A Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains

Abstract

A method for converting Tibetan-Sanskrit spelling sequences based on grammatical rule chains is proposed, relating to the fields of Tibetan speech synthesis and linguistics. By analyzing features such as superposition, reverse writing, and long vowels in Tibetan-Sanskrit, a grammatical rule chain framework is designed, comprising pre-processing, Sanskrit-Tibetan conversion mapping, and post-processing. Combined with an automated data conversion mapping library, the conversion of Tibetan-Sanskrit syllables to standard Tibetan syllables is achieved. This method has two major innovations: a deterministic rule matrix framework built upon traditional authoritative spelling theories; and the introduction of superimposed consonant serialization and special vowel normalization techniques to accurately decompose complex reduplicated character structures. Experimental results show that the method achieves a data conversion accuracy of 99.8% on a complex Tibetan-Sanskrit test set, while effectively reducing the pronunciation error rate of the downstream speech synthesis system to 2.82%. This method solves the problems of Tibetan-Sanskrit text parsing failure and synthesized sound loss in existing Tibetan speech synthesis, and is of great significance for promoting the development of ethnic language information processing technology.

Description

Technical Field

[0001] This invention relates to the field of Tibetan speech synthesis and linguistics, specifically to a method for automatically converting Sanskrit and Tibetan pronunciation data. Background Technology

[0002] Tibetan Sanskrit phonics refers to a method of pronunciation that uses the "fifty-letter Tibetan alphabet system" to standardize the representation of Sanskrit sounds and applies specific rules for interpretation. In modern Tibetan texts, Tibetan Sanskrit is widely used in common fields such as Tibetan medicine (e.g., "གྷི་ཝཾ།"), place names (e.g., "ཀིརྟི་དགོན།"), and personal names (e.g., "ཀརྨ་སྐྱིད།"). Due to its extremely complex structure and typical long-tail distribution in real-world corpora, existing Tibetan speech synthesis models frequently suffer from pronunciation distortion or dropped sounds when processing such texts. Therefore, there is an urgent need for a data conversion and rule processing engine that can seamlessly integrate Tibetan Sanskrit. Overcoming this technical bottleneck is of great significance for promoting the development of Tibetan speech synthesis technology, and will directly improve the performance and accuracy of Tibetan speech synthesis systems and Sanskrit standardized pronunciation teaching aids.

[0003] In the field of linguistics, the study of the evolution of Tibetan from ancient times to the present has always been inseparable from a deep analysis of the Sanskrit phonological system. Throughout its long history, scholars have written numerous theoretical works systematically exploring Sanskrit spelling. Among them, the *Concise Collection of Mantra Recitation Methods* compiled by the translator Natang in 1420 and *The Method of Reading the Alphabet for Children* by Sakya Sonam Tsemo are widely recognized as the most systematic and authoritative guiding documents. In recent years, modern linguists have built upon this foundation to conduct more systematic diachronic and synchronic phonological studies. For example, by excavating Sanskrit spelling materials from different historical periods, they have explored the intrinsic mapping relationship between the historical evolution of Tibetan phonetics and Tibetan Sanskrit spelling methods, systematically sorted out the historical origins of different Tibetan spelling notations, and summarized the diachronic phonological characteristics of Tibetan and Sanskrit. Related research focuses on the phonological changes of multidimensional superimposed characters in complex contexts such as liaison, euphony, and cessation.

[0004] In the field of Tibetan speech processing, with the rapid development of deep learning, Tibetan speech synthesis technology has made significant progress, transitioning from traditional waveform concatenation methods to end-to-end generative models based on neural networks. Despite substantial advancements in acoustic models, the core challenges in Tibetan speech synthesis have gradually shifted to the design of basic front-end text processing and modeling units. Early research typically used "Tibetan character components" as text input primitives to effectively balance the integrity of large units with the flexibility of small units. Subsequent low-resource, multi-dialect Tibetan synthesis methods based on Tibetan character components have greatly reduced the difficulty of model fitting. However, current component-based processing primarily focuses on morpheme construction and model training for standard Tibetan. When faced with the structurally complex and typically long-tailed Tibetan Sanskrit, existing front-end technologies exhibit significant limitations. Although some studies have attempted to mechanically add a small number of Sanskrit characters to the standard syllables, the failure to exhaustively cover the reduplicated combinations, special vowels, and reverse spelling rules of Sanskrit has led to frequent text parsing failures and synthesized sound loss when the models process real documents or general texts containing Sanskrit (such as medical terms, place names, and personal names). Summary of the Invention

[0005] The purpose of this invention is to address the problem of ineffective synthesis of Tibetan Sanskrit syllables in existing Tibetan speech synthesis technologies, and the lack of standardized spelling rules in Sanskrit learning. It innovatively proposes a Tibetan-Sanskrit spelling data conversion method based on grammatical rules and applies it to Tibetan speech synthesis systems and Sanskrit spelling learning platforms. This method is based on the authoritative traditional Tibetan-Sanskrit spelling theories, *A Concise Collection of Mantra Recitation Methods* and *The Introduction to Alphabet Reading for Children*, systematically analyzing the spelling rules of superimposed consonants, reversed consonants, and various vowel symbols in Tibetan-Sanskrit, and constructing an automated data conversion mapping library. Specifically, the method first constructs a preset mapping library based on Tibetan grammatical rules; second, it converts the input Tibetan-Sanskrit syllables into spelling data using the preset mapping library; finally, it applies the converted spelling data to downstream tasks. Experimental results show that the accuracy of the proposed spelling data conversion method in Tibetan-Sanskrit pronunciation data conversion reaches a state-of-the-art level, significantly outperforming existing mainstream large language models. In a comparative experiment on Tibetan speech synthesis, the text processed using this method showed a significant reduction in pronunciation errors, resulting in a substantial improvement in speech synthesis quality. This research not only provides an effective technical solution for converting Tibetan Sanskrit spelling data, but also offers new insights into addressing the problem of ineffective synthesis of Tibetan Sanskrit syllables in Tibetan speech synthesis technology.

[0006] To achieve the above objectives, this invention provides a Tibetan-Sanskrit spelling sequence conversion method based on grammatical rule chains, which reduces the time and manpower costs of rebuilding and training Tibetan speech synthesis models and data, and provides high-quality speech synthesis data with strong interpretability and zero cost for existing Tibetan speech synthesis models.

[0007] A method for converting Tibetan-Sanskrit spelling sequences based on grammatical rule chains includes the following steps: S1: Text preprocessing: Cleaning and syllable segmentation of the Tibetan text to be processed; S2: Tibetan Sanskrit Recognition: Using the principles of Tibetan syllable construction, identify Tibetan Sanskrit syllable sequences from the sequence after syllable segmentation; S3: Preset Mapping Rule Base Configuration: Configure a preset mapping rule base based on grammar rule chains. The processing flow of the preset mapping rule base includes preprocessing, Sanskrit-Tibetan conversion mapping, and postprocessing: First, preprocessing is used to unify encoding and vertically segment syllables; second, specific processing is performed using Sanskrit-Tibetan conversion mapping methods, including special consonant preprocessing, superimposed consonant serialization, pronunciation standardization, and vowel normalization; finally, postprocessing is used to merge sequences and remove redundant information, and to check whether the converted syllables conform to the Tibetan syllable construction principles. S4: Tibetan-Sanskrit spelling sequence conversion: The Tibetan-Sanskrit syllable sequence is converted into standard Tibetan syllables using the preset mapping rule library.

[0008] Furthermore, the Tibetan Sanskrit recognition method specifically involves: first, receiving a list of multiple syllables as input, i.e., the syllable list after S1 segmentation; then, performing a spelling check on each syllable in the list based on the Tibetan syllable construction principles; finally, filtering out syllables that do not meet the syllable construction principles and are Tibetan consonant combinations, and obtaining the index position of the syllable in the list.

[0009] Furthermore, step S3 specifically includes the following steps: S301: Execute the encoding unification method, specifically: determine whether there is a preset specific consonant or long vowel in the Tibetan Sanskrit syllable to be processed; if so, replace its combined encoding with the corresponding standard encoding; if not, skip it. S302: Execute the vertical segmentation method, specifically: based on S301, segment the character based on the construction principle of Tibetan character T and its Unicode encoding combination characteristics, and output a sequence composed of characters T; S303: Execute the special consonant preprocessing method, specifically: First, determine whether there is a preset specific consonant in the segmentation sequence of S302. If so, replace it according to the preset mapping table; second, further determine whether there is a special overlapping consonant in the sequence. If so, replace the overlapping consonant with a single consonant; finally, further determine whether there is a special overlapping consonant in the sequence. If so, replace the first character of the overlapping consonant with the suffix of the previous character; if not, skip. S304: Execute the superimposed consonant serialization method, specifically: determine whether there are underpinned characters in the processed sequence of S303; if so, replace all underpinned characters with their corresponding standard consonants and save them separately in the sequence; if not, skip this step. S305: Execute the pronunciation standardization method, specifically: determine whether a preset specific consonant exists in the processing sequence of S304; if it exists, replace it according to the preset mapping table; if it does not exist, skip it. S306: Execute the vowel normalization method, specifically: determine whether there are preset specific vowels such as long vowels or nasal consonants in the processing sequence of S305; if they exist, replace them according to the preset mapping table; if they do not exist, skip. S307: Execute the sequence merging method, specifically: based on the processed sequence in S306, use Tibetan syllable symbols to concatenate strings and return the concatenated string; S308: Execute the redundancy removal method, specifically: based on the processing sequence of S307, use a preset character filtering matrix to match and remove redundant Tibetan syllables from the string; S309: Output the standard Tibetan syllables after the spelling data conversion, and end the process.

[0010] The advantages of this invention are as follows: By systematically analyzing the superposition logic and speech mapping rules of Sanskrit sounds, traditional grammatical knowledge is transformed into a computer-executable rule matrix. Compared to purely data-driven models, rule-based methods based on linguistic prior knowledge possess higher logical completeness and fidelity, providing controllable and deterministic front-end constraints. This not only significantly reduces the front-end parsing difficulty of speech synthesis models, effectively reducing the pronunciation error rate by 2.82% when processing texts containing Tibetan Sanskrit syllables, thus effectively improving the predicament of the inability to effectively synthesize Tibetan Sanskrit syllables; but also, this method achieves the direct output of zero-cost, highly interpretable, high-quality speech synthesis data, thereby eliminating the significant time, space, and human resource investment required for secondary construction and repeated training of existing speech synthesis models and data, providing a highly practical optimization path for the efficient deployment of related speech synthesis systems and learning platforms. Attached Figure Description

[0011] To more clearly illustrate the technical solution of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Wherein: Figure 1 This is an overall flowchart of the Tibetan-Sanskrit spelling sequence conversion method of the present invention; Figure 2 This is a flowchart of the Tibetan-Sanskrit character recognition process of the present invention; Figure 3 This is a flowchart of the grammar-based rule chain method of the present invention; Figure 4 This is a schematic diagram of the unified encoding mapping library of the present invention; Figure 5 This is a flowchart of the special consonant preprocessing of the present invention; Figure 6 This is a schematic diagram of the Sanskrit pronunciation standardization mapping library of the present invention; Figure 7 This is a flowchart of the vowel normalization algorithm of the present invention. Detailed Implementation

[0012] To more clearly illustrate the technical solution of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and specific examples. The specific embodiments of the present invention are not limited to the following examples, and appropriate changes and modifications can be made according to actual needs without affecting the substance and scope of the present invention.

[0013] Figure 1 This is an overall flowchart of the Tibetan-Sanskrit spelling sequence conversion method of the present invention, as follows: Figure 1 As shown, a method for converting Tibetan Sanskrit spelling sequences based on grammar rule chains is presented, which mainly includes text preprocessing, Sanskrit recognition, and grammar rule chains.

[0014] S1: Perform text preprocessing, including text cleaning and syllable segmentation; Specifically, this embodiment segments the Tibetan text into syllables using Tibetan syllable symbols ("་"), Tibetan vertical symbols ("།"), Tibetan numerals ("༠-༩"), Tibetan special symbols, and non-Tibetan characters, and returns a list composed of Tibetan syllables.

[0015] S2: Perform Tibetan and Sanskrit character recognition. Figure 2 This is a flowchart of Tibetan and Sanskrit character recognition according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: S201: Input Tibetan syllable sequence; S202: Perform spelling checks on each syllable in the Tibetan syllable sequence based on syllable construction principles; S203: Determine whether the current syllable conforms to the Tibetan syllable construction principles. If it does, proceed to S204; otherwise, proceed to S205. S204: Mark the current syllable as label "F" to indicate a non-Tibetan Sanskrit syllable, and proceed to S208; S205: Determine whether the current syllable is a combination of Tibetan letters, i.e., any combination within the Unicode encoding range [\u0F40-\u0FBC]. If it matches, proceed to S206; otherwise, proceed to S207. S206: Mark the current syllable as the label "T", indicating a Tibetan Sanskrit syllable, and proceed to S208; S207: Mark the current syllable with the label "F" to indicate a non-Tibetan Sanskrit syllable; S208: Output the syllable sequence with labels and end the process.

[0016] S3: Using a pre-defined mapping rule base based on grammar rule chains, the Tibetan Sanskrit syllables identified by S2 are converted into spelling data. Figure 3 This is a flowchart of the grammar rule chain processing according to an embodiment of the present invention, such as... Figure 3 As shown, the data transformation process includes three steps: Specifically, the processing flow of the preset mapping rule base includes preprocessing, Sanskrit-Tibetan conversion mapping, and postprocessing: the preprocessing includes encoding unification and syllable vertical segmentation; the Sanskrit-Tibetan conversion mapping includes special consonant preprocessing, superimposed consonant serialization, pronunciation standardization, and vowel normalization; the postprocessing includes sequence merging, redundancy removal, and syllable construction principle checking, totaling 9 steps.

[0017] S301: Perform encoding standardization. Figure 4 This is a schematic diagram of the unified encoding mapping library according to an embodiment of the present invention, as shown below. Figure 4 As shown, the process includes the following steps: S30101: Input Tibetan Sanskrit syllables; S30102: Determine whether the combined code in the preset encoding library exists in the currently input Tibetan Sanskrit syllable. If it exists, proceed to S30103; otherwise, proceed to S30104. S30103: Replace the combined encoding with the corresponding single encoding (e.g., the Unicode encoding of the consonant "གྷ" is a combination of \u0F42 and \u0FB7, which will be replaced with the single encoding \u0F43). S30104: Output the encoded Tibetan Sanskrit syllables and end the process.

[0018] S302: In this embodiment, syllable vertical segmentation will be performed, specifically: based on the construction principle of Tibetan syllable components and the Unicode encoding combination rule, that is, use a preset regular expression to segment the result of step S301, and output a sequence composed of syllable components (for example, the vertical segmentation result of "པདྨ" is ["པ", "དྨ"]).

[0019] S303: Perform special consonant preprocessing. Figure 5 This is the flowchart of the special consonant preprocessing in the embodiment of the present invention. As Figure 5 shown, this process includes the following steps: S30301: Input a sequence of Tibetan syllable components. S30302: Determine whether there is a preset specific consonant in the syllable component sequence and whether there is a syllable sign before it. If so, enter S30303; otherwise, enter S30304. S30303: Add the Tibetan consonant "འ" before this syllable component (for example, replace "གུ" with "འགུ"), update the sequence, and enter S30306. S30304: Determine whether there is a special identical stacked consonant in the syllable component sequence and whether there is a syllable sign before it. If so, enter S30305; otherwise, enter S30306. S30305: Replace the identical stacked consonant with the corresponding single consonant (for example, replace "་ཀྐ" with "་ཀ"), and update the sequence. S30306: Starting from the second position of the syllable component sequence, determine whether the current syllable component has a stacked consonant and its initial consonant is a specific consonant. If so, enter S30307; otherwise, enter S30308. S30307: Replace the initial consonant of the current syllable component with the postfix of the previous syllable component (for example, replace ["པ", "དྨ"] with ["པད", "ྨ"]), and update the sequence. S30308: Output the updated syllable component sequence, and end the process.

[0020] S304: In this embodiment, stacked consonant serialization will be performed, specifically: determine whether each syllable component in the syllable component sequence has a subscript character. If so, uniformly replace all subscript characters, that is, consonant characters in the Unicode encoding range [\u0F90-\u0FBC], with their corresponding standard basic consonant encodings and save them separately in the syllable component sequence (but 7 specific consonants such as "ཀྱ" cannot be converted), and output the updated syllable component sequence (for example, the syllable component "ཀྵྨྼྻུ" is serialized into ["ཀྵ་", "མ་", "ར་", "ཡུ"]).

[0021] S305: Perform pronunciation normalization. Figure 6This is a schematic diagram of the Sanskrit pronunciation standardization mapping library according to an embodiment of the present invention, as shown below. Figure 6 As shown, the process includes the following steps: S30501: Input the Tibetan character sequence; S30502: Determine whether the consonant in the preset specific pronunciation standard library exists in the current input sequence. If it exists, proceed to S30503; otherwise, proceed to S30504. S30503: Replace specific consonants with their corresponding standard Tibetan pronunciations (e.g., replace the consonant “གྷ” with “རྒ”). S30504: Output the sequence after phonetic normalization and end the process.

[0022] S306: Perform vowel normalization. Figure 7 This is a flowchart of the vowel normalization algorithm according to an embodiment of the present invention, as follows: Figure 7 As shown, the process includes the following steps: S30601: Input the given character sequence; S30602: Load the Tibetan consonant classification dictionary and long vowel replacement library; S30603: Determine whether there is a specific Tibetan Sanskrit syllable "ཨོཾ" or "ཧཱུཾ" in the sequence of characters. If it exists, proceed to S30604; otherwise, proceed to S30605. S30604: Replace the special Tibetan Sanskrit syllables “ཨོཾ” and “ཧཱུཾ” with “ཨོམ” and “ཧུང” respectively; S30605: Determine whether the special vowel in the preset long vowel replacement library exists in the current character sequence. If it exists, proceed to S30606; otherwise, proceed to S30607. S30606: Replace the corresponding long vowel with the preset standard Tibetan vowel; S30607: Determine whether the following sound "ཾ" exists in the currently input character sequence. If it exists, proceed to S30608; otherwise, proceed to S30609. S30608: Based on the consonant of the character following the consonant mark or the consonant preceding it, obtain the corresponding replacement result from the Tibetan consonant classification dictionary and perform the replacement. S30609: Determine whether the final rhyme “ཿ” exists in the current character sequence. If it exists, proceed to S30610; otherwise, proceed to S30611. S30610: Replace the rhyme punctuation mark with the suffix "ས"; S30611: Output the vowel-normalized sequence of characters and end the process.

[0023] S307: Perform a sequence merging method, specifically: use Tibetan syllables to concatenate strings and return the concatenated string.

[0024] S308: Perform a redundancy removal method, specifically: remove redundant Tibetan syllables and preset redundant information, and at the same time perform a spelling check based on syllable construction principles. If the spelling matches, the conversion is considered successful; otherwise, the conversion is considered to have failed.

[0025] In this embodiment, several mainstream existing large language models with Tibetan language generation and comprehension capabilities were selected as comparison models for experimental verification. The specific comparison results are shown in Table 1: Table 1 Existing large language model A 22.4 7.2 Existing large language model B 26.4 6.4 Existing large language model C 22.4 10.4 Existing large language model D 27.2 8.8 Existing Tibetan language large language model 21.6 4.0 The method of the present invention 100 99.8 Table 1 compares the accuracy of the spelling conversion results using the method of this invention with mainstream large-scale models on simple and complex test sets. As can be seen from Table 1, although mainstream large-scale models possess a certain logical reasoning ability under strong cue word guidance, their accuracy drops significantly on complex test sets due to the randomness and illusion problems inherent in their generation mechanism, particularly when dealing with the vertical and demanding phonological task of Tibetan and Sanskrit. In contrast, the method of this invention solidifies linguistic prior knowledge into a deterministic algorithm, achieving accuracy rates of 100% and 99.8% on the two test sets respectively, demonstrating extremely strong logical completeness. Experimental results show that the method of this invention achieves the highest performance indicators in both simple and complex tasks.

[0026] Table 2 Original text 8.29 7.08 15.20 After using the method of the present invention 9.42 9.16 2.82 Table 2 shows a comparison of the pronunciation accuracy of Tibetan syllables using the method of this invention and the method without this invention in Tibetan speech synthesis (test data are sentences containing Tibetan Sanskrit). As can be seen from Table 2, without Tibetan Sanskrit conversion, the existing Tibetan speech synthesis model has a high pronunciation error rate. The fundamental reason is that the feature extraction logic of the acoustic model collapses when faced with unfamiliar complex superimposed structures, leading to cascading distortions or loss of pronunciation. After preprocessing with the rule engine using the method of this invention, the pronunciation error rate is significantly reduced.

[0027] The specific embodiments of the present invention have been described above, but are merely illustrative examples, and the related interface diagrams are schematic diagrams. Various modifications can be made to these embodiments in practical applications without departing from the principles and essence of the present invention. Therefore, the scope of protection of the present invention is limited by the appended claims.

Claims

1. A method for converting Tibetan-Sanskrit spelling sequences based on grammatical rule chains, characterized in that, The Tibetan-Sanskrit spelling sequence conversion method includes: Text preprocessing: Cleaning and syllable segmentation of the Tibetan text to be processed; Tibetan-Sanskrit Recognition: Using the principles of Tibetan syllable construction, Tibetan-Sanskrit syllable sequences are identified from the syllable segmentation sequence. Preset mapping rule base configuration: Configure a preset mapping rule base based on grammar rule chains. The processing flow of the preset mapping rule base includes preprocessing, Sanskrit-Tibetan conversion mapping, and postprocessing: First, preprocessing performs encoding unification and syllable vertical segmentation; second, specific processing is performed using Sanskrit-Tibetan conversion mapping methods, including special consonant preprocessing, superimposed consonant serialization, pronunciation standardization, and vowel normalization; finally, postprocessing performs sequence merging and redundant information removal, and checks whether the converted syllables conform to the Tibetan syllable construction principles. Tibetan-Sanskrit spelling sequence conversion: The Tibetan-Sanskrit syllable sequence is converted into standard Tibetan syllables using the preset mapping rule library.

2. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 1, characterized in that, The specific steps of the encoding unification process are as follows: determine whether there is a preset combination encoding of specific consonants or long vowels in the Tibetan Sanskrit syllables to be processed; if so, replace the combination encoding with the corresponding standard single encoding.

3. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 1 or 2, characterized in that, The syllable vertical segmentation step is as follows: for the syllable sequence after the encoding is unified, further segmentation is performed in combination with the construction principle of Tibetan characters and their Unicode encoding combination features, and the resulting sequence of characters composed of characters is output.

4. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 1 or 3, characterized in that, The special consonant preprocessing steps include: For the character sequence, firstly, determine whether there is a preset specific consonant in each character and whether there is a syllable symbol before the character; if so, add the Tibetan consonant "འ" before the character and update the character sequence; Secondly, determine whether there is a preset overlapping consonant in each character and whether there is a syllable symbol before the character; if so, replace the overlapping consonant with the corresponding single consonant and update the character sequence. Then, starting from the second position of the character sequence, determine whether the current character has a superimposed consonant and its first consonant is a specific consonant; if so, replace the first consonant of the current character with the suffix of the previous character and output the updated character sequence.

5. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 4, characterized in that, The consonant stacking sequencer step includes: For the given character sequence, determine whether there is a subscript in each character; if so, replace the codes of all subscripts with their corresponding standard basic consonant codes, and update the character sequence to the sequence of all basic consonants.

6. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 5, characterized in that, The pronunciation standardization steps include: Determine whether any consonant in the preset specific pronunciation standard library exists in the character sequence; if it exists, replace the specific consonant with the corresponding standard Tibetan pronunciation and output the pronunciation standardization sequence.

7. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 6, characterized in that, The vowel normalization step includes: Configure a preset Tibetan consonant classification dictionary and long vowel replacement library; First, determine whether the sequence of characters contains a preset special Tibetan Sanskrit syllable "ཨོཾ" or "ཧཱུཾ"; if it does, replace the special Tibetan Sanskrit syllables "ཨོཾ" and "ཧཱུཾ" with "ཨོམ" and "ཧུང" respectively; Secondly, determine whether the special vowel in the preset long vowel replacement library exists in the word sequence; if it exists, replace the corresponding long vowel with the preset standard Tibetan vowel. Then, determine whether the consonant "ཾ" exists in the sequence of characters; if it exists, then according to the consonant of the character following the consonant symbol or the consonant preceding it, obtain the corresponding suffix from the preset Tibetan consonant classification dictionary and replace it; Finally, determine whether the stop rhyme "ཿ" exists in the sequence of characters; if it exists, replace the stop rhyme symbol with the suffix "ས" and output a sequence composed of standard Tibetan syllables.

8. The Tibetan-Sanskrit spelling sequence conversion method based on grammar rule chains according to claim 7, characterized in that, The sequence merging method specifically involves merging the sequences using Tibetan syllable symbols.

Images