding to all the sample dialects, which are multiple pre-determined dialect pinyin codes.

[0016] As an optional implementation, in the first aspect of the present invention, determining the set of dialect initial consonant strings and the set of dialect final vowel strings based on the basic phonetic features includes:

[0017] Phonetic analysis of all phonemes in the standard language is performed to extract the phonetic features of the standard language;

[0018] By comparing the pronunciation differences between the basic phonetic features and the phonetic features of the standard language, a pronunciation difference comparison result is obtained. Based on the pronunciation difference comparison result, the initial consonant strings of the standard language are modified to obtain a set of dialect initial consonant strings. Based on the pronunciation difference comparison result, the final vowel strings of the standard language are modified to obtain a set of dialect final vowel strings.

[0019] As an optional implementation, in the first aspect of the present invention, the step of encoding the original unit based on the target dialect pinyin encoding for any one of the original units to obtain the target unit of the original unit includes:

[0020] When the original unit is the numeric string, the current application scenario of the target dialect is determined, and the dialect numeric pronunciation codes of all single-digit numbers in the current application scenario and the dialect numeric combination pronunciation methods of number combinations are obtained from the target dialect pinyin encoding. It is then determined whether the numeric string is a single-digit number. If the numeric string is determined to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to the single-digit numbers in the current application scenario to obtain the target unit of the numeric string. If the numeric string is determined not to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to all single-digit numbers included in the numeric string in the current application scenario and the dialect numeric combination pronunciation methods of the number string combinations in the current application scenario. The dialect pronunciation of the number combination corresponding to the number combination is used to encode the number string to obtain the target unit of the number string. The dialect pronunciation encoding of each single digit is the same in the same application scenario, and the dialect pronunciation encoding of each single digit is different in different application scenarios. The dialect pronunciation of each number combination is the same in the same application scenario, and the dialect pronunciation of each number combination is different in different application scenarios. The number combination includes at least two single digits. The type of the number combination is determined by the context of the number combination in the text to be annotated. The dialect pronunciation encoding is a unique encoding of the syllable corresponding to the single digit in the target dialect in the current application scenario.

[0021] When the original unit is the Chinese character, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the target dialect pinyin code as the Chinese character pronunciation code of the Chinese character. The Chinese character is then encoded using the Chinese character pronunciation code to obtain the target unit of the Chinese character.

[0022] When the original unit is the Pinyin string, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the Pinyin encoding of the target dialect as the Pinyin pronunciation code of the Pinyin string. The Pinyin pronunciation code is used to encode the Pinyin string to obtain the target unit of the Pinyin string.

[0023] As an optional implementation, in the first aspect of the present invention, the original unit further includes punctuation characters, and before sorting all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated to obtain the dialect prosodic text corresponding to the text to be annotated, the method further includes:

[0024] Based on the type of the target dialect, determine the punctuation code corresponding to each punctuation character. The punctuation code is used to represent the pause duration of the punctuation character at the corresponding position in the speech stream. Encode the punctuation character according to the punctuation code to obtain the target unit of the punctuation character.

[0025] The punctuation mark encoding is determined in the following way:

[0026] Based on the type of the target dialect, at least one speech pause pattern is predefined, and each speech pause pattern maintains a silence mode for a different duration in the speech stream;

[0027] Define a unique speech pause pattern encoding symbol for each of the aforementioned speech pause patterns;

[0028] The speech pause pattern encoding symbol matching each of the sample punctuation characters in all sample punctuation characters of the target dialect is determined to obtain the punctuation code, and all sample punctuation characters contain the punctuation character.

[0029] As an optional implementation, in the first aspect of the present invention, before segmenting the text to be annotated in the target dialect according to a preset text segmentation method, the method further includes:

[0030] Based on the type of the target dialect, a text cleaning method is determined, and text cleaning operations are performed on the text to be labeled according to the determined text cleaning method. The text cleaning operations include deleting duplicate characters and / or deleting pre-determined illegal characters; and / or

[0031] Based on the type of the target dialect, determine the punctuation deletion method, and perform punctuation deletion operation on the text to be annotated according to the punctuation deletion method; and / or

[0032] Determine whether there are other language texts in the text to be labeled that are different from the target dialect type. If there are other language texts, convert the other language texts into texts that are the same as the target dialect type.

[0033] As an optional implementation, in the first aspect of the present invention, after sorting all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated to obtain the dialect prosodic text corresponding to the text to be annotated, the method further includes:

[0034] The dialect prosodic text is converted into a playable dialect speech file using a pre-determined dialect speech synthesis model.

[0035] The dialect speech synthesis model is obtained by inputting the dialect prosodic text into the speech synthesis model corresponding to the standard language and performing transfer learning.

[0036] A second aspect of the present invention discloses a prosodic text generation device for dialects, the device comprising:

[0037] The segmentation module is used to segment the text to be labeled in the target dialect according to a preset text segmentation method to obtain all the original units of the text to be labeled. All the original units include one or more of Chinese characters, Pinyin strings and number strings.

[0038] The determining module is used to determine, based on the type of the target dialect, a target dialect pinyin code that matches the target dialect from a plurality of pre-determined dialect pinyin codes, wherein the target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect;

[0039] An encoding module is used to encode any one of the original units based on the target dialect pinyin encoding to obtain the target unit of the original unit;

[0040] The sorting module is used to sort all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, so as to obtain the dialect prosodic text corresponding to the text to be annotated.

[0041] As an optional implementation, in a second aspect of the invention, the apparatus further includes:

[0042] The phonetic analysis module is used to perform phonetic analysis on all phonemes of any one of the multiple sample dialects and extract basic phonetic features. The multiple sample dialects include the target dialect.

[0043] The dialect syllable parsing module is used to determine the set of dialect initial consonant strings and the set of dialect final vowel strings based on the basic phonetic features. The set of dialect initial consonant strings includes all dialect initial consonant strings representing the pronunciation of the beginning part of all syllables in the sample dialect, and the set of dialect final vowel strings includes all dialect final vowel strings representing the pronunciation of the non-beginning part of all syllables in the sample dialect.

[0044] The dialect syllable parsing module is also used to determine a dialect tone character set based on the basic phonetic features. The dialect tone character set includes all dialect tone characters representing all pronunciation tones in the sample dialect.

[0045] The dialect syllable encoding module is used to determine the syllable pinyin encoding corresponding to any syllable of the sample dialect according to the dialect initial consonant string, dialect final vowel string and dialect tone character of the syllable in a predetermined dialect pinyin encoding order, and to use the syllable pinyin encoding corresponding to all syllables of the sample dialect as the dialect pinyin encoding corresponding to the sample dialect, wherein the dialect initial consonant string is a string that can be omitted.

[0046] The determining module is further configured to, after the dialect syllable encoding module determines the dialect pinyin encoding corresponding to all the sample dialects, determine the dialect pinyin encoding corresponding to all the sample dialects as a plurality of pre-determined dialect pinyin encodings.

[0047] As an optional implementation, in the second aspect of the present invention, the dialect syllable parsing module determines the dialect initial consonant string set and the dialect final vowel string set based on the basic phonetic features in the following specific way:

[0048] Phonetic analysis of all phonemes in the standard language is performed to extract the phonetic features of the standard language;

[0049] By comparing the pronunciation differences between the basic phonetic features and the phonetic features of the standard language, a pronunciation difference comparison result is obtained. Based on the pronunciation difference comparison result, the initial consonant strings of the standard language are modified to obtain a set of dialect initial consonant strings. Based on the pronunciation difference comparison result, the final vowel strings of the standard language are modified to obtain a set of dialect final vowel strings.

[0050] As an optional implementation, in the second aspect of the present invention, the encoding module encodes any one of the original units based on the target dialect pinyin encoding to obtain the target unit of the original unit in the following specific manner:

[0051] When the original unit is the numeric string, the current application scenario of the target dialect is determined, and the dialect numeric pronunciation codes of all single-digit numbers in the current application scenario and the dialect numeric combination pronunciation methods of number combinations are obtained from the target dialect pinyin encoding. It is then determined whether the numeric string is a single-digit number. If the numeric string is determined to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to the single-digit numbers in the current application scenario to obtain the target unit of the numeric string. If the numeric string is determined not to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to all single-digit numbers included in the numeric string in the current application scenario and the numeric character encoding methods of the current application scenario. The dialect pronunciation of the number combination corresponding to the number string is used to encode the number string to obtain the target unit of the number string. The dialect pronunciation encoding of each single digit is the same in the same application scenario, and the dialect pronunciation encoding of each single digit is different in different application scenarios. The dialect pronunciation of each number combination is the same in the same application scenario, and the dialect pronunciation of each number combination is different in different application scenarios. The number combination includes at least two single digits. The type of the number combination is determined by the context of the number combination in the text to be annotated. The dialect pronunciation encoding is the unique encoding of the syllable corresponding to the single digit in the target dialect in the current application scenario.

[0052] When the original unit is the Chinese character, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the target dialect pinyin code as the Chinese character pronunciation code of the Chinese character. The Chinese character is then encoded using the Chinese character pronunciation code to obtain the target unit of the Chinese character.

[0053] When the original unit is the Pinyin string, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the Pinyin encoding of the target dialect as the Pinyin pronunciation code of the Pinyin string. The Pinyin pronunciation code is used to encode the Pinyin string to obtain the target unit of the Pinyin string.

[0054] As an optional implementation, in a second aspect of the invention, the original unit further includes punctuation characters, and the device further includes:

[0055] The second encoding module is used to, before the sorting module sorts all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated to obtain the dialect prosodic text corresponding to the text to be annotated, determine the punctuation code corresponding to each punctuation character according to the type of the target dialect, wherein the punctuation code is used to represent the pause duration of the position of the punctuation character in the text to be annotated and the corresponding position in the speech stream; and encode the punctuation character according to the punctuation code to obtain the target unit of the punctuation character;

[0056] The punctuation mark encoding is determined in the following way:

[0057] Based on the type of the target dialect, at least one speech pause pattern is predefined, and each speech pause pattern maintains a silence mode for a different duration in the speech stream;

[0058] Define a unique speech pause pattern encoding symbol for each of the aforementioned speech pause patterns;

[0059] The speech pause pattern encoding symbol matching each of the sample punctuation characters in all sample punctuation characters of the target dialect is determined to obtain the punctuation code, and all sample punctuation characters contain the punctuation character.

[0060] As an optional implementation, in a second aspect of the invention, the apparatus further includes:

[0061] The text cleaning module is used to determine the text cleaning method according to the type of the target dialect before the segmentation module segments the text to be labeled in the target dialect according to the preset text segmentation method, and to perform text cleaning operation on the text to be labeled according to the text cleaning method. The text cleaning operation includes deleting duplicate characters and / or deleting predetermined illegal characters.

[0062] The punctuation deletion module is used to determine the punctuation deletion method according to the type of the target dialect before the segmentation module segments the text to be annotated in the target dialect according to the preset text segmentation method, and to perform punctuation deletion operation on the text to be annotated according to the punctuation deletion method.

[0063] The language conversion module is used to determine whether there are other language texts in the text to be labeled that are different from the target dialect type before the segmentation module segments the text to be labeled in the target dialect according to a preset text segmentation method. If there are other language texts, the other language texts are converted into texts that are the same as the target dialect type.

[0064] As an optional implementation, in a second aspect of the invention, the apparatus further includes:

[0065] The speech synthesis module is used to sort all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, and obtain the dialect prosodic text corresponding to the text to be annotated, and then use a pre-determined dialect speech synthesis model to convert the dialect prosodic text into a playable dialect speech file.

[0066] The dialect speech synthesis model is obtained by inputting the dialect prosodic text into the speech synthesis model corresponding to the standard language and performing transfer learning.

[0067] A third aspect of the present invention discloses another prosodic text generation apparatus for dialects, the apparatus comprising:

[0068] Memory containing executable program code;

[0069] A processor coupled to the memory;

[0070] The processor calls the executable program code stored in the memory to execute some or all of the steps in the method for generating prosodic text for dialects disclosed in the first aspect of the present invention.

[0071] The fourth aspect of the present invention discloses a computer-storable medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the method for generating prosodic texts in dialects disclosed in the first aspect of the present invention.

[0072] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0073] In this embodiment of the invention, the text to be annotated in the target dialect is segmented according to a preset text segmentation method to obtain all original units of the text to be annotated. All original units include one or more of Chinese characters, pinyin strings, and numeric strings. According to the type of the target dialect, a target dialect pinyin code matching the target dialect is determined from a plurality of pre-determined dialect pinyin codes. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect. For any original unit, the original unit is encoded based on the target dialect pinyin code to obtain the target unit of the original unit. According to the sorting order of the original units corresponding to each target unit in the text to be annotated, all target units are sorted to obtain the dialect prosodic text corresponding to the text to be annotated. As can be seen, this invention divides the text to be annotated into original units, determines the target dialect pinyin code according to the type of the target dialect, and uses the target dialect pinyin code to sequentially encode the text to be annotated in units of original units. By using the dialect prosodic features carried by the target dialect pinyin code, the text to be annotated is converted into a dialect prosodic text that conforms to the pronunciation characteristics of the dialect. This can capture its accurate prosodic characteristics, thereby improving the accuracy of generating dialect prosodic text, and thus helping to improve the synthesis of speech that accurately expresses the prosodic characteristics of the dialect. Attached Figure Description

[0074] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0075] Figure 1 This is a flowchart illustrating a method for generating prosodic text in dialects, as disclosed in an embodiment of the present invention.

[0076] Figure 2 This is a flowchart illustrating another method for generating prosodic text in dialects, as disclosed in an embodiment of the present invention.

[0077] Figure 3 This is a schematic diagram of the structure of a prosodic text generation device for dialects disclosed in an embodiment of the present invention;

[0078] Figure 4 This is a schematic diagram of another prosodic text generation device for dialects disclosed in an embodiment of the present invention;

[0079] Figure 5 This is a schematic diagram of another prosodic text generation device for dialects disclosed in an embodiment of the present invention. Detailed Implementation

[0080] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0082] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0083] The prosodic characteristics of dialects differ significantly from those of standard languages ​​such as Mandarin. Prosodic texts based on standard languages ​​cannot reflect the pronunciation features of dialects. This invention provides a method for generating prosodic texts applicable to dialects, capable of producing prosodic texts that reflect the pronunciation characteristics of dialects. Based on this prosodic text, further speech processing can be performed, including but not limited to generating synthesized speech with dialect pronunciation features for use in navigation, voice assistant, and other systems; applying the prosodic text to dialect recognition and conversion systems to identify the dialect type of speech and convert it to speech based on standard languages; and applying the prosodic text to voice control systems to accurately recognize dialect-based voice control commands. This invention does not limit the applications of the prosodic text. The following detailed description of specific embodiments further illustrates this invention.

[0084] Example 1

[0085] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for generating prosodic text in dialects, as disclosed in an embodiment of the present invention. Figure 1The method shown can be applied to various Chinese dialect application scenarios equipped with a prosodic text generation device for dialects, such as Northeastern Mandarin, Chongqing dialect, and Cantonese, etc., and is not limited to any particular dialect in this embodiment. Figure 1 As shown, the method for generating prosodic text for dialects may include the following operations:

[0086] 101. Segment the text to be annotated in the target dialect according to the preset text segmentation method to obtain all the original units of the text to be annotated. All original units include one or more of Chinese characters, Pinyin strings and number strings.

[0087] The text segmentation method in this embodiment can refer to splitting all Chinese characters in the text to be labeled into several Chinese character units, splitting all numbers in the text to be labeled into several number strings, and splitting all pinyin letters in the text to be labeled into several pinyin strings, with each consecutive pinyin character as the unit. For example, after splitting the text "My ID is 123, her ID is 456", the following original units are obtained in sequence: "I", "of", "ID", "is", "123", "she", "of", "ID", "is", "456".

[0088] 102. Based on the type of the target dialect, determine the target dialect pinyin code that matches the target dialect from among a number of pre-determined dialect pinyin codes. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect.

[0089] 103. For any original unit, encode the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit.

[0090] 104. Based on the sorting order of the original units corresponding to each target unit in the text to be annotated, sort all target units to obtain the dialect prosodic text corresponding to the text to be annotated.

[0091] It is evident that implementation Figure 1 The described method segments the text to be annotated into original units, determines the target dialect pinyin code based on the type of the target dialect, and uses the target dialect pinyin code to sequentially encode the text to be annotated in units of original units. By using the dialect prosodic features carried by the target dialect pinyin code, the text to be annotated is converted into dialect prosodic text that conforms to the pronunciation characteristics of the dialect. This method can capture its accurate prosodic characteristics, thereby improving the accuracy of dialect prosodic text generation and thus helping to improve the synthesis of speech that accurately expresses the prosodic characteristics of the dialect.

[0092] In an optional embodiment, the method for generating prosodic text applied to dialects may further include:

[0093] For any one of the multiple sample dialects, perform phonetic analysis on all phonemes in the sample dialect to extract basic phonetic features. The multiple sample dialects include the target dialect.

[0094] Based on basic phonetic features, a set of dialect initial consonant strings and a set of dialect final vowel strings are determined. The set of dialect initial consonant strings includes all dialect initial consonant strings representing the pronunciation of the beginning part of all syllables in the sample dialect, and the set of dialect final vowel strings includes all dialect final vowel strings representing the pronunciation of the non-beginning part of all syllables in the sample dialect.

[0095] Based on basic phonetic features, a set of dialect tone characters is determined. The set of dialect tone characters includes all dialect tone characters that represent all pronunciation tones in the sample dialect.

[0096] For any syllable of the sample dialect, the syllable pinyin code is determined according to the dialect initial consonant string, dialect final vowel string and dialect tone character corresponding to the syllable in a predetermined dialect pinyin encoding order. The syllable pinyin codes corresponding to all syllables of the sample dialect are used as the dialect pinyin code corresponding to the sample dialect. Among them, the dialect initial consonant string is a string that can be omitted.

[0097] Determine the dialect pinyin codes corresponding to all sample dialects, which are multiple pre-determined dialect pinyin codes.

[0098] Among them, a phoneme is the smallest unit of pronunciation in a language, the smallest unit that makes up a syllable or the smallest speech segment, and is the smallest linear speech unit divided from the perspective of voice quality; a syllable is a unit of sound continuously emitted by the airflow through the vocal organs during speaking, and is also the smallest speech unit in which a single vowel phoneme and a consonant phoneme are combined to pronounce. A single vowel phoneme can also form a syllable by itself; basic speech features refer to the acoustic and physiological characteristics formed by phonemes during the pronunciation process, specifically including elements such as pitch (tone), length (duration), voice quality (including the opening and closing, front and back positions, tongue position of vowels, etc., and the voicelessness / voicedness, place of articulation, and manner of articulation of consonants), etc. They jointly define the unique pronunciation attributes of a syllable; the pronunciation at the beginning of a syllable refers to the initial consonant or consonant combination in a syllable; the pronunciation of the non-beginning part of a syllable refers to the vowel and subsequent consonants in a syllable that are located after the pronunciation at the beginning of the syllable; tone refers to the pitch change of a syllable or phoneme; phonetic analysis is a systematic study of the acoustic and articulatory characteristics of speech, including the production method of phonemes, audio characteristics, etc. Taking the Chinese character "马" as an example, the pronunciation corresponding to the pinyin mǎ in Mandarin is a syllable. The pronunciation corresponding to the pinyin m in Mandarin is the pronunciation at the beginning of this syllable, and the pronunciation corresponding to the pinyin a in Mandarin is the pronunciation of the non-beginning part of this syllable. The pronunciations corresponding to the pinyin m and the pinyin a in Mandarin are both separate phonemes. The tone of the Chinese character "马" is a falling-rising tone, that is, it starts from a medium pitch, then decreases and then increases.

[0099] When the dialect pinyin codes corresponding to all sample dialects are determined, an operation of determining the target dialect pinyin code that matches the target dialect from multiple pre-determined dialect pinyin codes according to the type of the target dialect is triggered and executed.

[0100] It can be seen that the optional embodiment determines the set of dialect initial strings and the set of dialect final strings according to the basic speech features, and then generates the dialect pinyin codes corresponding to various sample dialects, which can generate accurate target dialect pinyin codes, accurately represent the prosodic features of the dialect, and is beneficial to improving the generation accuracy of the prosodic text corresponding to the text to be annotated, and further helps the inheritance and promotion of the dialect.

[0101] In this optional embodiment, further optionally, determining the set of dialect initial strings and the set of dialect final strings according to the basic speech features includes:

[0102] Performing phonetic analysis on all phonemes in the standard language and extracting the speech features of the standard language;

[0103] Compare the pronunciation differences between the basic speech features and the standard language speech features to obtain the pronunciation difference comparison results. According to the pronunciation difference comparison results, modify the initial consonant string of the standard language to obtain a set of dialect initial consonant strings; according to the pronunciation difference comparison results, modify the final vowel string of the standard language to obtain a set of dialect final vowel strings.

[0104] Among them, the standard language is a language form widely recognized and used in a social, political or cultural group, and is usually standardized and promoted by official institutions. In this embodiment, Mandarin is the standard language.

[0105] Taking Chongqing dialect as an example in this embodiment, it specifically explains how to determine the target dialect pinyin encoding of Chongqing dialect. Compare the pronunciation differences between Chongqing dialect and Mandarin, and delete the phonemes that do not exist in Chongqing dialect, such as the retroflex initial consonants zh in Mandarin, the retroflex initial consonants ch in Mandarin, the retroflex initial consonants sh in Mandarin, the final vowel eng in Mandarin, the final vowel ing in Mandarin, etc.; add the phonemes unique to Chongqing dialect, such as the pronunciation of the non-beginning part of the syllable corresponding to "han" in the phrase "lao hanr" in Chongqing dialect. Modify the initial consonants of Mandarin to obtain a set of initial consonant strings of Chongqing dialect:

[0106] aa, b, c, d, ee, f, g, h, ii, j, k, l, m, oo, p, q, r, s, t, uu, vv, x, z.

[0107] Among them, the initial consonants aa, ee, ii, oo, uu, vv are the pseudo-initial consonants corresponding to the vowels a, e, i, o, u, v respectively. The pseudo-initial consonant means that no sound is emitted in the speech stream. When the initial consonant in a syllable is missing, the pseudo-initial consonant corresponding to the vowel included in the syllable is used as the initial consonant of the syllable. For example, the syllable of the Chinese character "e" has no initial consonant, and the initial consonant in its dialect pinyin encoding is set as the pseudo-initial consonant ee corresponding to the vowel e. By introducing pseudo-initial consonants in the set of initial consonant strings, it is ensured that all syllables in the dialect pinyin encoding include initial consonants, which can simplify the encoding method of the original units, prevent errors caused by complex encoding methods, and improve the accuracy of prosody text generation.

[0108] Modify the final vowels of Mandarin to obtain a set of final vowel strings of Chongqing dialect:

[0109] a, ai, an, ang, ao, e, ei, en, er, i, iu, iy, o, ong, ou, u, ueng, ui, un, v, van, ve, vn, ng, uong, ar, uar, ier, iar, uer.

[0110] Among them, the pronunciation of the final ar is the non-initial part of the syllable corresponding to "Han" in "laohan'er" in Chongqing dialect; the pronunciation of the final iy is the non-initial part of the syllable corresponding to "chi" in Chongqing dialect; the pronunciation of the final ueng is the non-initial part of the syllable corresponding to "yue" in Chongqing dialect; the pronunciation of the final van is the non-initial part of the syllable corresponding to "wan" in Chongqing dialect; the pronunciation of the final ng is the non-initial part of the syllable corresponding to "ying" in Chongqing dialect; and the pronunciation of the final uong is the non-initial part of the syllable corresponding to "yong" in Chongqing dialect.

[0111] The tone types in Chongqing dialect are the same as those in Standard Mandarin. However, some Chinese characters have different tones in Chongqing dialect and Standard Mandarin. Therefore, when encoding the original unit corresponding to a character, the encoding should be based on the tone of that character in Chongqing dialect.

[0112] As can be seen, this further optional embodiment performs phonetic analysis on all phonemes of the standard language, extracts the phonetic features of the standard language, compares them with the basic phonetic features of the target dialect, and modifies the initial consonant strings and final vowel strings of the standard language according to the comparison results, thereby obtaining an accurate set of dialect initial consonant strings and a set of dialect final vowel strings. This can improve the accuracy of determining the pinyin encoding of the target dialect, which in turn helps to improve the accuracy of generating dialect prosodic text, and can also reduce the development time of prosodic text generation devices applied to dialects, saving development resources.

[0113] In another optional embodiment, for any original unit, the original unit is encoded based on the target dialect pinyin encoding to obtain the target unit of the original unit, including:

[0114] When the original unit is a numeric string, determine the current application scenario of the target dialect, and obtain from the target dialect pinyin encoding all the dialect numeric pronunciation encodings of single-digit numbers and the dialect numeric combination pronunciation methods of numeric combinations in the determined current application scenario, and determine whether the numeric string is a single-digit number; when it is determined that the numeric string is a single-digit number, encode the numeric string according to the dialect numeric pronunciation encoding corresponding to the single-digit number in the current application scenario to obtain the target unit of the numeric string; when it is determined that the numeric string is not a single-digit number, encode the numeric string according to the dialect numeric pronunciation encodings corresponding to all the single-digit numbers included in the numeric string in the current application scenario and the dialect numeric combination pronunciation method corresponding to the numeric combination formed by the numeric string in the current application scenario to obtain the target unit of the numeric string. The dialect numeric pronunciation encoding of each single-digit number is the same in the same application scenario of the dialect, and the dialect numeric pronunciation encodings of each single-digit number are different in different application scenarios. The dialect numeric combination pronunciation method of each numeric combination is the same in the same application scenario, and the dialect numeric combination pronunciation methods of each numeric combination are different in different application scenarios. A numeric combination includes at least two single-digit numbers, and the type of the numeric combination is determined by the context of the numeric combination in the text to be annotated. The dialect numeric pronunciation encoding is the unique encoding of the syllable corresponding to the single-digit number in the target dialect in the current application scenario;

[0115] When the original unit is a Chinese character, determine from the target dialect pinyin encoding the unique encoding that can describe the syllable corresponding to the Chinese character in the target dialect as the Chinese pronunciation encoding of the Chinese character, and use the Chinese pronunciation encoding to encode the Chinese character to obtain the target unit of the Chinese character;

[0116] When the original unit is a pinyin string, determine from the target dialect pinyin encoding the unique encoding that can describe the syllable corresponding to the Chinese character corresponding to the pinyin string in the target dialect as the pinyin pronunciation encoding of the pinyin string, and use the pinyin pronunciation encoding to encode the pinyin string to obtain the target unit of the pinyin string.

[0117] In this optional embodiment, single-digit numbers refer to the numbers 0-9, and the pronunciation methods of single-digit numbers may be different in different application scenarios. For example, the number 1 is often pronounced the same as "yao" in Mandarin in a phone number, the number 0 is often pronounced the same as "dong" in Mandarin in a phone number, and the number 2 is often pronounced as "liang" before a measure word. Moreover, the pronunciation methods of different types of numeric combinations may also be different. For example, the numeric combination 1200 in the phone number type is pronounced as "one two zero zero", and the numeric combination 1200 before a measure word is pronounced as "one thousand two hundred" in some application scenarios and "one thousand two" in some application scenarios.

[0118] As can be seen, this optional embodiment improves the encoding accuracy of number strings in different application scenarios by determining the pronunciation encoding of single digits in different application scenarios, the pronunciation of different digit combination types in different application scenarios, and encoding the number string according to the current application scenario and the number of digits in the number string. Encoding the Chinese character based on the precise target dialect pinyin encoding and the syllable corresponding to the Chinese character in the target dialect accurately reflects the pronunciation of the Chinese character in the target dialect, thus improving the encoding accuracy of the Chinese character. Furthermore, encoding the pinyin string based on the precise target dialect pinyin encoding and the syllable corresponding to the Chinese character in the target dialect accurately reflects the pronunciation of the Chinese character corresponding to the pinyin string in the target dialect, thus improving the encoding accuracy of the pinyin string, and consequently improving the accuracy of dialect prosodic text generation.

[0119] In another optional embodiment, the original unit further includes punctuation characters. Before sorting all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, and obtaining the dialect prosodic text corresponding to the text to be annotated, the method for generating prosodic text applied to dialects further includes:

[0120] Based on the type of the target dialect, determine the punctuation code corresponding to each punctuation character. The punctuation code is used to represent the pause duration at the corresponding position in the speech stream of the punctuation character in the text to be annotated. Encode the punctuation characters according to the punctuation code to obtain the target units of the punctuation characters. And trigger the execution of sorting all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated to obtain the dialect prosodic text corresponding to the text to be annotated.

[0121] The punctuation mark encoding is determined in the following way:

[0122] Based on the type of the target dialect, at least one speech pause pattern is predefined, and each speech pause pattern maintains a silence mode for a different duration in the speech stream;

[0123] Define a unique speech pause pattern encoding symbol for each speech pause pattern;

[0124] The speech pause pattern encoding symbol is determined for each sample punctuation character in all sample punctuation characters of the target dialect, thus obtaining the punctuation code. All sample punctuation characters contain punctuation characters.

[0125] As can be seen, in this optional embodiment, by matching punctuation marks with corresponding speech pause patterns to represent different pause durations in the speech stream, the prosodic text can represent the speech rhythm and pauses that simulate human pronunciation, making the synthesized speech more in line with the prosody of the target dialect, thereby improving the accuracy of dialect prosodic text generation.

[0126] In yet another optional embodiment, before segmenting the text to be annotated in the target dialect according to a preset text segmentation method, the method for generating prosodic text applied to the dialect further includes:

[0127] Based on the type of the target dialect, determine the text cleaning method, perform text cleaning operations on the text to be labeled according to the text cleaning method, and trigger the execution of the operation of segmenting the text to be labeled in the target dialect according to the preset text segmentation method. The text cleaning operation includes deleting duplicate characters and / or deleting predetermined illegal characters.

[0128] Among them, illegal characters refer to characters that cannot be recognized as Chinese characters, pinyin strings, or number strings when the text to be labeled in the target dialect is segmented according to the preset text segmentation method.

[0129] As can be seen, this optional embodiment reduces noise in prosodic text by deleting duplicate and / or illegal characters, thereby improving the accuracy and efficiency of generating dialect prosodic text.

[0130] In yet another optional embodiment, before segmenting the text to be annotated in the target dialect according to a preset text segmentation method, the method for generating prosodic text applied to the dialect further includes:

[0131] Based on the type of the target dialect, determine the punctuation deletion method, perform punctuation deletion operation on the text to be annotated according to the punctuation deletion method, and trigger the execution of the operation of segmenting the text to be annotated in the target dialect according to the preset text segmentation method.

[0132] Among them, determining the punctuation deletion method based on the type of the target dialect can refer to determining the type of punctuation marks that do not carry pause duration as the type of punctuation marks to be deleted based on the type of the target dialect, and deleting all punctuation marks of the same type as the punctuation marks to be deleted one by one or in batches in the text to be annotated.

[0133] As can be seen, this optional embodiment eliminates the interference caused by irrelevant punctuation marks by deleting punctuation marks that do not carry the duration of pauses, thereby improving the accuracy of dialect prosodic text generation, as well as improving the generation efficiency of dialect prosodic text and simplifying the structure of prosodic text.

[0134] In yet another optional embodiment, before segmenting the text to be annotated in the target dialect according to a preset text segmentation method, the method for generating prosodic text applied to the dialect further includes:

[0135] Determine whether there is other language text in the text to be annotated that is different from the target dialect type. If there is other language text, convert the other language text into text that is the same as the target dialect type, and trigger the operation of segmenting the text to be annotated in the target dialect according to the preset text segmentation method.

[0136] As can be seen, this optional embodiment can ensure that all texts in the text to be annotated are of the same language type as the target dialect by converting other language texts into texts of the same type as the target dialect, thereby improving the accuracy of dialect prosodic text generation and thus improving the accuracy of dialect prosodic expression during speech synthesis.

[0137] In another optional embodiment, after encoding the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit, the method for generating prosodic text applied to dialects further includes:

[0138] For each target unit that is pre-determined to require prolonged pronunciation, a corresponding prolonged pronunciation code is added. The prolonged pronunciation code for each target unit represents the pronunciation of the non-starting part of the syllable of the target dialect corresponding to that target unit in the speech stream that needs to be prolonged.

[0139] Specifically, adding a prolonging sound code to the target unit can be done by placing the prolonging sound code in the position after the target unit or in the position before the target unit; this embodiment of the invention does not impose any limitation. The target unit that needs to prolong the pronunciation can be any one of the following: a target unit corresponding to a Chinese character, a target unit corresponding to a Pinyin string, or a target unit corresponding to a number string; this embodiment of the invention does not impose any limitation.

[0140] As can be seen, this optional embodiment, by incorporating elongation encoding, enables prosodic text to represent the natural speech features of extending the non-initial parts of certain syllables in specific contexts when humans speak, thereby improving the accuracy of dialect prosodic text generation and further enhancing the accuracy and naturalness of speech synthesis.

[0141] In another optional embodiment, after encoding the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit, the method for generating prosodic text applied to dialects further includes:

[0142] Add a corresponding phonetic repetition code to each target unit that needs to be re-read in a pre-determined manner. The phonetic repetition code for each target unit represents whether the volume of the syllable of the target dialect corresponding to that target unit needs to be increased or decreased in the speech stream.

[0143] Specifically, adding phonetic re-encoding to the target unit can be done by placing the phonetic re-encoding in the position following the target unit or in the position preceding the target unit; this embodiment of the invention does not impose any limitation. The target unit to which phonetic re-encoding needs to be added can be any one of the following: a target unit corresponding to a Chinese character, a target unit corresponding to a Pinyin string, or a target unit corresponding to a number string; this embodiment of the invention does not impose any limitation.

[0144] After the corresponding actions are performed on all target units, the operation of sorting all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated is triggered to obtain the dialect prosodic text corresponding to the text to be annotated. Specifically, the corresponding actions refer to adding elongated sound encoding and / or sound repetition encoding to the target units.

[0145] As can be seen, this optional embodiment, by incorporating phonetic re-encoding, enables prosodic text to represent the emphasis and weakening in natural language, thereby improving the accuracy of dialect prosodic text generation and further enhancing the emotional expressiveness of the synthesized speech. On the other hand, clearly indicating which syllables in the speech stream need to be raised or lowered in volume helps listeners better capture key information and improves the comprehension of the speech content.

[0146] In yet another optional embodiment, before segmenting the text to be annotated in the target dialect according to a preset text segmentation method, the method for generating prosodic text applied to the dialect further includes:

[0147] Replace the pre-determined text that needs dialect synonym replacement with dialect text that has the same semantics, and trigger the execution of the step of segmenting the target dialect text to be labeled according to the preset text segmentation method.

[0148] In this embodiment, dialect synonyms refer to the common substitutes for words in the standard language in the dialect. Taking Chongqing dialect as an example, "thinking of" is usually replaced by Chongqing dialect with the pronunciation mié. The Chongqing dialect with the pronunciation mié is the common substitute for "thinking of" in the standard language in Chongqing dialect.

[0149] As can be seen, this optional embodiment, by replacing the text to be replaced with dialect synonyms, better simulates the real pronunciation habits of dialect users, improves the accuracy of dialect prosodic text generation, and thus helps to improve the synthesis of speech that accurately expresses the prosodic characteristics of dialect; moreover, the generated prosodic text is more in line with the expression habits of dialect, improves the naturalness and local flavor of speech synthesis output, makes the synthesized speech more friendly and realistic, and thus enhances the auditory experience of dialect users.

[0150] Example 2

[0151] Please see Figure 2 , Figure 2 This is a flowchart illustrating another method for generating prosodic text in dialects, as disclosed in an embodiment of the present invention. Figure 2 The method shown can be applied to various Chinese dialect application scenarios of the dialect prosodic text generation device, including various Chinese dialects such as Northeastern Mandarin, Chongqing dialect, and Cantonese, etc., and the embodiments of this invention are not limited thereto. Figure 2 As shown, the method for generating prosodic text for dialects may include the following operations:

[0152] 201. Segment the text to be annotated in the target dialect according to the preset text segmentation method to obtain all the original units of the text to be annotated. All original units include one or more of Chinese characters, Pinyin strings and number strings.

[0153] 202. Based on the type of the target dialect, determine the target dialect pinyin code that matches the target dialect from among a number of pre-determined dialect pinyin codes. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect.

[0154] 203. For any original unit, encode the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit.

[0155] 204. Based on the sorting order of the original units corresponding to each target unit in the text to be annotated, sort all target units to obtain the dialect prosodic text corresponding to the text to be annotated.

[0156] 205. Use a pre-determined dialect speech synthesis model to convert dialect prosodic text into a playable dialect speech file;

[0157] Dialect speech synthesis models are obtained by inputting dialect prosodic text into the speech synthesis model corresponding to the standard language and then performing transfer learning.

[0158] In this embodiment of the invention, the description of steps 201-204 is the same as that of steps 101-104 in Embodiment 1, and will not be repeated here.

[0159] As can be seen, the method described in Example 2 segments the text to be annotated into original units, determines the target dialect pinyin encoding based on the type of the target dialect, and sequentially encodes the text to be annotated using the target dialect pinyin encoding on a unit basis. By using the dialect prosodic features carried by the target dialect pinyin encoding, the text to be annotated is converted into dialect prosodic text that conforms to the characteristics of dialect pronunciation. This captures its accurate prosodic characteristics, thereby improving the accuracy of dialect prosodic text generation and thus contributing to the improvement of synthesized speech that accurately expresses the prosodic features of the dialect. Furthermore, the method described in Example 2 utilizes existing standard language speech synthesis models as a foundation, adapting to the target dialect through transfer learning, reducing the training time of the target dialect speech synthesis model and improving the efficiency of dialect speech synthesis. On the other hand, transfer learning allows the model to inherit the generalization ability of the standard language speech synthesis model, including the understanding of basic speech features, enabling the target dialect speech synthesis model to adapt to new dialect data more quickly, further improving the efficiency of dialect speech synthesis.

[0160] Example 3

[0161] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a prosodic text generation device for dialects disclosed in an embodiment of the present invention. Figure 3 The device shown can be applied to various Chinese dialects, such as Northeastern Mandarin, Chongqing Mandarin, and Cantonese. This invention does not limit the application of such dialects.

[0162] like Figure 3 The apparatus shown includes:

[0163] The segmentation module 301 is used to segment the text to be labeled in the target dialect according to a preset text segmentation method to obtain all the original units of the text to be labeled. All the original units include one or more of Chinese characters, Pinyin strings and number strings.

[0164] The determining module 302 is used to determine the target dialect pinyin code that matches the target dialect from a plurality of pre-determined dialect pinyin codes according to the type of the target dialect. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect.

[0165] Encoding module 303 is used to encode any original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit;

[0166] The sorting module 304 is used to sort all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, so as to obtain the dialect prosodic text corresponding to the text to be annotated.

[0167] visible, Figure 3 The described device segments the text to be annotated into raw units, determines the target dialect pinyin code according to the type of the target dialect, and uses the target dialect pinyin code to sequentially encode the text to be annotated in units of raw units. By using the dialect prosodic features carried by the target dialect pinyin code, the text to be annotated is converted into dialect prosodic text that conforms to the pronunciation characteristics of the dialect. It can capture its accurate prosodic characteristics, thereby improving the accuracy of dialect prosodic text generation, which in turn helps to improve the synthesis of speech that accurately expresses the prosodic characteristics of the dialect.

[0168] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0169] The phonetic analysis module 305 is used to perform phonetic analysis on all phonemes of any sample dialect from multiple sample dialects, and extract basic phonetic features. The multiple sample dialects include the target dialect.

[0170] The dialect syllable parsing module 306 is used to determine the set of dialect initial consonant strings and the set of dialect final vowel strings based on basic phonetic features. The set of dialect initial consonant strings includes all dialect initial consonant strings representing the pronunciation of the beginning part of all syllables in the sample dialect, and the set of dialect final vowel strings includes all dialect final vowel strings representing the pronunciation of the non-beginning part of all syllables in the sample dialect.

[0171] The dialect syllable parsing module 306 is also used to determine the dialect tone character set based on basic phonetic features. The dialect tone character set includes all dialect tone characters representing all pronunciation tones in the sample dialect.

[0172] The dialect syllable encoding module 307 is used to determine the syllable pinyin encoding corresponding to any syllable of the sample dialect according to the dialect initial consonant string, dialect final vowel string and dialect tone character of the syllable in a predetermined dialect pinyin encoding order. The syllable pinyin encoding corresponding to all syllables of the sample dialect is used as the dialect pinyin encoding corresponding to the sample dialect. The dialect initial consonant string is a string that can be omitted.

[0173] The determination module 302 is also used to determine the dialect pinyin codes corresponding to all sample dialects after the dialect syllable encoding module 307 has determined the dialect pinyin codes corresponding to all sample dialects, which are multiple dialect pinyin codes that have been predetermined.

[0174] In this embodiment, a phoneme is the smallest unit of articulation in a language, the smallest unit constituting a syllable or the smallest speech segment, and is the smallest linear unit of speech defined from the perspective of sound quality. A syllable is a unit of sound continuously emitted by airflow through the vocal organs during speech; it is also the smallest unit of speech in a language where a single vowel phoneme and a combination of consonant phonemes are pronounced. A single vowel phoneme can also form a syllable on its own. Basic speech features refer to the acoustic and physiological characteristics formed by phonemes during articulation, specifically including pitch (tone), duration, and sound quality (including the opening and closing of vowels, their front and back positions, tongue position, etc., and the voicing, place of articulation, and manner of articulation of consonants). These elements collectively define the unique articulation attributes of a syllable. Syllable beginning part articulation refers to the initial consonant or consonant combination in a syllable. Syllable non-beginning part articulation refers to the vowel and subsequent consonant in a syllable that follow the beginning part of the syllable articulation. Tone refers to the pitch variation of a syllable or phoneme. Phonetic analysis is a systematic study of the acoustic and articulation features of speech, including the way phonemes are produced and their audio characteristics.

[0175] Once the dialect syllable encoding module 307 determines the dialect pinyin encoding corresponding to all sample dialects, it triggers the determination module 302 to perform the operation of determining the target dialect pinyin encoding that matches the target dialect from among the multiple pre-determined dialect pinyin encodings, based on the type of the target dialect.

[0176] As can be seen, this optional embodiment determines the set of dialect initials and the set of dialect finals based on basic phonetic features, and then generates dialect pinyin codes corresponding to various sample dialects. It can generate accurate target dialect pinyin codes, accurately represent the prosodic features of the dialect, and help improve the accuracy of generating prosodic text corresponding to the text to be annotated, thereby contributing to the inheritance and promotion of dialects.

[0177] In this optional embodiment, further optionally, the dialect syllable parsing module 306 determines the dialect initial consonant string set and the dialect final vowel string set based on basic phonetic features in the following specific way:

[0178] Phonetic analysis of all phonemes in the standard language is performed to extract the phonetic features of the standard language;

[0179] By comparing the pronunciation differences between basic phonetic features and standard language phonetic features, a pronunciation difference comparison result is obtained. Based on the pronunciation difference comparison result, the initial consonant strings of the standard language are modified to obtain a set of dialect initial consonant strings. Based on the pronunciation difference comparison result, the final vowel strings of the standard language are modified to obtain a set of dialect final vowel strings.

[0180] In this context, a standard language is a widely recognized and used form of language within a social, political, or cultural group, typically regulated and promoted by official institutions. In this embodiment, Mandarin Chinese is the standard language.

[0181] As can be seen, this further optional embodiment performs phonetic analysis on all phonemes of the standard language, extracts the phonetic features of the standard language, compares them with the basic phonetic features of the target dialect, and modifies the initial consonant strings and final vowel strings of the standard language according to the comparison results, thereby obtaining an accurate set of dialect initial consonant strings and a set of dialect final vowel strings. This can improve the accuracy of determining the pinyin encoding of the target dialect, which in turn helps to improve the accuracy of generating dialect prosodic text, and can also reduce the development time of prosodic text generation devices applied to dialects, saving development resources.

[0182] In another optional embodiment, the encoding module 303 encodes any original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit in the following specific way:

[0183] When the original unit is a numeric string, the current application scenario of the target dialect is determined. From the target dialect's pinyin encoding, the dialect pronunciation codes for all single-digit numbers in the current application scenario and the dialect pronunciation methods for number combinations are obtained. It is then determined whether the numeric string is a single digit. If the numeric string is determined to be a single digit, it is encoded according to the dialect pronunciation codes for the single-digit numbers in the current application scenario to obtain the target unit of the numeric string. If the numeric string is determined to be not a single digit, it is encoded according to the dialect pronunciation codes for all single-digit numbers included in the numeric string in the current application scenario and the dialect pronunciation codes for the numeric string in the current application scenario. The dialect pronunciation of the number combination is used to encode the number string to obtain the target unit of the number string. The dialect pronunciation code of each single digit is the same in the same application scenario, and the dialect pronunciation code of each single digit is different in different application scenarios. The dialect pronunciation of each number combination is the same in the same application scenario, and the dialect pronunciation of each number combination is different in different application scenarios. The number combination includes at least two single digits. The type of number combination is determined by the context of the number combination in the text to be annotated. The dialect pronunciation code is the unique code of the syllable corresponding to the single digit in the target dialect in the current application scenario.

[0184] When the original unit is a Chinese character, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the target dialect pinyin code as the Chinese character pronunciation code. The Chinese character is then encoded using the Chinese character pronunciation code to obtain the target unit of the Chinese character.

[0185] When the original unit is a Pinyin string, a unique code that can describe the syllable of the Chinese character corresponding to the Pinyin string in the target dialect is determined from the Pinyin encoding of the target dialect. This unique code is then used to encode the Pinyin string to obtain the target unit of the Pinyin string.

[0186] In this optional embodiment, a single-digit number refers to the number 0-9. The pronunciation of a single-digit number may vary in different application scenarios, and the pronunciation of different types of number combinations may also vary.

[0187] As can be seen, this optional embodiment improves the encoding accuracy of number strings in different application scenarios by determining the pronunciation encoding of single digits in different application scenarios, the pronunciation of different digit combination types in different application scenarios, and encoding the number string according to the current application scenario and the number of digits in the number string. Encoding the Chinese character based on the precise target dialect pinyin encoding and the syllable corresponding to the Chinese character in the target dialect accurately reflects the pronunciation of the Chinese character in the target dialect, thus improving the encoding accuracy of the Chinese character. Furthermore, encoding the pinyin string based on the precise target dialect pinyin encoding and the syllable corresponding to the Chinese character in the target dialect accurately reflects the pronunciation of the Chinese character corresponding to the pinyin string in the target dialect, thus improving the encoding accuracy of the pinyin string, and consequently improving the accuracy of dialect prosodic text generation.

[0188] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0189] The text cleaning module 308 is used to determine the text cleaning method according to the type of the target dialect before the segmentation module 301 segments the text to be labeled in the target dialect according to the preset text segmentation method, and to perform text cleaning operation on the text to be labeled according to the text cleaning method. The text cleaning operation includes deleting duplicate characters and / or deleting predetermined illegal characters, and triggering the segmentation module 301 to perform the operation of segmenting the text to be labeled in the target dialect according to the preset text segmentation method.

[0190] The punctuation deletion module 309 is used to determine the punctuation deletion method according to the type of the target dialect before the segmentation module 301 segments the text to be annotated in the target dialect according to the preset text segmentation method, perform punctuation deletion operation on the text to be annotated according to the punctuation deletion method, and trigger the segmentation module 301 to perform the operation of segmenting the text to be annotated in the target dialect according to the preset text segmentation method.

[0191] The language conversion module 310 is used to determine whether there are other language texts of different types than the target dialect in the text to be labeled before the segmentation module 301 segments the text to be labeled in the target dialect according to the preset text segmentation method. If there are other language texts, the other language texts are converted into texts of the same type as the target dialect, and the segmentation module 301 is triggered to perform the operation of segmenting the text to be labeled in the target dialect according to the preset text segmentation method.

[0192] As can be seen, in this optional embodiment, by deleting duplicate characters and / or illegal characters, noise in the prosodic text is reduced, improving the accuracy and efficiency of dialect prosodic text generation; by deleting punctuation marks that do not carry pause duration, interference from irrelevant punctuation marks is eliminated, improving the accuracy and efficiency of dialect prosodic text generation and simplifying the structure of the prosodic text; by converting other language texts into text of the same type as the target dialect, it is ensured that the language type of all texts in the text to be annotated is the same as the target dialect, improving the accuracy of dialect prosodic text generation, and thus improving the accuracy of dialect prosodic expression during speech synthesis.

[0193] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0194] The elongated sound encoding module 311 is used to add a corresponding elongated sound encoding to each target unit that needs to be elongated after the encoding module 303 encodes the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit. The elongated sound encoding corresponding to each target unit represents that the non-starting part of the pronunciation of the syllable of the target dialect corresponding to the target unit needs to be elongated in the speech stream.

[0195] Specifically, adding a prolonging sound code to the target unit can be done by placing the prolonging sound code in the position after the target unit or in the position before the target unit; this embodiment of the invention does not impose any limitation. The target unit that needs to prolong the pronunciation can be any one of the following: a target unit corresponding to a Chinese character, a target unit corresponding to a Pinyin string, or a target unit corresponding to a number string; this embodiment of the invention does not impose any limitation.

[0196] After the elongated sound encoding module 311 performs the action of adding elongated sound encoding to all target units, it triggers the sorting module 304 to perform the operation of sorting all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, so as to obtain the dialect prosodic text corresponding to the text to be annotated.

[0197] As can be seen, this optional embodiment, by incorporating elongation encoding, enables prosodic text to represent the natural speech features of extending the non-initial parts of certain syllables in specific contexts when humans speak, thereby improving the accuracy of dialect prosodic text generation and further enhancing the accuracy and naturalness of speech synthesis.

[0198] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0199] The speech repetition encoding module 312 is used to add corresponding speech repetition codes to each target unit that needs to be re-read in a predetermined manner after the encoding module 303 encodes the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit. The speech repetition code corresponding to each target unit represents whether the volume of the syllable of the target dialect corresponding to the target unit needs to be increased or decreased in the speech stream.

[0200] Specifically, adding phonetic re-encoding to the target unit can be done by placing the phonetic re-encoding in the position following the target unit or in the position preceding the target unit; this embodiment of the invention does not impose any limitation. The target unit to which phonetic re-encoding needs to be added can be any one of the following: a target unit corresponding to a Chinese character, a target unit corresponding to a Pinyin string, or a target unit corresponding to a number string; this embodiment of the invention does not impose any limitation.

[0201] After the phonetic re-encoding module 312 performs the phonetic re-encoding action on all target units, it triggers the sorting module 304 to sort all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, so as to obtain the dialect prosodic text corresponding to the text to be annotated.

[0202] As can be seen, this optional embodiment, by incorporating phonetic re-encoding, enables prosodic text to represent the emphasis and weakening in natural language, thereby improving the accuracy of dialect prosodic text generation and further enhancing the emotional expressiveness of the synthesized speech. On the other hand, clearly indicating which syllables in the speech stream need to be raised or lowered in volume helps listeners better capture key information and improves the comprehension of the speech content.

[0203] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0204] The replacement module 313 is used to replace the text that needs to be replaced with dialect text with the same semantic meaning before the segmentation module 301 segments the text to be labeled in the target dialect according to the preset text segmentation method, and to trigger the segmentation module 301 to perform the step of segmenting the text to be labeled in the target dialect according to the preset text segmentation method.

[0205] In this embodiment, dialect synonyms refer to commonly used alternatives for words in the standard language in the dialect.

[0206] As can be seen, this optional embodiment, by replacing the text to be replaced with dialect synonyms, better simulates the real pronunciation habits of dialect users, improves the accuracy of dialect prosodic text generation, and thus helps to improve the synthesis of speech that accurately expresses the prosodic characteristics of dialect; moreover, the generated prosodic text is more in line with the expression habits of dialect, improves the naturalness and local flavor of speech synthesis output, makes the synthesized speech more friendly and realistic, and thus enhances the auditory experience of dialect users.

[0207] In yet another alternative embodiment, such as Figure 4 As shown, the original unit also includes punctuation characters, and the prosodic text generation device applied to dialects also includes:

[0208] The second encoding module 314 is used to determine the punctuation code corresponding to each punctuation character according to the type of the target dialect before the sorting module 304 sorts all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, and obtains the dialect prosodic text corresponding to the text to be annotated. The punctuation code is used to represent the pause duration of the position of the punctuation character in the text to be annotated and the corresponding position in the speech stream. The punctuation characters are encoded according to the punctuation code to obtain the target unit of the punctuation character. The second encoding module 314 is used to trigger the sorting module 304 to sort all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, and obtain the dialect prosodic text corresponding to the text to be annotated.

[0209] The punctuation mark encoding is determined in the following way:

[0210] Based on the type of the target dialect, at least one speech pause pattern is predefined, and each speech pause pattern maintains a silence mode for a different duration in the speech stream;

[0211] Define a unique speech pause pattern encoding symbol for each speech pause pattern;

[0212] The speech pause pattern encoding symbol is determined for each sample punctuation character in all sample punctuation characters of the target dialect, thus obtaining the punctuation code. All sample punctuation characters contain punctuation characters.

[0213] As can be seen, in this optional embodiment, by matching punctuation marks with corresponding speech pause patterns to represent different pause durations in the speech stream, the prosodic text can represent the speech rhythm and pauses that simulate human pronunciation, making the synthesized speech more in line with the prosody of the target dialect, thereby improving the accuracy of dialect prosodic text generation.

[0214] In yet another alternative embodiment, such as Figure 4 As shown, the prosodic text generation device applied to dialects also includes:

[0215] The speech synthesis module 315 is used to sort all target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated in the sorting module 304, and after obtaining the dialect prosodic text corresponding to the text to be annotated, use a pre-determined dialect speech synthesis model to convert the dialect prosodic text into a playable dialect speech file.

[0216] Dialect speech synthesis models are obtained by inputting dialect prosodic text into the speech synthesis model corresponding to the standard language and then performing transfer learning.

[0217] As can be seen, this optional embodiment converts dialect prosodic text that conforms to the pronunciation characteristics of the dialect into a playable dialect speech file, capturing the accurate prosodic features of the dialect, thereby improving the accuracy of dialect prosodic text generation and thus contributing to the improvement of synthesized speech that accurately expresses the prosodic features of the dialect. Furthermore, this optional embodiment utilizes existing standard language speech synthesis models as a foundation, adapting to the target dialect through transfer learning, reducing the training time of the target dialect speech synthesis model and improving the efficiency of dialect speech synthesis. On the other hand, transfer learning allows the model to inherit the generalization ability of the standard language speech synthesis model, including the understanding of basic speech features, enabling the target dialect speech synthesis model to adapt to new dialect data more quickly, further improving the efficiency of dialect speech synthesis.

[0218] Example 4

[0219] Please see Figure 5 , Figure 5 This is a schematic diagram of another prosodic text generation device for dialects disclosed in an embodiment of the present invention. Figure 5 The described device can be used in application servers. For example... Figure 5 As shown, the device may include:

[0220] Memory 501 storing executable program code;

[0221] Processor 502 coupled to memory 501;

[0222] Furthermore, it may also include an input interface 503 and an output interface 504 coupled to the processor 502;

[0223] The processor 502 calls the executable program code stored in the memory 501 to execute the steps of the method described in any one of the embodiments of the present invention, from Embodiment 1 to Embodiment 2.

[0224] Example 5

[0225] This invention discloses a computer-storable medium storing computer instructions, which, when invoked, execute the steps of the method described in any one of the embodiments of this invention (Embodiment 1 to Embodiment 2).

[0226] Example 6

[0227] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the method described in any one of the embodiments of this invention, from Embodiment 1 to Embodiment 2.

[0228] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0229] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0230] Finally, it should be noted that the prosodic text generation method and apparatus for dialects disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating prosodic text in dialects, characterized in that, The method includes: For any of the sample dialects in the multiple sample dialects, for any syllable of the sample dialect, the syllable pinyin code corresponding to the syllable is determined according to the dialect initial consonant string, dialect final vowel string and dialect tone character of the syllable in a predetermined dialect pinyin encoding order. The syllable pinyin codes corresponding to all syllables of the sample dialect are used as the dialect pinyin codes corresponding to the sample dialect, wherein the dialect initial consonant string is a string that can be omitted. Determine the dialect pinyin codes corresponding to all the sample dialects, which are multiple pre-determined dialect pinyin codes; The text to be labeled in the target dialect is segmented according to a preset text segmentation method to obtain all the original units of the text to be labeled. All the original units include one or more of Chinese characters, Pinyin strings and number strings. Based on the type of the target dialect, a target dialect pinyin code that matches the target dialect is determined from a plurality of pre-determined dialect pinyin codes. The target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect. For any of the original units, the original units are encoded based on the target dialect pinyin encoding to obtain the target units of the original units; Based on the sorting order of the original units corresponding to each target unit in the text to be annotated, all target units are sorted to obtain the dialect prosodic text corresponding to the text to be annotated.

2. The method for generating prosodic text in dialects according to claim 1, characterized in that, The method further includes: For any one of the sample dialects in the multiple sample dialects, perform phonetic analysis on all phonemes in the sample dialect to extract basic phonetic features, wherein the multiple sample dialects include the target dialect; Based on the basic phonetic features, a set of dialect initial consonant strings and a set of dialect final vowel strings are determined. The set of dialect initial consonant strings includes all dialect initial consonant strings representing the pronunciation of the beginning part of all syllables in the sample dialect. The set of dialect final vowel strings includes all dialect final vowel strings representing the pronunciation of the non-beginning part of all syllables in the sample dialect. Based on the basic speech features, a set of dialect tone characters is determined, which includes all dialect tone characters representing all pronunciation tones in the sample dialect.

3. The method for generating prosodic text in dialects according to claim 2, characterized in that, The step of determining the set of dialect initial consonant strings and the set of dialect final vowel strings based on the basic phonetic features includes: Phonetic analysis of all phonemes in the standard language is performed to extract the phonetic features of the standard language; By comparing the pronunciation differences between the basic phonetic features and the phonetic features of the standard language, a pronunciation difference comparison result is obtained. Based on the pronunciation difference comparison result, the initial consonant strings of the standard language are modified to obtain a set of dialect initial consonant strings. Based on the pronunciation difference comparison result, the final vowel strings of the standard language are modified to obtain a set of dialect final vowel strings.

4. The method for generating prosodic text in dialects according to any one of claims 1-3, characterized in that, For any of the original units, encoding the original unit based on the target dialect pinyin encoding to obtain the target unit of the original unit includes: When the original unit is the numeric string, the current application scenario of the target dialect is determined, and the dialect numeric pronunciation codes of all single-digit numbers in the current application scenario and the dialect numeric combination pronunciation methods of number combinations are obtained from the target dialect pinyin encoding. It is then determined whether the numeric string is a single-digit number. If the numeric string is determined to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to the single-digit numbers in the current application scenario to obtain the target unit of the numeric string. If the numeric string is determined not to be a single-digit number, it is encoded according to the dialect numeric pronunciation codes corresponding to all single-digit numbers included in the numeric string in the current application scenario and the dialect numeric combination pronunciation methods of the number string combinations in the current application scenario. The dialect pronunciation of the number combination corresponding to the number combination is used to encode the number string to obtain the target unit of the number string. The dialect pronunciation encoding of each single digit is the same in the same application scenario, and the dialect pronunciation encoding of each single digit is different in different application scenarios. The dialect pronunciation of each number combination is the same in the same application scenario, and the dialect pronunciation of each number combination is different in different application scenarios. The number combination includes at least two single digits. The type of the number combination is determined by the context of the number combination in the text to be annotated. The dialect pronunciation encoding is a unique encoding of the syllable corresponding to the single digit in the target dialect in the current application scenario. When the original unit is the Chinese character, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the target dialect pinyin code as the Chinese character pronunciation code of the Chinese character. The Chinese character is then encoded using the Chinese character pronunciation code to obtain the target unit of the Chinese character. When the original unit is the Pinyin string, a unique code that can describe the syllable corresponding to the Chinese character in the target dialect is determined from the Pinyin encoding of the target dialect as the Pinyin pronunciation code of the Pinyin string. The Pinyin pronunciation code is used to encode the Pinyin string to obtain the target unit of the Pinyin string.

5. The method for generating prosodic text in dialects according to any one of claims 1-3, characterized in that, The original unit also includes punctuation characters. Before sorting all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, to obtain the dialect prosodic text corresponding to the text to be annotated, the method further includes: Based on the type of the target dialect, determine the punctuation code corresponding to each punctuation character. The punctuation code is used to represent the pause duration of the punctuation character at the corresponding position in the speech stream. Encode the punctuation character according to the punctuation code to obtain the target unit of the punctuation character. The punctuation mark encoding is determined in the following way: Based on the type of the target dialect, at least one speech pause pattern is predefined, and each speech pause pattern maintains a silence mode for a different duration in the speech stream; Define a unique speech pause pattern encoding symbol for each of the aforementioned speech pause patterns; The speech pause pattern encoding symbol matching each of the sample punctuation characters in all sample punctuation characters of the target dialect is determined to obtain the punctuation code, and all sample punctuation characters contain the punctuation character.

6. The method for generating prosodic text in dialects according to any one of claims 1-3, characterized in that, Before segmenting the text to be labeled in the target dialect according to a preset text segmentation method, the method further includes: Based on the type of the target dialect, a text cleaning method is determined, and text cleaning operations are performed on the text to be labeled according to the determined text cleaning method. The text cleaning operations include deleting duplicate characters and / or deleting pre-determined illegal characters; and / or Based on the type of the target dialect, determine the punctuation deletion method, and perform punctuation deletion operation on the text to be annotated according to the punctuation deletion method; and / or Determine whether there is other language text in the text to be labeled that is different from the target dialect type. If there is other language text, convert the other language text into text that is the same as the target dialect type.

7. The method for generating prosodic text in dialects according to any one of claims 1-3, characterized in that, After sorting all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, to obtain the dialect prosodic text corresponding to the text to be annotated, the method further includes: The dialect prosodic text is converted into a playable dialect speech file using a pre-determined dialect speech synthesis model. The dialect speech synthesis model is obtained by inputting the dialect prosodic text into the speech synthesis model corresponding to the standard language and performing transfer learning.

8. A device for generating prosodic text for dialects, characterized in that, The device includes: The dialect syllable encoding module is used to determine the syllable pinyin encoding for any syllable of any sample dialect from multiple sample dialects, based on the dialect initial consonant string, dialect final vowel string, and dialect tone character corresponding to the syllable in a predetermined dialect pinyin encoding order. The syllable pinyin encoding corresponding to all syllables of the sample dialect is used as the dialect pinyin encoding for the sample dialect. The dialect initial consonant string is a string that can be omitted. The segmentation module is used to segment the text to be labeled in the target dialect according to a preset text segmentation method to obtain all the original units of the text to be labeled. All the original units include one or more of Chinese characters, Pinyin strings and number strings. The determining module is configured to determine, based on the type of the target dialect, a target dialect pinyin code that matches the target dialect from a plurality of pre-determined dialect pinyin codes, wherein the target dialect pinyin code is used to represent the pronunciation of all syllables in the target dialect; and is further configured to, after the dialect syllable encoding module determines the dialect pinyin codes corresponding to all the sample dialects, determine the dialect pinyin codes corresponding to all the sample dialects as a plurality of pre-determined dialect pinyin codes; An encoding module is used to encode any one of the original units based on the target dialect pinyin encoding to obtain the target unit of the original unit; The sorting module is used to sort all the target units according to the sorting order of the original units corresponding to each target unit in the text to be annotated, so as to obtain the dialect prosodic text corresponding to the text to be annotated.

9. A device for generating prosodic text for dialects, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the prosodic text generation method applied to dialects as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when invoked, are used to execute the prosodic text generation method for dialects as described in any one of claims 1-7.

Images