A rule-driven Tibetan inscription class i case particle marking method

Abstract

The application discloses a rule-driven Tibetan inscription i-class case particle marking method, wherein the i-class case particle presents systematic variation around the 9th century in Tibetan, which makes traditional syntactic analysis complicated and difficult, in order to solve the problem, the rule-driven Tibetan inscription i-class case particle marking method constructs a diachronic rule adaptation framework, and designs a set of dynamic rule adaptation algorithms based on the framework, can automatically call the adapted grammar rule set for marking according to the age metadata of the inscription, the system sorts out the related rules in the basic Tibetan grammar 'Thirty Odes' and 'Character Nature Organization Method', combines historical linguistics analysis, constructs a formal diachronic rule knowledge base, and clearly distinguishes the two grammar systems before and after the Tibetan standardization; the key use cases of the Enlan Dazhalugonggong stele and the Tang-Ban alliance stele are verified, and the results show that the method can logically and consistently handle most of the standard and variation phenomena in the two steles.

Description

Technical Field

[0001] The present invention relates to the technical field of labeling class I case particles in Tibetan inscriptions, and specifically provides a method for labeling class I case particles in Tibetan inscriptions based on rule-driven Background Art

[0002] The digitalization and informatization processing of ancient Tibetan have great strategic significance for protecting and inheriting Tibetan cultural heritage and deepening historical linguistics research. Among them, the case particle system in Tibetan inscription documents reflects the standardization process of grammar before and after the finalization of Tibetan; the information processing of ancient Tibetan faces three core dilemmas, which restrict the in-depth excavation and utilization of inscription documents:

[0003] I. Complexity of character variation: The omission phenomenon of the Tibetan postfix "" involves the combination rules of three upper add-on characters "" and four lower add-on characters "", and the writing habit differences among different inscriptions lead to increased difficulty in character recognition; for example, in ancient Tibetan documents and other forms with the postfix are simplified to in modern writing, with meanings of "I, horse" respectively. This diachronic difference reduces the applicability of character matching rules in inscriptions; etc. with the postfix are simplified to in modern writing, with meanings of "I, horse" respectively. This diachronic difference reduces the applicability of character matching rules in inscriptions;

[0004] II. Problems of grammatical multi-functionality and diachronic variation: Class I case particles (genitive "", often corresponding to the structural particle 'of / ' in Chinese translation, mainly indicating syntactic and semantic associations such as possessive relationship, restrictive modification, attribute dependence, and appositive co-reference, etc.) and ergative ( "" indicates the agent of an action, and also indicates tools, reasons, manners, etc.) have significant functional generalization phenomena. The genitive not only indicates relationships such as possession (e.g., "" Tibetan culture), but can also indicate non-core functions such as contrast (e.g., "" We are nationals, not foreigners), tense (e.g., "" being sent), etc.; more critically, the case particle addition methods in early inscriptions (such as the Dazhalugong Stele) differ from the grammar norms based on the Thirty Verses in terms of addition rules. For example, after the masculine character is followed by the feminine case particle This diachronic variation makes it difficult to adjust processing models based on the grammar norms of the Thirty Verses, etc.; "Tibetan culture), but can also indicate non-core functions such as contrast (e.g., "" We are nationals, not foreigners), tense (e.g., "" being sent), etc.; more critically, the case particle addition methods in early inscriptions (such as the Dazhalugong Stele) differ from the grammar norms based on the Thirty Verses in terms of addition rules. For example, after the masculine character "We are nationals, not foreigners), tense (e.g., "" being sent), etc.; more critically, the case particle addition methods in early inscriptions (such as the Dazhalugong Stele) differ from the grammar norms based on the Thirty Verses in terms of addition rules. For example, after the masculine character "being sent), etc.; more critically, the case particle addition methods in early inscriptions (such as the Dazhalugong Stele) differ from the grammar norms based on the Thirty Verses in terms of addition rules. For example, after the masculine character is followed by the feminine case particle This diachronic variation makes it difficult to adjust processing models based on the grammar norms of the Thirty Verses, etc.;

[0005] Third, the scarcity of annotation resources: The existing ancient Tibetan corpus is limited in size, and the linguistic levels covered by its annotations are relatively limited, resulting in a significant deficiency in the generalization ability of deep learning models and other technologies in the processing of epigraphic documents due to data sparsity. Therefore, exploring a rule-driven method that conforms to Tibetan grammatical norms and can explain the diachronic variations of type i case particles has important theoretical and applied value for the information processing of ancient Tibetan epigraphic documents. Summary of the Invention

[0006] In view of the problems existing in the prior art, this invention discloses a rule-driven method for annotating i-type case particles in Tibetan inscriptions. The technical solution adopted is that the method achieves annotation through a dynamic rule adaptation algorithm (DRDA), specifically including the following steps:

[0007] Step 1: Input and metadata parsing. Receive the inscription text T and its key metadata M. The metadata M includes the core control variable, the date attribute M.era. The value range of M.era is {"before determination", "after determination"}.

[0008] Step 2: Preprocessing and host word identification. Natural language preprocessing is performed on the Tibetan inscription text T, including word segmentation and stemming analysis, to accurately locate the host word H that needs to be added with class i case particles.

[0009] Step 3: Linguistic attribute extraction. For each host word H, the algorithm extracts its key grammatical features.

[0010] Step 4: Rule activation and matching. Based on the value of M.era, dynamically load the corresponding historical rule subset. Load the rule subset before determination and the rule subset after determination. Then match the suffix attribute of the host word H with the condition part of the loaded historical rule subset.

[0011] Step 5: Decision and Ambiguity Resolution. Based on the rule matching results, the labeling decision is executed. If a unique result rule is matched, the case particle form and grammatical function tag are directly output. If a candidate result set rule is matched, the ambiguity resolution module is activated to make a decision and then output the result.

[0012] Step Six: Structured Output. The annotation results are formatted into unified structured data units containing host words, case particles, grammatical functions, and rules. The collection of all units constitutes the complete i-type case particle annotation for the input text T. The i-type case particles include genitive case particles. and as case particles (gis,kyis,gyis,'is,yis), the diachronic rule subset is provided by a formal diachronic rule knowledge base built on the Thirty Odes and the Word Organization Method.

[0013] Furthermore, each rule in the historical rule subset is formally defined using a triplet structure: rule R = <Condition, Action, Metadata Constraint>; where the condition is a logical expression that makes logical judgments based on the suffix attribute of the host word, the action is the output i-type case particle form and its grammatical function label, and the metadata constraint specifies that the rule will only be activated and participate in the calculation if the input metadata such as the inscription date meets specific conditions.

[0014] Furthermore, the rules in the formalized diachronic rule knowledge base are divided into a general rule subset, a pre-determination rule subset, and a post-determination rule subset. The general rule subset includes standard neutral mapping rules, standard positive mapping rules, standard negative mapping rules, and genitive derivation rules. The genitive derivation rules are obtained by adding a prefix to the labeled genitive form. The ergative form is obtained, and the rules for adding ergative particles are consistent with the rules for adding genitive particles.

[0015] Furthermore, the adjudication process of the ambiguity resolution module in step five is as follows: First, the preset inscription-specific variation dictionary is queried. If the dictionary can adjudicate the candidate result set, the annotation result is automatically output. If the dictionary does not cover the candidate result set, the test case is submitted to the manual collaboration interface for expert adjudication, and the expert adjudication result is fed back to the system to optimize the inscription-specific variation dictionary or the formalized historical rule knowledge base.

[0016] Furthermore, the formalized diachronic rule knowledge base also includes inscription-specific rules, which have a higher priority than general rules and are used to handle diachronic variations and special grammatical uses of type i case particles in Tibetan inscriptions.

[0017] Furthermore, the inscription specificity rules include the positive rule for the Dazha Lugong Stele and the positive special case rule for the Tang-Tibet Stele; the positive rule for the Dazha Lugong Stele is that after adding a positive character in the Dazha Lugong Commemorative Stele, a negative auxiliary word is always used. The rule for the positive exception of the Tang-Tibet Stele is that it applies to the addition of characters to the Tang-Tibet Alliance Stele. Follow-match genitive particle The rule is to mark it as a modal particle.

[0018] Furthermore, in step four, the condition part of the historical rule subset is an IF-THEN production rule structure, and the rule matching is to substitute the post-word attribute of the host word H into the IF part of the production rule structure for logical judgment.

[0019] A rule-driven Tibetan inscription i-type case particle annotation system, used to execute a rule-driven Tibetan inscription i-type case particle annotation method, including:

[0020] The data input layer is used to support the import of inscription text and metadata;

[0021] The rules engine layer, the core module, includes an embedded formal rule library and executes the DRDA algorithm;

[0022] The manual collaboration interface provides an entry point for manual adjudication for special instances or ambiguous situations not covered by the rule base, and its feedback can be used to iteratively optimize the rules.

[0023] The output layer is used to generate labeled corpora in a standard format.

[0024] Furthermore, the formal diachronic rule knowledge base in the rule engine layer is computable, interpretable, and scalable: the rules adopt an IF-THEN production structure to achieve computability, the annotation results can be traced back to the linguistic basis through rule ID to achieve interpretability, and the hierarchical storage structure supports incremental addition of new rules or addition of epigraphic exception dictionaries to achieve scalability.

[0025] The application of the rule-driven Tibetan inscription i-type case particle annotation method in the syntactic analysis of Tibetan inscriptions allows the structured output of the method to resolve syntactic structural ambiguities in Tibetan inscriptions, construct comparable Tibetan annotation data over time, and train or validate ancient Tibetan dependency syntax analysis and semantic role annotation models. This provides a data foundation for comparative studies of inscriptions before and after Tibetan text standardization and for studies on the evolution of Tibetan grammar.

[0026] The beneficial effects of this invention are:

[0027] 1. The Dynamic Rule Adaptation Algorithm (DRDA) proposed in this invention uses the metadata of the inscription's era as the core control variable to realize the dynamic loading and matching of the diachronic rule subset before and after the Tibetan text is defined. It can logically and consistently handle most of the normative and variation phenomena of type i case particles in Tibetan inscriptions, and solve the annotation adaptability problem caused by diachronic variation.

[0028] 2. The formal diachronic rule knowledge base is constructed using a triplet structure to define rules, which are divided into a general rule subset, a pre-determination rule subset, and a post-determination rule subset. Inscription-specific rules are also set, and the rule priorities are clear. It can uniformly handle standard syntax mapping and diachronic variation phenomena. At the same time, the knowledge base has computability, interpretability, and scalability. It can be directly converted into program code to achieve automated annotation. The annotation results can be traced back to linguistic basis and are convenient for subsequent rule expansion and optimization.

[0029] 3. An ambiguity resolution module was set up, which combined the inscription-specific variation dictionary and the manual collaboration interface to resolve ambiguities. The resolution results can be fed back to the system for optimization, realizing a closed-loop iteration of the annotation process and effectively improving the accuracy and adaptability of the annotation.

[0030] 4. The system architecture of this invention is clearly hierarchical, with the data input layer, rule engine layer, manual collaboration interface, and result output layer each performing their respective functions and cooperating with each other. It not only realizes automated annotation but also retains the entry point for manual optimization, making it highly practical. The constructed knowledge base and algorithm framework can be directly used as the development basis for annotation tools or parsers, providing a methodological and resource foundation for building a high-quality diachronic tree bank and deepening the research on the evolution of Tibetan grammar. Attached Figure Description

[0031] Figure 1 This is an organizational structure diagram of the Tibetan i-class case particle diachronic rule knowledge base of this invention;

[0032] Figure 2 This is a flowchart of the rule-driven time-adaptive annotation (DRDA) algorithm of this invention. Detailed Implementation

[0033] Example 1

[0034] like Figures 1 to 2 As shown, this invention discloses a rule-driven method for annotating i-type case particles in Tibetan inscriptions. The technical solution adopted is that the method uses a Dynamic Rule Adaptation Algorithm (DRDA) to achieve annotation, specifically including the following steps:

[0035] Step 1: Input and metadata parsing. Receive the inscription text T and its key metadata M. The metadata M includes the core control variable, the era attribute M.era, whose value range is {"before determination", "after determination"}. This step provides the initial conditions for subsequent chronological rule adaptation.

[0036] Step 2: Preprocessing and host word identification. Natural language preprocessing is performed on the Tibetan inscription text T, including word segmentation and stemming analysis, to accurately locate the host word H that needs to be added with class i case particles.

[0037] Step 3: Linguistic attribute extraction. For each host word H, the algorithm extracts its key grammatical features. The core is to identify and judge its "suffix" attribute. Based on the "Character Classification Method", the algorithm identifies and judges the attribute of its suffix, classifying the suffix as positive, negative or neutral, providing accurate input features for rule matching.

[0038] Step 4: Rule Activation and Matching. This step is crucial for the algorithm to achieve "historical adaptation". Based on the value of M.era, the corresponding historical rule subset is dynamically loaded. Before determination, the rule subset before determination is loaded, and after determination, the rule subset after determination is loaded. Then, the post-term attribute of the host word H is matched with the condition part, such as the IF part, of the loaded historical rule subset. The condition part of the historical rule subset is an IF-THEN production rule structure. Rule matching involves substituting the post-term attribute of the host word H into the IF part of the production rule structure for logical judgment.

[0039] Step 5: Decision-making and ambiguity resolution. Based on the rule matching results, the annotation decision is executed. If a unique result rule is matched, such as rule R1 or R2, the case particle form and grammatical function tag are directly output. If a candidate result set rule is matched, such as rule E1, the ambiguity resolution module is activated to adjudicate and then output. The adjudication process of the ambiguity resolution module is as follows: First, the preset inscription-specific variation dictionary is queried. If the dictionary can adjudicate the candidate result set, the annotation result is automatically output. If the dictionary does not cover the candidate result set, the test case is submitted to the manual collaboration interface for expert adjudication, and the expert adjudication result is fed back to the system to optimize the inscription-specific variation dictionary or the formal historical rule knowledge base.

[0040] Step Six: Structured Output. The annotation results are formatted into unified structured data units containing host words, case particles, grammatical functions, and rules. The collection of all units constitutes the complete i-type case particle annotation for the input text T. The i-type case particles include genitive case particles. and as case particles The diachronic rule subset is provided by a formalized diachronic rule knowledge base constructed based on the "Thirty Odes" and the "Word Genre Organization Method." Each rule in the diachronic rule subset is formally defined using a triplet structure: rule R = <Condition, Action, MetadataConstraint>. The condition is a logical expression that makes logical judgments based on the suffix attribute of the host word; the action is the output i-type case particle form and its grammatical function tag; and the metadata constraint specifies that the rule will only be activated and participate in the calculation if the input metadata, such as the inscription date, meets specific conditions. The rules in the formalized diachronic rule knowledge base are divided into a general rule subset, a pre-determination rule subset, and a post-determination rule subset. The general rule subset includes standard neutral mapping rules, standard masculine mapping rules, standard feminine mapping rules, and genitive derivation rules. The genitive derivation rules are obtained by adding a suffix to the labeled genitive form. The ergative form is obtained, and the rules for adding ergative particles are consistent with the rules for adding genitive particles. The formalized diachronic rule knowledge base also includes inscription-specific rules, which have higher priority than general rules and are used to handle diachronic variations and special grammatical uses of type i case particles in Tibetan inscriptions. The inscription-specific rules include the masculine rule for the Dazha Lugong Stele and the masculine special case rule for the Tang-Tibet Stele. The masculine rule for the Dazha Lugong Stele is to use a feminine particle after adding a masculine character in the Dazha Lugong Commemorative Stele. The rule for the positive exception of the Tang-Tibet Stele is that it applies to the addition of characters to the Tang-Tibet Alliance Stele. Follow-match genitive particle The rule is to mark it as a modal particle.

[0041] A rule-driven Tibetan inscription i-type case particle annotation system, used to execute a rule-driven Tibetan inscription i-type case particle annotation method, including:

[0042] The data input layer is used to support the import of inscription text and metadata;

[0043] The rule engine layer, a core module, embeds a formal rule base and executes the DRDA algorithm. The formal diachronic rule knowledge base in the rule engine layer is computable, interpretable, and scalable: the rules adopt an IF-THEN production structure to achieve computability, the annotation results can be traced back to the linguistic basis through rule ID to achieve interpretability, and the hierarchical storage structure supports incremental addition of new rules or addition of epigraphic exception dictionaries to achieve scalability.

[0044] The manual collaboration interface provides an entry point for manual adjudication for special instances or ambiguous situations not covered by the rule base, and its feedback can be used to iteratively optimize the rules.

[0045] The output layer is used to generate labeled corpora in a standard format.

[0046] The application of the rule-driven Tibetan inscription i-type case particle annotation method in the syntactic analysis of Tibetan inscriptions allows the structured output of the method to resolve syntactic structural ambiguities in Tibetan inscriptions, construct comparable Tibetan annotation data over time, and train or validate ancient Tibetan dependency syntax analysis and semantic role annotation models. This provides a data foundation for comparative studies of inscriptions before and after Tibetan text standardization and for studies on the evolution of Tibetan grammar.

[0047] Example 2

[0048] Definition and grammatical functions of class i case particles

[0049] Class i case particles refer to genitive and ergative particles, because both types of particles contain vowels. (represented by i), therefore they are collectively called i-type case particles; the Tibetan fundamental grammar, the Thirty Verses, was produced as early as the eighth century, in which the rules for adding i-type case particles were clearly pointed out and are still in use today. "; its meaning is the first of ten added characters. And the second Add the first postscript character The third The fifth one and the tenth Added later The seventh Add the seventh character after The rest are followed by words Added later Finally in Add vowels above This forms the genitive particle. It should be noted that, although not explicitly stated here... But it is with The usage and grammatical function of this are equivalent, and when followed by a word... Later, when it is necessary to add syllables (generally in poetic genres), or to connect the relationships between paragraphs, etc., additional syllables can be added. Additionally, according to the rules, a word is usually added after it. Add later (You can also add) ,For example (In the sky), however, in such cases, a word is usually added after it. Omitted, forming its abbreviation (In the heavens), for example, (Birds in the sky); Add a tenth suffix after the above five genitive particles. Can form As a case particle, its addition rules are the same as those for genitive particles;

[0050] Regarding added words Although it is attached to each base character, its addition and abbreviation methods have undergone certain development and changes. For example, in the ancient Tibetan documents written during the reign of the Tibetan king Tride Tsugten Moa Tsung, there is a passage from Dunhuang: (Then, in the sixth year...that is, the Year of the Rabbit...it was agreed to burn wood at the horn) etc., among which... All three were followed by... However, in modern writing, it is always included Omitted, written as (na,ste,du), its omission rule is, " This means that when a consonant is a single consonant letter, or when a consonant is accompanied by any of the following suffixes: r, l, s, i, u, e, o, or r, l, y, w, no suffix is ​​added. However, if the above situations do not apply, and a consonant is preceded by a syllable, then a suffix must be added. In fact, this is not only about The same applies to the other nine suffixes; otherwise, it would be easy to confuse the consonants with the suffixes, thus affecting the expression of meaning. This was standardized by translators Gawa Pelze, Jue Ru Luye Jianshen, and Xiang Yixi De during the reign of King Trizin Tsultrim Lodro, for the sake of writing convenience when the script was standardized.

[0051] Distribution and Comparison of Class i Case Particles in the Two Inscriptions

[0052] The inscriptions from the Tubo period record the events of that time in prose form, and belong to historical prose and narrative prose. Since the inscriptions were written by the ruling Tubo dynasty, the content is true and reliable, without any exaggeration or embellishment. The syntactic structure is rigorous, the word choice is appropriate and accurate, and the pronunciation system and word choice are relatively regular. They have high historical value and can be used as one of the standards for studying the language and script of that time.

[0053] Use Case Analysis of the Enlan Dazhalu Gong Memorial Monument

[0054] The Stele Commemorating the Merits and Achievements of Takdra Lukhong was erected in 764 AD by Trisong Detsen, the Tibetan king (reigned 755-797). It is a historical document. The stele commemorates the achievements of the minister Takdra Lukhong, and the king inscribed on it the immense honors bestowed upon him and his descendants, along with the important positions he held. This stele is the earliest existing Tubo stele and has significant historical value. "The background and social standing of Dazha Lugong were previously vaguely described in historical records, lacking detail. The inscription fills this gap and verifies the authenticity of certain historical materials." Although the grammatical system of the *Thirty Odes* existed at the time of the stele's erection, the method of adding genitive cases in the inscription differs from modern writing. Firstly, there are a total of 66 genitive case particles, of which 53 are consistent with the grammar of the *Thirty Odes*, and 13 are inconsistent, as shown in the table below:

[0055] Table 1:

[0056]

[0057] Note: The "Rule Consistency" column in this table refers to whether it is consistent with the rules for adding class i case particles in the grammar of the Thirty Odes;

[0058] In addition to the inscription outside, Later, it was also added. For example, appearing in rows 42-43 on the south side, (King Trisong Detsen was profound, decisive, and skillful in deliberation.) Added later The form; followed by a word Later, it was also added. For example, appearing in rows 38-39 on the north side, " (Regarding the father of Dazha Lugong, the sons and grandsons of the Grand Duke were numerous and all were granted the title of Shanglun Changshi). Nowadays, in such cases, the word "again" is generally omitted. Added later The inscription appears There is only one example, and it is inconsistent with the rule, appearing in rows 16-17-18 on the south side. (Lu Gongnai reported the facts of the rebellion of Mo Dongzebu and Lang Maishe to Prince Zanpu Trisong Detsen Shengcong), for Added later The form; Add all back, or All are added after no following word, and all are consistent with the rules; It did not appear in the inscription;

[0059] Secondly, there are a total of 14 ergative particles, of which 8 are consistent with the grammatical rules of the Thirty Odes, and 6 are inconsistent, as shown in the table below:

[0060] Table 2:

[0061]

[0062] Note: The "Rule Consistency" column in this table refers to whether it is consistent with the rules for adding class i case particles in the grammar of the Thirty Odes;

[0063] Consistent with the rules for adding genitive cases, in Added later ;Appear There is only one example, appearing in rows 4-5 on the east side. "(Already appointed as)" is followed by another word. Added later The same form appears in line 28 on the south side, " "But then another word was added." Added later In this situation, the form is now commonly omitted and followed by a word. Added later ; Add all Then, it conforms to the rules; and It did not appear in the inscription;

[0064] Example 3

[0065] Use Case Analysis of the Tang-Fan Alliance Stele

[0066] The Monument to the Tang-Tubo Alliance was erected in 821-822 by Trisong Detsen, the Tibetan king (reigned 815-838), and in 823. The "Alliance" inscription belongs to the historical document category. The Tang Dynasty and Tibet held eight alliance meetings throughout history. This stele was erected in 821 AD (the first year of the Changqing era of the Tang Dynasty) by the Tang emperors Trizin and Emperor Muzong, who decided to inscribe the content of the eighth alliance meeting. The original text reads: In 821 AD, envoys from both the Tang Dynasty and Tibet first swore an oath in Chang'an, the Tang capital; the following year, the Han Dynasty renewed the alliance in Lhasa, Tibet; in 823 AD, the alliance text was inscribed on a stele erected in front of the Jokhang Temple in Lhasa, known as the Changqing Nephew-Uncle Alliance Stele, also called the Tang-Tibet Peace Alliance Stele, or in Tibetan as Tsuglagkhang Dorje. It serves as a testament to the unity and friendship between the Han and Tibetan peoples. The method of adding class i case particles in the inscription differs from that of the Dazha Lugong Commemorative Stele. Firstly, there are a total of 152 class i case particles, of which 148 conform to the grammar of the "Thirty Verses," and 4 do not, as shown in the table below:

[0067] Table 3:

[0068]

[0069] Note: The "Rule Consistency" column in this table refers to whether it is consistent with the rules for adding class i case particles in the grammar of the Thirty Odes;

[0070] Unlike the Dazha Lugong Memorial Stele, this stele's inscription contains... Added later There is only one instance of deviating from the rules, appearing in the 15th row on the east side. (Although both can compete for victory on the battlefield), for Added later The form; Most of them were added later. However, there are still three instances of additions that do not conform to the rules, appearing in line 70 on the east side, and lines 71-72 and 73-74 on the left side of the west side. (The same inscription on the monument) (Both Han and non-Han rulers and their ministers jointly reported and made an oath) (The verification of the two gentlemen was done with official seals), for Added later or གྱྀ In this example sentence, the form is the same as the following sentence: The phrase “(the ministers who ascended the altar personally signed their names)” forms the only genitive particle in the entire inscription that does not indicate a genitive case meaning; it functions as a modal particle. Add all back, or All of them are added after no following word, and all conform to the rules; Not mentioned in the inscription, see the table below:

[0071] Table 4:

[0072]

[0073] Note: The "Rule Consistency" column in this table refers to whether it is consistent with the rules for adding class i case particles in the grammar of the Thirty Odes;

[0074] Unlike the Dazha Lugong Memorial Stele, the rules for adding erroneous characters are consistent with the grammatical system. Added later ; Added later ; Added later One of them is followed by a word; Added later The form appears in the 50th line on the east side. (They suddenly became enemy states) When adding Add after omission if there is no following word Indicates a case particle, It appears once in the inscription, on the 60th line on the east side. (The Tang court presided over the alliance)

[0075] Analysis of rule evolution from the perspective of historical linguistics

[0076] Judging from the construction dates of the Dazha Lugong Memorial Stele and the Tang-Tibet Alliance Stele, which are only 59 years apart, the rules for adding class i case particles have changed somewhat, mainly reflected in... and Added to; on the Dazha Road Commemorative Monument, and The following were all added This is because and They belong to the same group of consonants; in terms of place of articulation, they are retroflex consonants; in terms of manner of articulation, they are plosives. It belongs to the central vowel on the tongue surface; and letters in the same place of articulation or in the same group of consonants can be interchanged, which is one of the most distinctive grammatical features of the inscription.

[0077] Secondly, "the organizational structure of Tibetan script is explained from the perspective of the phonetic relationship between spelling." This is the content of the "Law on the Organization of Characters by Spelling." During the reign of King Tride Songtsen (798-815), Tibetan script underwent a standardization process, which standardized its structure and spelling rules. Furthermore, the rules for adding function words were specified based on the spelling of the syllable following the preceding syllable. "Spellings are generally divided into three categories: masculine characters belong to strong vowels, unaspirated voiceless sounds; feminine characters belong to weak vowels, unaspirated voiced sounds; and neutral characters belong to a combination of strong, weak, and neutral sounds." Therefore, the Tang-Tibet Alliance Stele, erected after this standardization, mentions... and The addition of was differentiated, because (da, ba, sa) are masculine words, and adding them after them creates masculine base words. , It is a feminine character, followed by a feminine base character. This follows the same principle as the same gender, and the rules for adding ergative particles are also the same.

[0078] Tibetan is derived from Sanskrit. Sanskrit grammar emphasizes gender, number, and case in word formation and sentence construction. Tibetan grammar also mentions determining the gender of a phrase or affix based on the gender of the suffix following the preceding syllable, which will not be elaborated here. The "Gender Organization Law" states that "the gender of the base word of a syllable used as a suffix, case particle, or function word must be consistent with the gender of the suffix of the preceding syllable; that is, a masculine suffix is ​​followed by a masculine base word; a feminine suffix is ​​followed by a feminine base word; and a neuter suffix is ​​followed by a neuter base word. This principle focuses on the harmony of pronunciation." Essentially, this refers to the same-gender principle, which effectively avoids the interchangeability of base words within the same consonant group, thus contributing to the refinement of function word addition rules and the uniformity of affixes in word formation and sentence construction. For example, in the word "summer,"... There are neutral words Add neutral text later and Neutral words Add positive character later The two writing styles are clearly different upon comparison. (dbyarkha) conforms to the principle of same gender, which is conducive to the uniformity of its affixes;

[0079] Another one is and It did not appear in the previous inscription, but appeared in the later inscription. It appeared once, and It is not explicitly stated in the grammar, so it is understandable that it does not appear or appears very rarely in the inscription, which is consistent with the rules of addition. It is just that as the grammar system developed, its additions became more and more numerous due to the needs. The addition rules also follow the same principle, with the last character added. It is a feminine character. and It is also a feminine character;

[0080] The three revisions of the Tibetan script throughout history have not only made the grammatical system more standardized and complete, but also promoted the standardization and normalization of the Tibetan script. They have played a positive role in the unification and popularization of the Tibetan script, as well as the development of Tibetan culture, and have been used to this day.

[0081] Based on the formal definition of the rules, Table 5 lists the core rules constructed for this paper; these rules clearly explain how the knowledge base uniformly handles "standard mapping" and "diachronic variation", as shown in the table below:

[0082] Table 5:

[0083]

[0084] Example 4

[0085] Based on the deductive verification of the above cases, due to the scarcity of annotated corpora of epigraphic documents, it is currently difficult to conduct large-scale data training and quantitative analysis. Therefore, this paper adopts the deductive method based on known linguistic analysis to demonstrate the feasibility of the method. We regard the above detailed analysis results of the two epigraphic documents as a high-quality "standard test set". By systematically demonstrating the process and results of the DRDA method in processing these test cases, we aim to verify its logical consistency and explore its explanatory potential for diachronic grammatical phenomena.

[0086] Inductive analysis of key use cases for the two monuments

[0087] To verify the logical consistency and diachronic interpretability of the DRDA method, we applied it to a carefully constructed "standard test set," consisting of key use cases selected from the Enlan Dazha Lugong Memorial Stele (before its revision) and the Tang-Tibet Alliance Stele (after its revision). The core logic and results of the deduction and verification are summarized in the table below:

[0088] Table 6:

[0089]

[0090] As shown in Table 6, through the deduction of four typical use cases, we can draw the following conclusions:

[0091] 1. Handling of mainstream rules: For the standardized positive rules (R1) after definition and the stable negative rules before and after definition (R2), the DRDA method can uniquely and accurately match and output the results, proving its basic ability to handle standard grammar.

[0092] 2. Explanation of diachronic variations: For the historical variations of "masculine characters followed by feminine particles" before the determination, DRDA activates the variation rules that allow candidate sets (R1) through the diachronic rule hierarchical mechanism, and completes disambiguation with the help of the inscription-specific mapping table; this verifies the core innovation of this method - through metadata-driven rule adaptation, it is compatible with rather than negates historical grammatical phenomena.

[0093] 3. Tolerance for special cases outside the rules: When encountering a very small number of grammatical special cases that cannot be covered by the rules (such as the usage of modal particles), the system can identify and record them through a manual collaboration interface; this proves the scalability and practicality of the DRDA framework and leaves an interface for the iterative optimization of the knowledge base.

[0094] Supporting syntactic analysis tasks

[0095] Accurate annotation of case particles in epigraphic documents is a crucial foundation for achieving deep syntactic and semantic analysis in Tibetan information processing. The DRDA method implemented in this paper can provide the following structured and computable core information for subsequent parsing:

[0096] 1. Resolve structural ambiguity: Clearly identify genitive (attributive marker) and ergative (agent marker), providing a prerequisite for unambiguous division of syntactic components such as subject, predicate, object, and attributive in a sentence;

[0097] 2. Ensure consistency over time: Adaptive annotation is performed based on the grammatical rules of different periods to generate comparable annotation data over time, providing a solid and reliable data foundation for comparative studies of inscriptions before and after the Tibetan script standardization and for language evolution analysis.

[0098] 3. Empowering downstream tasks: The high-quality, rule-driven case marker corpus generated can be directly used to train or validate models for dependency parsing and semantic role labeling in Old Tibetan. In the field of Old Tibetan where training data is scarce, such high-precision and interpretable basic annotation resources have irreplaceable value.

[0099] Based on the above embodiments, and using the fundamental Tibetan grammar *Thirty Verses* and *Character Organization Law* as a basis, this paper systematically reveals the systematic diachronic evolution of Tibetan i-type case particles before and after the standardization of Tibetan writing through a comparative study of the Enlan Dazha Lugong Memorial Stele and the Tang-Tibet Alliance Stele. On this basis, this paper proposes and elaborates on a rule-driven diachronic dynamic annotation method (DRDA). The theoretical innovation of this method lies in establishing the diachronic dimension of "inscription date" as the core metadata, and constructing a hierarchical and scalable diachronic rule knowledge base based on this, thus solving the problem of consistency and adaptability in cross-period document grammatical annotation within a theoretical framework. Through deductive verification based on typical use cases, the logical consistency of this framework and its explanatory potential for diachronic grammatical phenomena are confirmed.

[0100] Components not described in detail in this article are existing technologies.

[0101] While the specific embodiments of the present invention have been described in detail above, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention, and modifications or variations without creative effort are still within the protection scope of the present invention.

Claims

1. A rule-driven method for annotating type i case particles in Tibetan inscriptions, characterized in that, This method implements annotation through the Dynamic Rule Adaptation (DRDA) algorithm, specifically including the following steps: Step 1: Input and metadata parsing. Receive the inscription text T and its key metadata M. The metadata M includes the core control variable, the date attribute M.era, and the value range of M.era is {"before determination", "after determination"}. Step 2: Preprocessing and host word identification. Natural language preprocessing is performed on the Tibetan inscription text T, including word segmentation and stemming analysis, to accurately locate the host word H that needs to be added with class i case particles. Step 3: Linguistic attribute extraction. For each host word H, the algorithm extracts its key grammatical features. Step 4: Rule activation and matching. Based on the value of M.era, dynamically load the corresponding historical rule subset. Load the rule subset before determination and the rule subset after determination. Then match the suffix attribute of the host word H with the condition part of the loaded historical rule subset. Step 5: Decision and Ambiguity Resolution. Based on the rule matching results, the labeling decision is executed. If a unique result rule is matched, the case particle form and grammatical function tag are directly output. If a candidate result set rule is matched, the ambiguity resolution module is activated to make a decision and then output the result. Step Six: Structured Output. The annotation results are formatted into unified structured data units containing host words, case particles, grammatical functions, and rules. The collection of all units constitutes the complete i-type case particle annotation for the input text T. The i-type case particles include genitive case particles. (gi,kyi,gyi,'i,yi) and ergative particles (gis,kyis,gyis,'is,yis), the diachronic rule subset is provided by a formal diachronic rule knowledge base built on the Thirty Odes and the Word Organization Method.

2. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 1, characterized in that: Each rule in the historical rule subset is formally defined using a triplet structure: rule R = <Condition, Action, Metadata Constraint>. Here, the condition is a logical expression that makes logical judgments based on the suffix attribute of the host word, the action is the output i-type case particle form and its grammatical function label, and the metadata constraint specifies that the rule will only be activated and participate in the calculation if the input metadata, such as the date of the inscription, meets specific conditions.

3. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 2, characterized in that: The formalized diachronic rule knowledge base is divided into a general rule subset, a pre-determination rule subset, and a post-determination rule subset. The general rule subset includes standard neutral mapping rules, standard positive mapping rules, standard negative mapping rules, and genitive derivation rules. The genitive derivation rules are derived by adding a prefix to the already labeled genitive form. (sa) is in the ergative form, and the rules for adding ergative particles are the same as those for adding genitive particles.

4. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 1, characterized in that: The adjudication process of the ambiguity resolution module in step five is as follows: First, the preset inscription-specific variation dictionary is queried. If the dictionary can adjudicate the candidate result set, the annotation result is automatically output. If the dictionary does not cover the candidate result set, the test case is submitted to the manual collaboration interface for expert adjudication, and the expert adjudication result is fed back to the system to optimize the inscription-specific variation dictionary or the formalized historical rule knowledge base.

5. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 1, characterized in that: The formalized diachronic rule knowledge base also includes inscription-specific rules, which have a higher priority than general rules and are used to handle diachronic variations and special grammatical uses of type i case particles in Tibetan inscriptions.

6. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 5, characterized in that: The specific rules for inscriptions include the positive rule for the Dazha Lugong Stele and the positive special case rule for the Tang-Tibet Stele; the positive rule for the Dazha Lugong Stele is that after adding a positive character in the Dazha Lugong Commemorative Stele, a negative auxiliary word is always used. The rule for (gyi) is that the positive special case rule for the Tang-Tibet Stele is for the addition of characters to the Tang-Tibet Alliance Stele. (sa) is followed by a genitive case particle. (gyi) and marked as a modal particle.

7. The rule-driven Tibetan inscription i-type case particle annotation method according to claim 1, characterized in that: In step four, the condition part of the historical rule subset is an IF-THEN production rule structure, and the rule matching is to substitute the post-word attribute of the host word H into the IF part of the production rule structure for logical judgment.

8. A rule-driven Tibetan inscription i-type case particle annotation system, used to perform the method described in any one of claims 1-7, characterized in that: include: The data input layer is used to support the import of inscription text and metadata; The rules engine layer, the core module, includes an embedded formal rule library and executes the DRDA algorithm; The manual collaboration interface provides an entry point for manual adjudication for special instances or ambiguous situations not covered by the rule base, and its feedback can be used to iteratively optimize the rules. The output layer is used to generate labeled corpora in a standard format.

9. A rule-driven Tibetan inscription i-type case particle annotation system according to claim 8, characterized in that: The formalized diachronic rule knowledge base in the rule engine layer is computable, interpretable, and scalable: the rules adopt an IF-THEN production structure to achieve computability, the annotation results can be traced back to the linguistic basis through rule ID to achieve interpretability, and the hierarchical storage structure supports incremental addition of new rules or addition of epigraphic exception dictionaries to achieve scalability.

10. The application of the rule-driven Tibetan inscription i-type case particle annotation method according to any one of claims 1-7 in the syntactic analysis of Tibetan inscriptions, characterized in that, The structured output of the method is used to resolve syntactic ambiguities in Tibetan inscriptions, construct comparable Tibetan annotation data over time, and train or validate ancient Tibetan dependency syntax analysis and semantic role labeling models, providing a data foundation for comparative studies of inscriptions before and after Tibetan text standardization and for studies on the evolution of Tibetan grammar.

Images