A method for extracting Tibetan verb valence information based on a dependency syntax tree library

Abstract

The application discloses a Tibetan verb valence information extraction method based on a dependency syntax tree library, relates to the cross technical field of computational linguistics and natural language processing, and comprises the following steps: acquiring a Tibetan raw corpus; the source of the Tibetan raw corpus comprises literature, news, history, novels, teaching materials and a cloud Tibetan brat syntax tree library tool; preprocessing the Tibetan raw corpus to generate a dependency syntax tree library; analyzing the dependency syntax tree library to determine automatic verb valence information extraction rules; the automatic verb valence information extraction rules comprise zero-valence verb extraction rules, one-valence verb extraction rules, two-valence verb extraction rules and three-valence verb extraction rules; acquiring a Tibetan sentence to be extracted for verb valence information, and executing the automatic verb valence information extraction rules to obtain Tibetan verb valence information in the Tibetan sentence. The method improves the generation accuracy of Tibetan verb valence information.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of computational linguistics and natural language processing, and in particular to a method for extracting valence information of Tibetan verbs based on a dependency syntax treebank. Background Technology

[0002] Tibetan, as an important language in the Sino-Tibetan language family, possesses rich morphological variations and complex syntactic structures. The study of verb valency in Tibetan is of great significance for Tibetan natural language processing. Verb valency research in Tibetan is one of the core tasks of Tibetan dependency grammar analysis, aiming to determine the number of arguments governed by a verb in a sentence and its syntactic-semantic roles.

[0003] Extraction of verb valency information in Tibetan generally falls into two categories: rule-based methods and statistical methods. Early research primarily employed rule-based methods, but these suffer from significant errors in the manual annotation of verb valency information. Statistical methods utilize co-occurrence frequency information from Tibetan dependency treebanks to infer verb valency information; however, the corpora available for Tibetan dependency syntax research are relatively limited, far from sufficient to support the effective application of statistical methods, resulting in low accuracy in generating Tibetan verb valency information. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for extracting Tibetan verb valence information based on a dependency syntax treebank to address the aforementioned technical problems. This method improves the accuracy of generating Tibetan verb valence information.

[0005] The present invention adopts the following technical solution: This invention provides a method for extracting Tibetan verb valence information based on a dependency syntax treebank, comprising: Access Tibetan corpora; the sources of Tibetan corpora include literature, news, history, novels, textbooks, and the Yunzang Brat syntax library tool; The Tibetan language corpus is preprocessed to generate a dependency syntax tree library. The dependency syntax tree library is then analyzed to determine the rules for automatic extraction of verb valence information. The rules for automatic extraction of verb valence information include rules for zero-valence verbs, rules for one-valence verbs, rules for two-valence verbs, and rules for three-valence verbs. Obtain Tibetan sentences from which verb valency information is to be extracted, and execute the automatic extraction rules for verb valency information to obtain Tibetan verb valency information in the Tibetan sentences; Tibetan verb valency information includes zero-valence verbs, monovalent verbs, divalent verbs, and trivalent verbs.

[0006] Preferably, the Tibetan sentence from which verb valency information is to be extracted is obtained, and the automatic extraction rules for verb valency information are executed to obtain the Tibetan verb valency information in the Tibetan sentence, specifically including: When the entity component of a Tibetan sentence is a preposition and the syntactic relationship of the Tibetan sentence is a serial verb relationship, an adverbial-head relationship, a verb-complement relationship, a predicate-subject relationship, or a tense relationship, the verb in the Tibetan sentence is determined to be a zero-valence verb. When the entity component of a Tibetan sentence is a subject component and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship, or when the entity component of a Tibetan sentence is an object component and the syntactic relationship of the Tibetan sentence is an object-verb relationship or an object-verb relationship, the verb in the Tibetan sentence is determined to be a monovalent verb. When the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-verb-object, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-related, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is topic and description, the verb in the Tibetan sentence is determined to be a divalent verb. When a Tibetan sentence has a subject-verb-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

[0007] Preferably, preprocessing includes sentence segmentation, word segmentation, part-of-speech tagging, and dependency parsing; preprocessing the Tibetan corpus to generate a dependency parsing treebase specifically includes: The Tibetan language corpus was processed sequentially into sentence segmentation and word segmentation. Part-of-speech tagging is performed on the results of word segmentation. Dependency syntax relations are labeled based on the part-of-speech tagging results, and a dependency syntax treebase is generated.

[0008] Preferably, the rule for extracting zero-valent verbs is as follows: when the entity component of a Tibetan sentence is a preposition and the syntactic relationship of the Tibetan sentence is a serial verb relationship, an adverbial-head relationship, a verb-complement relationship, a predicate-subject relationship, or a tense relationship, the verb in the Tibetan sentence is determined to be a zero-valent verb.

[0009] Preferably, the rule for extracting monovalent verbs is as follows: when the entity component of a Tibetan sentence is a subject component and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship, or when the entity component of a Tibetan sentence is an object component and the syntactic relationship of the Tibetan sentence is an object-verb relationship or an object-verb relationship, the verb in the Tibetan sentence is determined to be a monovalent verb.

[0010] Preferably, the rule for extracting divalent verbs is as follows: when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-verb-object, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-related, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is topic-explanation, the verb in the Tibetan sentence is determined to be a divalent verb.

[0011] Preferably, the rule for extracting trivalent verbs is as follows: when a Tibetan sentence has a subject-verb-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

[0012] This invention provides a Tibetan verb valence information extraction device based on a dependency syntax treebank, comprising: The acquisition module is used to acquire Tibetan language corpora; the sources of Tibetan language corpora include literature, news, history, novels, textbooks, and the Yunzang Brat syntax library tool; The preprocessing module is used to preprocess the Tibetan language corpus, generate a dependency syntax tree, analyze the dependency syntax tree, and determine the automatic extraction rules for verb valence information. The automatic extraction rules for verb valence information include zero-valence verb extraction rules, one-valence verb extraction rules, two-valence verb extraction rules, and three-valence verb extraction rules. The extraction module is used to obtain Tibetan sentences from which verb valency information is to be extracted, and to execute the automatic extraction rules for verb valency information to obtain Tibetan verb valency information in the Tibetan sentences; Tibetan verb valency information includes zero-valence verbs, monovalent verbs, divalent verbs and trivalent verbs.

[0013] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for extracting Tibetan verb valence information based on a dependency syntax treebank.

[0014] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned method for extracting Tibetan verb valence information based on a dependency syntax tree library.

[0015] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects: This method acquires a raw Tibetan corpus covering Tibetan texts from different fields and periods, providing rich raw data for linguistic research. The corpus is preprocessed to generate a dependency syntax treebase. Analysis of the treebase determines automatic verb valence information extraction rules, and structured analysis of sentence component dependency relationships lays the foundation for subsequent verb valence research. Tibetan sentences from which verb valence information is to be extracted are obtained, and the automatic extraction rules are applied to obtain Tibetan verb valence information within these sentences. By executing the extraction rules, verb valence information can be automatically extracted from the sentences, reducing errors from manual annotation. This method improves the accuracy of generating Tibetan verb valence information. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0017] Figure 1 This invention provides a conceptual design for extracting valence information from Tibetan verbs. Figure 2 A flowchart illustrating a method for extracting Tibetan verb valence information based on a dependency syntax treebank provided by this invention; Figure 3 The corpus source diagram provided by this invention; Figure 4 The flowchart of the human-computer collaborative annotation mode for extracting Tibetan verb valency information provided by the present invention is shown below; Figure 5 This is a schematic diagram of the first example provided by the present invention; Figure 6 This is a schematic diagram of a second example provided by the present invention; Figure 7 This is a schematic diagram of the third example provided by the present invention; Figure 8 This is a schematic diagram of the fourth example provided by the present invention; Figure 9 A diagram representing Tibetan verb valency information provided by this invention; Figure 10 This is a schematic diagram of the fifth example provided by the present invention; Figure 11 This is a schematic diagram of the sixth example provided by the present invention; Figure 12 This is a schematic diagram of the seventh example provided by the present invention; Figure 13 Classification diagram of automatic extraction rules for monovalent verb information provided by this invention; Figure 14 This is a schematic diagram of the eighth example provided by the present invention; Figure 15 This is a schematic diagram of the ninth example provided by the present invention; Figure 16 This is a schematic diagram of the tenth example provided by the present invention; Figure 17 This is a schematic diagram of the eleventh example provided by the present invention; Figure 18 This is a schematic diagram of the twelfth example provided by the present invention; Figure 19 This is a schematic diagram of the thirteenth example provided by the present invention; Figure 20 This is a schematic diagram of the fourteenth example provided by the present invention; Figure 21This invention provides a classification diagram of rules for automatically extracting information from divalent verbs. Figure 22 This is a schematic diagram of the fifteenth example provided by the present invention; Figure 23 This is a schematic diagram of the sixteenth example provided by the present invention; Figure 24 This is a schematic diagram of the seventeenth example provided by the present invention; Figure 25 This invention provides a classification diagram of rules for the automatic extraction of trivalent verb information. Figure 26 This is a schematic diagram of the eighteenth example provided by the present invention; Figure 27 This is a framework diagram for the automatic extraction of Tibetan verb valency information provided by the present invention; Figure 28 This invention provides a word cloud diagram of the top 30 high-frequency zero-valent verbs extracted from a tree bank. Figure 29 This invention provides a word cloud diagram of the top 30 high-frequency Tibetan monovalent verbs extracted from a tree bank. Figure 30 The word cloud diagram of the first 30 Tibetan divalent verbs provided for this invention; Figure 31 The word cloud diagram of the first 30 Tibetan trivalent verbs provided for this invention; Figure 32 A schematic diagram of a Tibetan verb valence information extraction device based on a dependency syntax treebank provided by the present invention; Figure 33 A schematic diagram of a computer device for implementing a method for extracting Tibetan verb valence information based on a dependency syntax treebank, provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0019] Devices such as desktop computers, servers, and laptops are capable of executing the solutions of this invention. For ease of explanation, the following description will focus on servers as the executing entity.

[0020] In Natural Language Processing (NLP), Information Extraction (IE) and Information Retrieval (IR) are the two most important methods for mining textual information, and also the most practical and commercialized achievements of NLP. However, the two differ: "The former identifies and extracts information of interest to humans from documents and transforms it into a structured form, which can help achieve data fusion, tracking, and monitoring on large-scale datasets. The latter retrieves the most relevant subset of documents for users from unstructured document sets, or sorts the retrieved documents according to their relevance, as a response to the queries submitted by users." The earliest research on information extraction began in the mid-1960s, with Yale University's FRUMP...

[0021] Represented by Linguistic String at New York University, information extraction research and applications only began to flourish in the late 1980s, during which time the Message Understanding Conference (MUC) series was held. From 1987 to 1997, the MUC conference was held seven times. MUC established specific tasks and a rigorous evaluation system for information extraction, and proposed a complete information extraction scheme based on template filling mechanisms, including named entity recognition, coreference resolution, relation extraction, and event extraction.

[0022] The conference attracted researchers from all over the world, promoting a continuous stream of research findings on information extraction, both theoretically and technically. Therefore, MUC has made significant contributions to the field of information extraction in natural language processing.

[0023] Domestic research on information extraction started relatively late, initially focusing mainly on name recognition. With the support of the MUC conference, Chinese information extraction has made significant progress, moving towards more advanced stages such as relation extraction and event extraction. In recent years, the field of information extraction has shown a vibrant trend, with new advancements in both theory and application.

[0024] With the advent of the information age, information extraction has become an important branch of NLP (Natural Language Processing) and a hot topic and challenge in natural language processing research. Information extraction methods generally fall into two categories: rule-based methods and statistical methods. Early research primarily employed rule-based methods, but these have limitations, such as the complexity of manual compilation and low efficiency. Therefore, later research gradually shifted towards statistical methods. While statistical information extraction can compensate for the shortcomings of statistical methods, further research revealed that statistical methods are not perfect. Subsequently, a strategy combining rule-based and statistical methods was considered to find a more effective information extraction solution.

[0025] The fundamental element of dependency grammar is analyzing the dependency relationships between words in a sentence. Researchers have proposed three basic properties of dependency relationships: "A dependency relationship is a binary relation between two words; the two are usually asymmetrical, with one word being the dominant word and the other the subordinate word; and it is marked." Therefore, based on these three properties, the application of this function can make it possible to extract binary entity relations. Taking “ཁོ་བོའི་མགོ་བོ་སྨད།” (His head is down) as an example, in the above structure, “ཁོ་བོའི་མགོ་བོ” (His head) and “སྨད” (down) form a subject-verb relationship, while “ཁོ” (He) and “མགོ་བོ” (Head) form a modifier-head relationship. Relying on dependency grammar, the relationships between entities in the above natural language sentence can be effectively reflected. At the same time, it also allows for the identification of the roles played by each entity in the sentence and the extraction of verb valence information.

[0026] This invention is mainly based on the research of Tibetan verb valence information extraction from dependency syntax treebank. On the basis of dependency syntax treebank, extraction rules are formulated, and Tibetan verb valence information is extracted based on a combination of extraction rules and statistical methods. Figure 1 The conceptual design diagram for extracting valence information of Tibetan verbs provided by this invention is as follows: Figure 1 As shown, firstly, the subject and core predicate are extracted through syntactic role labeling; secondly, taking the core predicate as the starting point, the direct and indirect objects in the sentence are extracted through the analysis of dependency syntax relations; finally, if a sentence has only one subject and predicate, then the sentence is classified as a monovalent verb; if a sentence has both a subject and an object, then the sentence is classified as a divalent verb; if a sentence has not only a subject and an object, but also has two objects, then the sentence is classified as a trivalent verb.

[0027] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Figure 2This is a flowchart illustrating a method for extracting Tibetan verb valence information based on a dependency syntax treebank, as described in this invention. The method specifically includes the following steps: S201: Obtain a Tibetan corpus; the sources of the Tibetan corpus include literature, news, history, novels, textbooks, and the Yunnan Tibetan Brat syntax library tool.

[0029] This invention provides a formal description of Tibetan case grammar. It selects 10,000 sentences from multiple fields including literature, news, history, novels, and textbooks. Each sentence undergoes word segmentation, part-of-speech tagging, and syntactic-semantic annotation to form a Tibetan grammar tree. Furthermore, it selects 10,000 already-annotated sentences from different literary genres from the Yunzang BART grammar tree tool. Figure 3 The source diagram of the corpus provided by this invention.

[0030] S202: Preprocess the Tibetan language corpus to generate a dependency syntax tree, analyze the dependency syntax tree, and determine the automatic extraction rules for verb valence information; the automatic extraction rules for verb valence information include zero-valence verb extraction rules, one-valence verb extraction rules, two-valence verb extraction rules, and three-valence verb extraction rules.

[0031] In an exemplary embodiment, preprocessing includes sentence segmentation, word segmentation, part-of-speech tagging, and dependency parsing. Preprocessing the Tibetan raw corpus to generate a dependency parsing tree specifically includes: performing sentence segmentation and word segmentation on the Tibetan raw corpus in sequence; performing part-of-speech tagging on the results of word segmentation; and performing dependency parsing based on the results of part-of-speech tagging to generate a dependency parsing tree.

[0032] Specifically, this invention uses a Tibetan automatic word segmentation and part-of-speech tagging tool proposed by researchers to perform word segmentation and part-of-speech tagging, and then performs manual proofreading on this basis.

[0033] The dependency syntax tags used in this invention include: Head (HED), Subject (SBV), Topic (TPC), Object (IOB), Object Involved (VOB), Verb Serial (VV), Complement (CMP), Adverbial (ADV), Tense (TEN), Voice (MT), Attributive (ATT), Quantity (QUN), Prefix (LAD), Postfix (RAD), Clause (IC), General Punctuation (PUN), Coordination (COO), Coordination Shared (COS), Appositive (APP), and Sentence-wide Punctuation (PUS). The dependency syntax tags are shown in Table 1.

[0034] Table 1 Figure 4This invention provides a flowchart of a human-computer collaborative annotation model for extracting Tibetan verb valency information. The invention employs a human-computer collaborative annotation model for extracting verb valency information, and the main process is as follows: Figure 4 As shown. The specific process is as follows: ① Word segmentation is automatically completed by Chinese word segmentation and part-of-speech tagging, supplemented by a small amount of manual proofreading; ② Under the BART syntactic analysis tool, after automatic syntactic tagging, dependency syntax relations are manually tagged and corrected; ③ After downloading the tagged dependency syntax relation library from the BART automatic syntactic tagging tool, the dependency syntax tag library is analyzed to determine the rules for extracting verb valence information; ④ In the Python language environment, verb valence information is automatically extracted; ⑤ Under the Access data management software, verbs are manually corrected and classified according to different valence numbers.

[0035] S203: Obtain the Tibetan sentence from which verb valency information is to be extracted, and execute the automatic extraction rules for verb valency information to obtain the Tibetan verb valency information in the Tibetan sentence; the Tibetan verb valency information includes zero-valence verbs, monovalent verbs, divalent verbs, and trivalent verbs.

[0036] In an exemplary embodiment, a Tibetan sentence from which verb valency information is to be extracted is obtained, and an automatic verb valency information extraction rule is executed to obtain the Tibetan verb valency information in the Tibetan sentence. Specifically, this includes: when the entity component of the Tibetan sentence is a prepositional component and the syntactic relation of the Tibetan sentence is a serial verb relationship, adverbial-head relationship, predicate-complement relationship, predicate-subject relationship, or tense relationship, the verb in the Tibetan sentence is determined to be a zero-valence verb; when the entity component of the Tibetan sentence is a subject component and the syntactic relation of the Tibetan sentence is a subject-predicate relationship, or when the entity component of the Tibetan sentence is an object component and the syntactic relation of the Tibetan sentence is an intervening verb relationship... When a Tibetan sentence has a verb-object relationship or an object-verb-object relationship, the verb in the sentence is determined to be a monovalent verb. When the subject component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship or an object-verb-object relationship, or when the subject component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship or an object-object relationship, or when the subject component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is a topic-description relationship, the verb in the Tibetan sentence is determined to be a divalent verb. When a Tibetan sentence has a subject-predicate-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

[0037] Specifically, if there are two entity relationships in a sentence, called the subject-object relationship, relationship extraction is to study the relationship existing between the subject and the object in unstructured or semi-structured data and represent it as an entity relationship triple, that is, (subject, relationship, object). Sometimes the above relationship is formally represented as (subject, predicate, object), that is, (subject, predicate, object). Therefore, the triple relationship is sometimes also called the "SPO triple". For example: "ང་རང་ཟང་ཟིང་ལ་མོས། (Chinese: I like excitement.)", "ང་རང (Chinese: myself, subject, subject)" and "ཟང་ཟིང་། (excitement, object, object)" are two entities respectively, and "མོས" (like) is the relationship between the two entities. Therefore, "ང་རང" (myself), "མོས" (like) and "ཟང་ཟིང་" (excitement) are the subject (subject), predicate (predicate), and object (object) of a sentence respectively. Therefore, an SPO triple relationship is formed. In valency grammar, "ང་རང", "མོས་" and "ཟང་ཟིང་" are the "first actant", "second actant" and "core predicate" respectively. Therefore, the core verb "མོས།" belongs to a bivalent verb.

[0038] In the theory of valency grammar, "valency refers to the ability of a verb (or adjective, noun) to govern other types of words (or phrases, sentences)". Valency grammar includes two components: the governor and the governed. The governor refers to the component in the governing position, and the governed refers to the component in the governed position. Researchers believe that "a sentence is not only composed of some words, but more importantly, there is a relationship between these words". Therefore, the fundamental task of studying a sentence lies in studying the structure of the sentence, that is, the relationship in the sentence. The words that make up a sentence are in the leading position and the subordinate position respectively in terms of structure. One leading position can control several subordinate positions, and in principle, one subordinate position can only attach to one leading position. Researchers also compare "the clause of a verb to a play, which consists of performers and a scene. The protagonist among the performers is the verb, and the supporting roles are the complements, which are played by nouns or other equivalent components; the scene is the state description component, indicating time, place, manner, etc., and is played by adverbs or their equivalent components".

[0039] The fundamental principle of valency grammar is that a sentence is constructed around a predicate verb, connecting other components. Simply put, it's the dependency relationship between the predicate, subject, and object. For monovalent verbs, there is no subject-object dichotomy; if there is a subject, there is no object, and vice versa (only one entity). For divalent verbs, a verb can govern two nominal components (two entities), hence the term divalent verb, which, syntactically speaking, corresponds to subject-verb-object or topic structures. For trivalent verbs, a verb can govern three nominal components (three entities), which is the typical Tibetan "བྱ་བྱེད་ལས" (agent subject, involved object, and recipient object) sentence structure.

[0040] In an exemplary embodiment, the zero-valent verb extraction rule is as follows: when the entity component of a Tibetan sentence is a preposition and the syntactic relationship of the Tibetan sentence is a serial verb relationship, an adverbial-head relationship, a verb-complement relationship, a predicate-subject relationship, or a tense relationship, the verb in the Tibetan sentence is determined to be a zero-valent verb.

[0041] In an exemplary embodiment, the rule for extracting unvalent verbs is as follows: when the entity component of a Tibetan sentence is a subject component and the syntactic relation of the Tibetan sentence is a subject-predicate relation, or when the entity component of a Tibetan sentence is an object component and the syntactic relation of the Tibetan sentence is an object-verb relation or an object-verb relation, the verb in the Tibetan sentence is determined to be an unvalent verb.

[0042] In an exemplary embodiment, the rule for extracting divalent verbs is as follows: when the entity component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-verb-object, or when the entity component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-related, or when the entity component of a Tibetan sentence is the subject and the object component is the object, and the syntactic relationship of the Tibetan sentence is topic-description, the verb in the Tibetan sentence is determined to be a divalent verb.

[0043] In an exemplary embodiment, the rule for extracting trivalent verbs is: when a Tibetan sentence has a subject-verb-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

[0044] Since entities in a sentence will inevitably appear as a phrase in a dependency structure, this dependency relationship also reflects the characteristics of the relationship between the corresponding entities.

[0045] 1. In valency grammar, the monovalent verb is equivalent to the relationship between an entity and its attribute values ​​in relation extraction. Figure 5 This is a schematic diagram of the first example provided by the present invention. Figure 5 The Chinese equivalent of the Tibetan phrase is: "Guided by the Publicity Department of Haixi Prefecture".

[0046] All the words in the sentence are connected by dependency relations. For example, in the sentence, “མཚོ་ནུབ་ཁུལ་ཨུའི་དྲིལ་སྒྲོག་པུའུ” (Haixi Prefecture Propaganda Department) is a subordinate noun, forming a dependency relation through the subject and the governing verb “མཛུབ་ཁྲིད” (guide). Therefore, the relation between them is an asymmetrical binary relation, i.e., a subject-verb relation. The syntactic relation of this example is expressed as:

[0047] Sub+bo+hed+com The dependency syntax relation is written as: SBV [Sub+bo+hed] + CMP [com].

[0048] 2. In valency grammar, the divalent verb is equivalent to an entity relation in relation extraction. There are two entities involved; this invention refers to these two entities as the subject and the object. Relation extraction involves identifying the relationship between the subject and the object in unstructured or semi-structured data and representing it as a triple entity relation, which is a dependency relationship between entities. Figure 6 This is a schematic diagram of the second example provided by the present invention. Figure 6 The Chinese equivalent of the Tibetan phrase is: "Wanma Sanzhi singer sings".

[0049] All the words in the sentence are connected by syntactic dependency relations. For example, in the sentence, “གླུ་བ་པད་མ་བསམ་འགྲུབ།” (Wanma Sanzhi) forms a dependency relation through the subject (sub) and “བཏང” (sing) (predicate verb, v), and “བཏང” (predicate verb, v) also forms an object dependency relation with “གཞས” (sing). Therefore, the relationship between them is a subject-object-verb structure (a triplet entity relation), i.e., (Subject, Verb, Object). Here, the object is the object of interest in Tibetan grammar, so the syntactic relation of this example is expressed as:

[0050] Sub+bo+Obj + hed The dependency syntax relation is written as: SBV+VOB+HED= SBV[Sub+bo] + VOB[obj]+HED.

[0051] Figure 7 This is a schematic diagram of the third example provided by the present invention. Figure 7 The Chinese equivalent of the Tibetan phrase is: He studied Tibetan medical theory, such as... Figure 7As shown, all the words in the sentence are connected by syntactic dependency relations. For example, in the sentence, "ཁོ" (he) is a subordinate noun, forming a dependency relationship with the noun "སློབ་གཉེར" (learn) through the subject (sub). "སློབ་གཉེར" also forms an object dependency relationship with the subordinate noun "བོད་ཀྱི་གསོ་བ་རིག་པ" (Tibetan medicine). Therefore, the relationship between them is also a subject-object-verb structure consisting of triplets, i.e., (Subject, Verb, Indirect-object). The object in this question is the objective object as defined in Tibetan grammar, followed by a case particle. Therefore, the syntactic relation table for this question is as follows:

[0052] Sub+bo+atr+gi+obj+ls+hed+com Dependency relations are written as: SBV+ATT+VOB+HED= SBV[Sub+bo]+ATT[art+gi]+VOB[obj]+HED[hed]+CMP[com] In valency grammar, a trivalent verb is equivalent to the extraction of entity relations from three entities in information extraction: the subject, the object involved, and the object being extracted (subject, object 1, object 2). According to valency theory, a core verb governs these three nominal components (entities), thus defining the dependency relationships between these entities.

[0053] Figure 8 This is a schematic diagram of the fourth example provided by the present invention. Figure 8 The Chinese equivalent of the Tibetan phrase is: I have bad news for you.

[0054] The noun “ང” (I) is a subordinate noun, forming a dependency relationship with the noun “བཤད” (tell, say) through the noun “sub” (subject). “བཤད” forms a dependency relationship with “ཁྱོད” (you) as obj2 (indirect object or object of speech), and the noun “བཤད” forms a dependency relationship with the noun “obj1” (direct object or object of interest). Therefore, the relationship between them is a subject-object-verb structure (with two objects), namely (Subject, Verb, Object1, Object2).

[0055] In Tibetan grammar, "Object1" is the object in question, and "Object2" is the object of the question. Therefore, the syntactic relation table for this question is: sub+bo+obj2+ls+obj1+rz+hed Dependency relations can be written as: SBV+IOB+VOB+HED=SBV[sub+bo]+IOB[obj2+ls]+VOB[obj1]+RAD[rz]+HED It can be seen that dependency grammar and entity relation extraction are interdependent, that is, dependency relation analysis can reveal syntactic structure by analyzing the dependency relations between various linguistic units. For example, sub (subject) and verb (verb) form a subject-verb relation, and obj (object) and verb (verb) form an object-verb relation. By identifying the grammatical components such as subject, object, predicate, attributive, adverbial, and complement in the sentence, and analyzing the grammatical relations between these components. Entity relation extraction also refers to the (binary) relationship between two entities

[192] , that is, identifying the subject, object, adverbial, etc. in the sentence. Entity relation extraction can be divided into binary, ternary, and multi-entity relations according to the entity index involved. Binary relation extraction refers to extracting the relationship between an entity and its attribute value. From the perspective of dependency grammar, dependency relation is also an asymmetric binary relation, equivalent to a valency grammar's monovalent verb. Triple entity relation extraction refers to extracting the relationship between two entities and entity relation words, equivalent to a valency relation's divalent verb. Multi-entity relation extraction refers to extracting the relationships between three or more entities, which is equivalent to the trivalent or quadvalent valency relation, etc.

[0056] In summary, valency relations are binary symmetric relations between verbs and the constituents they govern. The head word of a sentence is usually a verb, and all other words either depend on the head word or are connected to it via a path. Researchers have proposed the concepts of "complement" and "descriptive word," where complements correspond to the subject and object in entity relation extraction, while descriptive words correspond to adverbs, attributives, and complements in entity relation extraction. The valency core word, the verb, corresponds to the entity relation attribute value in entity relation extraction.

[0057] Since entities in a sentence will always appear as phrases within dependency structures, this dependency relationship also reflects the characteristics of the relationship between the corresponding entities. Therefore, it can be determined that valency and entity relationships have consistent research objects.

[0058] Figure 9 To characterize the multivariate entity group extraction model, the present invention provides a Tibetan verb valence information representation graph. The invention analyzes the dependency relations of the syntactic treebase, and through analysis, it is found that the dependency relations between entity words and relation words are as follows: Figure 9 As shown, it mainly includes the following types: (1) Valence: zero-valence verbs; Substantive components: prepositional components; Syntactic relations: serial verb relations, adverbial-head relations, predicate-subject relations, verb-complement relations.

[0059] (2) Valence: monovalent verb; Substance relation: subject component; Syntactic relation: subject-predicate relation.

[0060] (3) Valence: monovalent verbs; Entity relation: object component; Syntactic relation: object-verb relation involving the subject, object-verb relation involving the object.

[0061] (4) Valence: divalent verb; Substance component: subject-object component; Syntactic relationship: subject-verb-object relationship.

[0062] (5) Valency: trivalent verb; Substances: subject and object; Syntactic relationship: subject-verb-object relationship (double object).

[0063] Specifically, in this invention, noun phrases in a sentence are considered as candidate entity words, and the identification of candidate entity words is equivalent to the identification of noun phrases. For example, “ཡོངས་འཛིན་བོད་ཡིག” (Yunnan Tibetan script) is a noun phrase, and this invention also considers it as a candidate entity word. Treating noun phrases as entity units and identifying entities based on noun phrases avoids extracting only local entities like “ཡོངས་འཛིན་བོད་ཡིག” (Yunnan Tibetan script) instead of the whole entity, such as “ཡོངས་འཛིན།” (Yunnan Tibetan script) and “བོད་ཡིག” (Tibetan script). In entity triple relation extraction, to ensure entity integrity, firstly… Noun phrases are identified and labeled, and the identified noun phrases are designated as candidate entity words. Verbs in the sentence are considered candidate relation words. Six types of relations—subject-verb (SBV), verb-object (VOB), object-object (IOB), verb-verb-verb (VV), adverbial-head (ADV), and verb-complement (CMP)—are considered as dependency relation sets. If a candidate entity word and a candidate relation word have a reachable dependency relation path, and the path includes one of the relations from the dependency relation set, then this candidate entity word is an entity word of that candidate relation word. In valency grammar theory, this verb (candidate relation word) is called a monovalent verb; a verb with only one noun phrase is a monovalent verb. If a candidate relation word includes two or more entity words, it can form an entity relation group. Among them, if it contains two entity words, it is a two-entity relation group; if it contains multiple entity words, it is a multi-entity relation group. For example: “སྒྲོལ་དཀར་_nr གྱིས་_bo སློབ་གྲོགས་_nn རྣམས་_qj ལ་_lsསྒྲུང་_nn བཤད_vt །_ww” (in Chinese: Zhuoga tells stories to her classmates).

[0064] Candidate entity words: སྒྲོལ་དཀར། (Zhuoga) སློབ་གྲོགས་རྣམས། (Students) སྒྲུང་། (Story).

[0065] Candidate relational word: "བཤད" (to speak).

[0066] Dependency paths that comply with the rules for candidate entity words and relation words: HED (བཤད); SBV (སྒྲོལ་དཀར། Zhuoga, བཤད། speak); VOB (སྒྲུང་། story, བཤད། speak); IOB (སློབ་གྲོགས་རྣམས། classmates, བཤད། speak) Candidate multi-entity relation group: བཤད། speak <སྒྲོལ་དཀར། Zhuoga, སྒྲུང་། story, སློབ་གྲོགས་རྣམས། classmates>.

[0067] Due to the differences in Tibetan sentence structure and verb valence information, the present invention proposes the following extraction rules for verb valence information: Automatic extraction of zero-valent verb information: A zero-valent verb refers to a verb that cannot govern any nominal components. In traditional Tibetan grammar, zero-valent verbs are generally classified as the same case form of locative particles (i.e., དེ་ཉིད), without action elements, and those with only state elements are zero-valent verbs. The division of action elements and state elements in valency grammar is equivalent to the opposition between subject and object and attributive, adverbial, and complement in traditional grammar.

[0068] In Tibetan, the instrumental case particle is used after a noun or noun phrase to form an instrumental case particle structure, which modifies various adverbials of the predicate verb in a sentence. When a transitive verb that can generally take an affected object is used as the predicate, it mostly indicates the tools, materials, reasons, and manners used in the action, equivalent to "with", "by", and the structural particle "~地" in Chinese. When an intransitive verb is used as the predicate, it mostly indicates the state, manner of the action, and the reasons or conditions for促成 the result. Equivalent to "~地", "due to...".

[0069] Therefore, the present invention classifies prepositional components in sentence components: verb-verb relation, adverbial-head relation, predicate-same-body relation, and complement relation as zero-valent verbs. When syntactically annotating, it is marked as follows:

[0070] Verb-verb relation (བྱ་བ་སྔ་ཕྱི་མཚམས་སྦྱོར།) Verb-verb VV Adverbial-head relation (བྱ་བ་དང་བྱེད་སྟངས་ཀྱི་འབྲེལ་བ།) Adverbial ADV Complement relation (བྱ་བ་དང་ཁ་གསབ་ཀྱི་འབྲེལ་བ)། Complement CMP Verb-verb relation (Verb-verb, VV), when the central word of a sentence is a verb, if there are other verbs behind it indicating a series of consecutive actions, the dependency relationship between the central word and these words is a verb-verb relation, marked with "VV". For example:

[0071] Figure 10 Schematic diagram of the fifth example provided for the present invention, Figure 10The Chinese corresponding to the Tibetan is: We gathered there carelessly.

[0072] When concatenated with the verb "འཛོམས" (gather) to represent the action and behavior of the agent, the dependent relationship between the central word "སྣང་བ་གཡེང་བ" (carelessly) and the verb "འཛོམས" (gather) is a serial verb relationship, marked with "VV". In the valence grammar theory, a verb without an argument is a zero-valent verb, and this serial verb relationship is classified as a state argument. Therefore, the present invention treats the serial verb relationship as a zero-valent verb.

[0073] Adverbial relationship (Adverbial, ADV). The instrumental case particle is used after a noun or noun phrase to form an instrumental case particle structure, which acts as an adverbial modifying the predicate, indicating the thing used in the action, that is, tools, organs, materials, means, etc. It is generally used before the object and after the subject. The locative case particle is used after a time noun, a locative noun, or an adverbial phrase to form a locative case particle structure, and acts as an adverbial of location, time, purpose, scope, degree, etc. in the sentence according to the combination relationship with the predicate verb of different properties behind it. The adverbial is generally before the predicate, and the dependent relationship is ADV. Generally, except for negative adverbs, the rest of the adverbs are in an adverbial relationship when they are before the predicate. For example:

[0074] Figure 11 This is the schematic diagram of the sixth example provided by the present invention. Figure 11 The Chinese corresponding to the Tibetan is: Good fortune is like fire.

[0075] In the example, the relationship between "མེ་ལྟར" (like fire) and "འབར" (burn) is marked as "ADV". "མེ་ལྟར" is the adverbial of the core verb "འབར", that is, "མེ་ལྟར" modifies the predicate "འབར". If there is no subject "བསོད་ནམས" (good fortune), this question belongs to zero valence. Since the valence grammar does not count the state argument into the valence of the verb, the example is classified as a monovalent verb.

[0076] Complement relationship (Complement, CMP). The complement is a supplementary component of the predicate, which is the result of the change of the action and behavior and plays a supplementary role in explaining the predicate. The auxiliary verb or particle closely following the verb (such as metaphorical particle, tense particle, mood particle, wish particle, ending particle, judgment verb and existential verb after the predicate, etc.) has a certain meaning, but not completely. It closely follows the main verb to assist the main verb in explaining the result, degree, tendency, possibility, state, mood, etc. of the action and behavior. Note: Sometimes the auxiliary verb includes a modal auxiliary verb, such as: "འདོད" (desire, longing) in the example expresses mood.

[0077] Figure 12 This is the schematic diagram of the seventh example provided by the present invention. Figure 12 The Chinese corresponding to the Tibetan is: Aishu turned off the light.

[0078] In the example, "ཨེ་ཧྲབ" (Aishu) and "གློག་སྒྲོན" (lamp) are respectively the agent subject and the involved object of the sentence, and are the action elements of the core verb "གཟིམས" (close). "བྱུང་" (le) is the complement of the verb "གཟིམས", and the relationship is "CMP", which is a state element. Therefore, the example belongs to a bivalent verb. So, the relationship between "གཟིམས" and "བྱུང་།" in the example is a verb-complement relationship. According to valence grammar, it is classified as a state element. The zero-valent verb information extraction rule tag set is shown in Table 2.

[0079] Table 2 Automatic extraction of monovalent verb information.

[0080] A monovalent verb refers to a verb that has only one action element, that is, a verb that can govern one NP is called a monovalent verb. According to the characteristics of Tibetan grammar, a zero-valent word plus a subject (that is, adding one action element) can be classified as a monovalent verb.

[0081] From the perspective of syntactic structure, the information structure of a monovalent verb is equivalent to the subject-predicate relationship and the verb-object relationship. In Tibetan sentences, the subject-predicate relationship (Subject-verb, SBV) is composed of the subject and predicate that represent the object and description. Opposite to the predicate, usually, the core corresponding to the subject component is the predicate of the sentence. A typical subject is composed of a noun or a pronominal structure. The subject is mostly marked with the "instrumental case particle". The instrumental case particle is used after a noun, pronoun, or noun phrase to form an instrumental case structure, and is used as the subject of a transitive verb predicate sentence that can take an involved object.

[0082] The subject in Tibetan can be divided into an agent subject and a patient subject (the action described by the predicate is not emitted from the subject, but refers to the party that is承受 by the subject, that is, the subject is not the agent, but the patient, so it is a patient subject). For example: "ང་ལ་འཛུམ་ཆུང་ཞིག་བསྟན་བྱུང་།" (Chinese: Smile at me), where "ང" (I) is the patient subject, and the agent subject in the sentence is omitted, and "ང" (I) becomes the patient subject.

[0083] In Tibetan grammar, the object is divided into two types. One is that the object represents the thing involved in the action. In Tibetan, it is "བྱ་བ་དང་འབྲེལ་བའི་ལས་ཀྱི་དངོས་པོ།", and this invention calls it "involved object (ལས་ཀྱི་དངོས་པོ།)", and no case particle is added after it. The other is that the object represents the action of the agent and the object it points to. This invention calls it "object object (བྱ་བའི་ཡུལ།)", and the locative case particle must be added after it.

[0084] Figure 13 The classification diagram of the automatic extraction rules for monovalent verb information provided by this invention is as follows: Figure 13 As shown, this invention uses five rules to extract information from monovalent verbs: 1. Subject-verb relations are marked with SBV: Tibetan sentences mostly consist of two parts: a subject and a predicate. The subject is the object of the predicate's description, and the predicate describes the subject; there is a declarative relationship between the two. Semantically, subjects are generally divided into three categories: agentic subjects, patient subjects, and neutral subjects. Agentic subjects indicate the doer of the action, marked with an ergative case. Patient subjects indicate the action described by the predicate, but not originating from the doer of the subject; rather, they originate from the receiver of the action. In other words, the subject is not the agent, but the receiver. Neutral subjects represent things unrelated to the agent or receiver, such as objects or instruments. In other words, they represent the object of description, judgment, or explanation. Examples of these three types of subjects are given below:

[0085] (1) Subject-verb relationship of agent The agent-subject relationship is generally marked with SBV: In Tibetan, the agent-subject indicates the doer of the action, and the ergative sign (བྱེད་སྒྲའི་ཕྲད།) is added after the subject. For example: Figure 14 This is a schematic diagram of the eighth example provided by the present invention. Figure 14 The Chinese equivalent of the Tibetan phrase is: I don't know.

[0086] In the example, "ང" (I) is the agent subject, "ས" is the ergative marker, "ཤེས" (to know) is the core verb, and "མི" is the negation particle, which is the prefix of the verb "ཤེས". The verb "ཤེས" governs one action element, which is "ང". Therefore, this question belongs to the category of monovalent verbs.

[0087] (2) Neutral subject-verb relation A neuter subject is neither the agent nor the patient, but rather the object being described, explained, or judged; some grammar books call it a "relative subject." A linking verb connects the relative subject and the content being described, thus forming a sentence. In Tibetan, the predicate of a neuter subject is generally an intransitive verb, and no case marker is added after the verb subject, indicating the object being described, judged, or explained. For example: Figure 15 This is a schematic diagram of the ninth example provided by the present invention. Figure 15 The Chinese equivalent of the Tibetan phrase is: I'm awake.

[0088] In the example, "ང" (I) is the neutral subject, the object being described; "སད" (wake up) is the predicate, the core verb of the sentence; and "བྱུང" (already) is the aspect marker. The core verb "སད" governs an action element, which is "ང". Therefore, this question involves a monovalent verb. For example: Figure 16 This is a schematic diagram of the tenth example provided by the present invention. Figure 16 The Chinese equivalent of the Tibetan phrase is: The weather was very good that day.

[0089] In the example, “ལེགས” (good) is the core verb, “གནམ་གཤིས” (weather) is the neutral subject, which is the object being described, “ཉིན་མོ” (day) is an adverb of time, “དེ” (that) is the suffix of the adverb of time “ཉིན་མོ” (day), “ར” is the time case marker, which follows the adverb of time, and “ཧ་ཅང” (very) is an adverb of degree, which modifies the core predicate “ལེགས” (good). From a valency perspective, the core predicate "ལེགས" (good) governs an action element, which is the neutral subject "གནམ་གཤིས" (weather). The stative element is not included in the valency of the verb. Therefore, this question is classified as a monovalent verb.

[0090] (3) Subject-verb relationship of the object In Tibetan sentences, the object of an action is the recipient of the action, not the agent or performer of the action; that is, the person or thing to which the action is directed. For example: Figure 17 This is a schematic diagram of the eleventh example provided by the present invention. Figure 17 The Tibetan equivalent of this phrase in Chinese is: "defeated the opponent".

[0091] In the example, “འགྲན་ཡ” (opponent) is the patient subject, the receiver of the action; “ཆམ་ཕབ” (defeat) is the core verb; and “བཏང” (finished) is the auxiliary verb. “ལ” is the occult particle marker, immediately following the patient subject. The relationship between the patient subject “འགྲན་ཡ” (opponent) and the core verb “ཆམ་ཕབ” (defeat) is a subject-verb relationship, marked with SBV. The relationship between the core verb “ཆམ་ཕབ” (defeat) and the auxiliary verb “བཏང” (finished) is a verb-complement relationship, marked with CMP. From the perspective of valency grammar, the verb "ཆམ་ཕབ" (defeat) in this question governs an action element, which is the object subject "འགྲན་ཡ" (opponent). The verb-complement relationship belongs to the stative element, and in valency grammar, the stative element is not included in the valency of the verb. Therefore, this question is classified as a monovalent verb.

[0092] 2. Object-Verb relationship, generally marked with VOB and IOB: In Tibetan grammar, objects are divided into verbs of action (VOB) and verbs of object (IOB). Verbs of action refer to the things involved in the action and are in the zero-form locative case. Verbs of object refer to the object to which the action is directed and must be followed by a locative particle.

[0093] (1) Object-Verb relationship, marked with VOB. For example: Figure 18 This is a schematic diagram of the twelfth example provided by the present invention. Figure 18 The Chinese equivalent of the Tibetan phrase is: Read this letter.

[0094] In the example, “འཕྲིན་པ་འདི” (this letter) is the object of the action, which can generally be marked as the patient subject. There is an agent subject in meaning, but it is omitted due to context. “བཀླགས” (read) is the core verb, governing an action element, which is “འཕྲིན་པ་འདི” (this letter). Therefore, this example belongs to the category of monovalent verbs.

[0095] (2) Object-verb relationship (Indirect-object) Generally, a noun, pronoun, or noun phrase is used as the object of a verb, indicating an action that extends to a specific object. It must be followed by a case particle. In this invention, this is marked with IOB in the rules. For example: Figure 19 This is a schematic diagram of the thirteenth example provided by the present invention. Figure 19 The Tibetan equivalent of this phrase in Chinese is: He has arrived home.

[0096] In the example, “ཁོ་རང” (he) and “ཡུལ” (home) are the agent subject and object of the sentence, respectively. “ལ” is the locative marker following the object. The core verb is “ཐོན” (to arrive), which governs two action elements, “ཁོ་རང་” (he) and “ཡུལ” (home). It is a divalent verb. The relationship between “ཡུལ” and “ཐོན” (to arrive) is an “object-verb relationship”, marked with “IOB”. Therefore, “ཐོན” governs “ཡུལ”, which is a monovalent verb within the valency range.

[0097] Figure 20 This is a schematic diagram of the fourteenth example provided by the present invention. Figure 20 The Chinese equivalent of the Tibetan phrase is: Thank you.

[0098] In the example, the core verb is “བཀའ་དྲིན་ཆེ” (thank you). It has an action element, “ཁྱོད” (you) (patient subject). In fact, it is an object, but because the agent subject is omitted in the context, semantically, “ཁྱོད” (you) is the object to which the action expressed by the predicate verb “བཀའ་དྲིན་ཆེ” is directed, that is, the object component.

[0099] The set of relation tags for extracting information from monovalent verbs is shown in Table 3.

[0100] Table 3 Information on divalent verbs is automatically extracted.

[0101] A divalent verb is a verb with only two action elements that can combine with two nouns. It generally has a subject and an object, with the action elements being the agent and patient. In Tibetan, both transitive and intransitive verbs can function as divalent verbs. A typical Tibetan divalent verb semantically governs two action elements, meaning it takes two essential nominal elements: the agent and the patient, and syntactically functions as the subject and object.

[0102] This invention extracts valence information from verbs into three different subject-verb-object sentence structures: 1. Subject-object-verb structure with a transitive verb containing an object as the predicate (i.e., "subject-object relationship"). ི་འབྲེལ་བ།(Subject-Object Relationship))); II. Subject-Object-Verb Structure with Transitive Verbs as Predicates Without Objects (i.e., “Subject-Object Relationship”) (བྱེད་པོ་དང་བྱ་བའི་ཡུལ ་གྱི་འབྲེལ་བ།(Subject-Object Relationship))); III. The topic-explanation relationship in sentences with the "referential conjunction ནི", that is, the topic is a functional component at the level of discourse structure, with the characteristics of a subject, and is played by nouns or pronouns. Its core is generally a copula as the predicate. The explanation is the corresponding topic, with the characteristics of explaining the subject. It is also played by nouns or pronouns, and its core predicate verb is generally a copula, often following the "referential conjunction ནི".

[0103] From the perspective of valency grammar, a divalent verb is a verb that has only two action elements and can combine with two nouns. It generally has a subject and an object, and its action elements are composed of agents and patients. Therefore, the above sentence patterns all require two action elements, thus falling under the information rules of divalent verbs. Figure 21 This is a classification diagram of the automatic extraction rules for divalent verb information provided by the present invention.

[0104] 1. Subject + Indirect-object + Particle + Verb (Example: Subject, Indirect-object, Verb) Figure 22 This is a schematic diagram of the fifteenth example provided by the present invention. Figure 22 The Chinese equivalent of the Tibetan phrase is: Deji Lhamo calmly returned home.

[0105] In the example, the core verb "ལོག" (return) has two action elements: the first action element, the agentic subject "བདེ་སྐྱིད་ལྷ་མོ" (Dejram), and the second action element, the object "རང་ཁྱིམ" (one's own). "དལ་ལྷོད་ལྷོད" (unhurried) is an adverbial of manner, and "ལ" is a locative marker added to the object. In valency grammar, adverbials are equivalent to stative elements and are not included in the valency number. Therefore, the example belongs to the divalent verb category. In this invention, subject-object-verb structures (subject, object, and predicate) in the rule base are categorized into a binary verb information base and marked with SVI, where S is the subject, V is the predicate verb, and I is the indirect object or object.

[0106] 2. Subject + Object + Verb (Subject, object, Verb) For example: Figure 23 This is a schematic diagram of the sixteenth example provided by the present invention. Figure 23 The Tibetan translation of the Chinese phrase is: I wrote a novel about my hometown.

[0107] In the example, the core verb is "བྲིས།" (to write), which governs two action elements: the first is "ང" (I), and the second is "སྒྲུང་གཏམ" (novel). "ས" is the ergative marker, and "ཕ་ཡུལ་སྐོར" (about hometown) is the attributive modifier of the object. The relationship between them is a possessive-attributive-head relationship. "གྱི" is the genitive marker, indicating the possessive relationship between the attributive modifier and the head noun. From a valency grammar perspective, a core verb that governs two action elements is a divalent verb; therefore, this example uses a divalent verb. In this invention, subject-object-verb structures (subject, object, and predicate) in the valence information extraction rule base are categorized into the divalent verb information rule base and marked with SVI, where S is the subject, V is the predicate verb, and O is the direct object or object.

[0108] 3. Topic (subject) + ནི + Description (object) + Comment Verb (topic comment verb) In linguistics, "topic" is sometimes called "topic" and "comment" is sometimes called "comment". Terms such as “language,” “commentary,” “review,” and “topic” are used. The concept of “topic and description” was first proposed by researchers in their 1921 book, *Language*, which stated that “a sentence… is the linguistic expression of a proposition. It combines the speaker’s subject with the statement of the topic.”

[0109] In the field of Chinese linguistics, Zhao Yuanren was the first scholar to apply the concepts of topic and description to Chinese language research. He pointed out that "in Chinese, the grammatical relationship between subject and predicate is less a relationship between agent and action, and more a relationship between topic and description; agent and action can be seen as a special case of topic and description." Previously, there was considerable debate in linguistics regarding the understanding of topic in terms of its nature or grammatical status. Summarizing various long-standing views on topic, there are mainly two: one view holds that topic and subject are combined into one, and all topics are subjects. The other view holds that topic and subject are two different syntactic components; "a sentence can have both topic and subject, or subject and topic, or only topic without subject, and of course, it can have neither explicit topic nor explicit subject (such as a subjectless sentence)."

[0110] In modern linguistics, the topic has been grammaticalized into a syntactic component. Any sentence in practical use contains three planes: syntax, semantics, and pragmatics. In terms of nature, the topic is a pragmatic concept, as opposed to description, and is a pragmatic component of the sentence. The subject is a syntactic concept, as opposed to the predicate, and is a syntactic component of the sentence. The agent is a semantic concept, as opposed to the patient, and is a semantic component of the sentence. However, this invention can also acknowledge that the topic is a component of the sentence, but it is not a component of the inner structure of the sentence, but rather a component of the outer structure, and does not fall within the scope of syntactic analysis. Therefore, both the topic and the subject are syntactic components; the topic is merely a pragmatic component, and the subject is also a syntactic component. However, whether understood semantically (agent as subject) or structurally defined (sentence-initial component as subject), any component acting as the subject can be marked with the topic marker ནན to become the topic. For example: Figure 24 This is a schematic diagram of the seventeenth example provided by the present invention. Figure 24 The Chinese equivalent of the Tibetan phrase is: I am a wounded person.

[0111] In the example, the core verb “ཡིན” (to be) governs two action elements. The first action element is the topic (pragmatic plane) or subject (syntactic plane) “ང” (I), and the second action element is the description (pragmatic plane) or object (syntactic plane) “རྨས་མ” (the injured person). Here, “ནི” is a referential conjunction indicating the relationship between the topic and the subject, and “ཞིག” (particle, referring to something) is a suffix describing “རྨས་མ” (the injured person). Dependency syntax is marked with RAD.

[0112] In this invention, words with a topic-description relationship or topic structure (topic + ནི + description + predicate) are classified into a divalent verb information base for processing in the valence information extraction rule base and are marked with TCV, where T is topic, C is comment, and V is predicate verb.

[0113] The set of relation tags for extracting information from divalent verbs is shown in Table 4.

[0114] Table 4 Information on trivalent verbs is automatically extracted.

[0115] A trivalent verb is a verb that connects three actors: the agent (doer or subject), the receiver (patient or object), and the other actor that assists in completing the action (participant or object). In entity relation extraction, this is called a quadruple relation (subject, participant, object, relation attribute value), which can be represented by P(x, y, z), where x represents the subject, y represents the direct object, z represents the indirect object, and P represents the predicate relation. However, this invention attempts to abstract relation extraction into the extraction of several triples in the extraction rules, rather than extracting "quadriples" or "multiples".

[0116] In Tibetan sentence structure, trivalent verbs are all transitive verbs. Generally, trivalent verbs appear in sentences with case particles and occult particles, as well as in sentences with case particles, occult particles, or accusative particles. This refers to the so-called "བྱ་བྱེད་ལས་གསུམ" (subject-verb-object tripartite structure) in Tibetan grammar, which is a sentence structure with a transitive verb with an object as the predicate, forming a subject-verb-object tripartite structure. The agent (བྱེད་ཐ་དད་པ) of a transitive verb (བྱ་བྱེད་ཐ་དད་པ) is the agent (བྱེད་པ་), the instrument or manner of the action (བྱེད་པ།), and the present active tense (བྱེད་ལས་ད་ལྟ་བ།). The three active parts of the verb (predicate) are collectively referred to as the "agent" (དངོས་པོ་་བདག); the three passive parts of the verb (transitive verbs) are collectively referred to as the "object" (བྱ་བའི་ལས།), the object and location being asked about (བྱ་བའི་ཡུལ།), and the future tense of the verb (བྱ་ལས་མ་འོངས་པ།).

[0117] Figure 25 This is a classification diagram of the automatic extraction rules for trivalent verb information provided by the present invention.

[0118] From the perspective of valency theory, a verb that can interact with three nominal components is called a trivalent verb. In Tibetan, these are the agent subject (བྱེད་པོ།), the object involved (བྱ་བའི་ལས), and the object of reference (བྱ་བའི་ཡུལ). In Tibetan sentence structure, trivalent verbs are all transitive verbs. Generally, trivalent verbs appear both in sentences with case particles and occult particles, and in sentences with case particles, occult particles, or accusative particles combined. For example: Figure 26This is a schematic diagram of the eighteenth example provided by the present invention. Figure 26 The Tibetan equivalent of this is: I dreamed of her.

[0119] In the example, the core verb "རྨིས" (to dream) governs three action elements: the first is "ང" (I) (the agent subject), the second is "མོ་རང" (she) (the object involved), and the third is "རྨི་ལམ" (dream) (the object). "ས" is an ergative particle added after the agent subject, "དུ" is a locative particle added after the object, and "བྱུང་" (an auxiliary verb, which here can refer to the complement of the verb "རྨིས") is a complement, forming a verb-complement relationship with the verb "རྨིས". From the perspective of valency theory, a verb that can relate to three nominal elements is called a trivalent verb. The three nominal components in this question are “ང”, “མོ་རང”, and “རྨི་ལམ”, therefore, it is classified as a trivalent verb. In this invention, verbs with subject-object-verb structures (subject, object, indirect object, and verb) and double objects are processed in the trivalent verb information database, marked with SOIV, where S is the subject, O is the object, I is the indirect object, and V is the verb. In entity relation extraction, this is referred to as a multi-entity relation (subject, subject, object, relation attribute value).

[0120] In an exemplary embodiment, this invention uses the Python programming language to implement the extraction process of triple relations under the rules of valence information extraction from a dependency parsing treebase. The verb valence information extraction method based on a dependency parsing treebase proposed in this invention performs dependency relation analysis and entity triple relation extraction on annotated corpus text. Valence information of Tibetan verbs is obtained through extraction rules, and then the valence information of Tibetan verbs is filtered based on statistical methods. Finally, the extracted valence information is manually proofread and stored in a database. This invention provides... Figure 27 The diagram shown illustrates the framework for automatic extraction of valency information for Tibetan verbs. The system's input is a raw text corpus, and its output is a database of extracted valency verb information.

[0121] This system consists of four main modules: 1. Preprocessing module.

[0122] (1) Various collections and organization of raw language data (2) Perform natural language processing on raw corpora, including sentence segmentation, word segmentation, part-of-speech tagging, and dependency parsing. After processing, such as labeling legal relationships, the corpus becomes a collection of sentences with natural language processing tags.

[0123] 2. Extraction module.

[0124] The extraction module uses heuristic rules on a set of sentences to obtain sentence triple relations and thus the valence information of Tibetan verbs.

[0125] 3. Post-processing module.

[0126] The extraction results are standardized and then manually verified.

[0127] According to the automatic extraction results, a total of 4553 pieces of information were extracted from the corpus, including 375 zero-valent verbs, 1990 monovalent verbs, 1879 divalent verbs, and 309 trivalent verbs. First, the text classifies all verbs with serial verbs, adverbial-head relationships, and verb-complement relationships extracted from the corpus as zero-valent verbs. For example: “མི་_dfམཐུན་_viསོ་སོ_rzར་_lsབསྡད་_viན་_cv[ལེགས_ad]hed།_ww” (Different ideas, each in their own place). ), "ངོ་སྤྲོད་_nvམདོར་ཙམ་_dc[གནང་_vt] hedརོགས_uq།_ww" (please be brief), "མཐོ་_dcརུ་_l d[བཏེགས་_vt]hed།_ww (elevation), “དར་ཁྱབ་_nvཤིན་ཏུ་_dc[ཆེ_ad]hed།_ww (widespread) and “ཤེ ས་མེད་ཚོར་མེད་_xxདུ་_ld[སྐྱེས་_vi]h edཤིང་_cnརྒྱས་_viབཞིན་_usའདུག_vc།_ww (Growing unnoticed); among them, “མི་_dfམཐུན་_viསོ་སོ_rzར་_lsབསྡད་_viན་_cv[ལེགས_ad]hed།_w w (Different ideas, each stays in their place) and “ས་མེད་ཚོར་མེད་_xxདུ་_ld[སྐྱེས་_vi]hedཤིང་_cnརྒྱས་_viབཞིན In the sentence ་_usའདུག_vc།_ww (growing unconsciously), the relationships between "བསྡད" (stay) and "ལེགས" (good), and between "སྐྱེས" (grow) and "རྒྱས" (grow) are sequential verbs. The phrases "མི་མཐུན་སོ་སོ" (alien) and "ཤེས་མེད་ཚོར་མེད" (unconsciously) are adverbial modifiers. Therefore, there is no action element in this sentence. Additionally, “ངོ་སྤྲོད་_nvམདོར་ཙམ་_dc[གནང་_vt]hedརོགས_uq།_ww” (please keep it brief) refers to the grammatical necessity of certain elements in language communication. Under certain conditions, omitting or omitting these necessary elements will not affect the normal progress of communication. This invention incorporates this principle... Sentences with mandatory components are classified as zero-valent verbs. "མཐོ་རུ་བཏེགས།" (to elevate) and "དར་ཁྱབ་ཤིན་ཏུ་ཆེ།" (to be very widespread) are what traditional grammar calls "དེ་ཉིད" (same case), which are adjectives with missing suffixes used as same-case complements. Therefore, such sentences are treated as zero-valent verbs.

[0128] Secondly, this invention categorizes the subject-predicate relation, the object-verb relation, and the subject-verb-object relation in the dependency syntax relation set into monovalent verbs. For example, “[སྐུ་ཚེ་_nn]sub རིང་_ad དུ་_ld [བརྟན་_vi]hed ཞིག_uq །_ww” (longevity) and [ཞིང་མགོ་_nn]sub སེར་བ_nn ས་_bo མ་_df [བརྡུང་_vt]hed །_ww (the wheat was not affected by hail) is a subject-predicate relationship, where “སྐུ་ཚེ” (lifespan) and “ཞིང་མགོ” (wheat head) are subordinate words (subjects). The subject-predicate relationship is formed by sub (subject) and the verbs “བརྟན” (unchanged) and “བརྡུང” (suffered).

[0129] For example, [གཡོག་_nn]sub ལ་_ls [རེ་_vt]hed དགོས_vu །_ww (needs to rely on others) and [སྐབས་ཐོག་_dp གྲོགས་_nn ]obj ལ་_ls [བཀུག་_vt]pre ནས་_cn བསྡད་པ་_vt བཟང་_ad །_ww (treat as a friend temporarily) The subject is the object, where "གཡོག" (others) and "གྲོགས" (friends) are subordinate nouns (patient subjects). Through "obj" (object) and "རེ" (rely on) and "བཀུག" (summon), a verb-object relationship is formed (the object becomes the subject). For example, the subject with the case particle "la" is the patient subject, indicating the object that receives the action or behavior, that is, the object involved in the action or behavior. The semantic structure of subject and predicate is "object + action".

[0130] For example, “དཔའ་རྩལ་_nn གྱིས་_bo [དགྲ་མགོ་_nn]sub [མནོན_vt]hed །_ww ” (bravely subduing the enemy) is an object-related relationship, where “དགྲ་མགོ” (enemy head) is the subordinate noun (patient subject), and the object-related verb relationship is formed through obj (object-related) and the governing word “གནོན” (hit), and due to the context, the object-related verb becomes the patient subject.

[0131] Furthermore, this invention categorizes subject-object-verb relationships and topic-description relationships as divalent verbs, such as: "[བུ་ཕྲུག་na]subགིས་bo[ཕ་མ་nn]obj[བསྐྱངས།_vt]hed། _ww" (child-nurturing parent) vs. _nn]sub[གནམ་_nn]objལ་_ls[འཆོངས_vi]hed།_w The phrase "w" (Azure Dragon Leaps Over the Southern Mountain) is a subject-verb-object structure. The phrase "བུ་ཕྲུག" (children) and "ལྷོ་རི་གཡུ་འབྲུག་སྔོན་མོ" (Azure Dragon of the Southern Mountain) are subordinate nouns (subjects). The subject-verb-object structure is formed through "sub" (subject), the subordinate nouns "ཕ་མ" (parents) (object involved), "གནམ" (sky) (object object), and the governing nouns "བསྐྱངས" (raise) and "འཆོངས" (leap). "[XXna]sub ནི་_ciནུབ་བྱང་མི་རིགས་སློབ་ཆེན་_nnནས་_jg [ཡོང་པ་_vi]pre ཡིན་_vj །_ww (XX comes from Northwest University for Nationalities) is a topic-description relationship, where "XX" (person's name) is the topic (subject) and "ནུབ་བྱང་མི་རིགས་སློབ་ཆེན་" (Northwest University for Nationalities) is the descriptive phrase (object). The topic-description structure is formed through tpc and tps and the verb "ཡོང་བ" (from).

[0132] Finally, structures with two objects (object, related object, or direct and indirect object) in a subject-verb-object relationship are classified as trivalent verbs. For example, “[ཕ་མ_nn]sub ས་_bo [བུ་ཕྲུག་_nn]obj རྣམས་_qj ལ་_ls [རེ་བ་_nn]obj ནན་མོ་_dc [འདོན་_vt]hed དགོས_vu །_ww” (Parents have high expectations for their children) is a subject-verb-object structure (with two objects). In the sentence, "ཕ་མ" (parents) is the subject (subject), "བུ་ཕྲུག་རྣམས" (children) is the object (subject), followed by the locative marker "ལ", and "རེ་བ" (expectation) is the object (subject). They form a subject-object-verb relationship through sub (subject), obj (object), and the governing verb "འདོན" (make a request).

[0133] The word cloud plot of the first 30 Tibetan monovalent verbs extracted from the tree bank reveals many problems in the classification and statistical process: First, there is a high frequency of duplicate entries. This duplication issue arises because manual part-of-speech tagging or dependency parsing did not involve selecting all entries, marking them, or using section breaks. For example, the verb "ཡིན (is)" resulted in three different entries depending on whether a section break was used between it and the marker: "ཡིན_vj", "ཡིན་_vu", and "ཡིན་_vi".

[0134] Second, a verb may have two or more markers. A verb having several different markers results in different valence information for the verb. For example, the verb "ཡོད" has two different markers: "ཡོད་_vc" (existence, being) and "ཡོད་_ud", etc.

[0135] Third, a verb can have different valences depending on the context. For example, "བཏང་" (to distribute) can function as a primary, secondary, or tertiary verb depending on the context of the sentence.

[0136] Fourth, the same verb can have different valence information in a sentence due to the different annotations of the action element and the state element. For example, in “རྒྱ་བཟའ་གྱིམ་ཤིང་ཀོང་ཇོ_nr ས་_boབོད་_ne དུ་_ls གསོ་རིག་_nn གི་_gi དཔེ་ཆ་_nnཁྱེར་_vtཡོང་_vu In the sentence “།_ww” (Princess Wencheng brought medical books to the Tubo Dynasty), the verb “ཁྱེར” (bring) is sometimes classified as a trivalent verb. This is because “རྒྱ་བཟའ་ཀོང་ཇོ” (Princess Wencheng) is marked as the agent, “བོད” (Tibet) as the object, and “གསོ་རིག་གི་དཔེ་ཆ” (medical books) as the referent. Therefore, the verb “ཁྱེར” (bring) is classified as a trivalent verb. Some argue that the verb "ཁྱེར" (to bring) should be classified as a divalent verb because "རྒྱ་བཟའ་ཀོང་ཇོ" (Princess Wencheng) is marked as the agent subject, "བོད" (Tibet) is marked as the adverbial of place, i.e., the stative element, and "གསོ་རིག་གི་དཔེ་ཆ" (medical books) is marked as the object of the action. Therefore, the verb "ཁྱེར" (to bring) is a divalent verb.

[0137] According to the automatic extraction results, there are 375 zero-valent verbs in the corpus. This invention classifies and counts the zero-valent verbs, including 109 monosyllabic zero-valent verbs, 243 disyllabic zero-valent verbs, 15 trisyllabic zero-valent verbs, and 8 quadrysyllabic zero-valent verbs.

[0138] The results for zero-valent verbs are shown in Table 5.

[0139] Table 5 Table 6 below shows the frequency distribution of the highest frequency verbs with zero valence in Tibetan.

[0140] Table 6 The Chinese equivalents of the Tibetan words in the sixth column of Table 6 (example) are: 1. How long did it take?

[0141] 2. Just do it that way.

[0142] 3. Avoid conflicts.

[0143] 4. Increase exports.

[0144] 5. Extremely rare.

[0145] 6. It becomes difficult to handle.

[0146] 7. Issue calls that cannot be made public.

[0147] 8. It is necessary to understand.

[0148] 9. Very widespread.

[0149] 10. It is easy to be hit.

[0150] According to corpus statistics, the top ten most frequent monovalent verbs are, in order: "འགྲོ་". _vi (walk), བྱས་_vt (do, past tense), བྱེད་_vt (do, present tense), གཏོང་_vt (give, distribute), མཐོང་_vt (see), གྱུར་_vi (become), འབོད་སྐུལ་_vt (call), ཡོང་བ་_vi (come), ཆེ_ad (big) and ཕོག་_vi (hit), etc.

[0151] Figure 28 This invention provides a word cloud diagram of the top 30 high-frequency zero-valent verbs extracted from the tree bank. Figure 28 Tibetan verbs include: "འགྲོ་" _vi (walk), བྱས་_vt (do, past tense), བྱེད་_vt (do, present tense), གཏོང་_vt (give, distribute), མཐོང་_vt (see), གྱུར་_vi (become), འབོད་་_vt (call), སྐུལ་_vt (call), ཡོང་བ་_vi (come), ཆེ_ad (big) and ཕོག་_vi (hit), etc.

[0152] (II) Results and Discussion of Monovalent Verb Extraction According to the automatic extraction results, there are 1990 monovalent verbs in the corpus. This invention classifies and counts the monovalent verbs, including 719 monosyllabic monovalent verbs, 1097 disyllabic monovalent verbs, 112 trisyllabic monovalent verbs, and 12 quadryllabic monovalent verbs.

[0153] The results of monovalent verbs extracted from the corpus are shown in Table 7.

[0154] Table 7 According to corpus statistics, the top ten most frequent monovalent verbs are, in order: “སོང་_vi (go),” “མེད ་_vj (do not),” “བྱུང་_vi (happen),” “འགྲོ་_vi (go),” “ཡོད་པ་_vc (exist),” “བྱེད་_vt (do),” “གྱུར་_vi (become),” “མེད་པ་_vc (do not),” “བཤད་_vt (speak),” and “མཐོང་_vi (see),” etc.

[0155] The frequency distribution of the highest frequency of monovalent verbs in Tibetan is shown in Table 8.

[0156] Table 8 The Chinese equivalents of the Tibetan words in the sixth column (example) of Table 8 are: 1. The condition has improved.

[0157] 2. One cannot be without money or valuables.

[0158] 3. At that time, there were even people who blamed and criticized.

[0159] 4. Where are you going? 5. The calligraphy possesses amazing skills.

[0160] 6. Women who wear this will walk the bridal entrance show.

[0161] 7. These accomplished disciples became renowned inheritors of their respective schools of thought.

[0162] 8. An insatiable king is like a beggar.

[0163] 9. This theory states that all external things are caused by the four elements.

[0164] 10. After seeing Che Zeng and the others, the three men stood up and continued their discussion.

[0165] Figure 29 This invention presents a word cloud diagram of the top 30 high-frequency Tibetan monovalent verbs extracted from the tree bank. Figure 29Tibetan verbs include: སོང་ (go); བྱུང་ (finished); མེད (not); འགྲོ (go); མཐོང་། (see); བཤད (speak); གཏོང་ (put); ཡིན་པ (yes); ཡོད་པ་ (exist); གྱུར (change); འདུག (exist); བྱེད (do); མེད་པ (No); བརྟེན (rely on); བལྟས (see); ཤེས (know); སྐྱེས (live); དགའ (happy); སྲུང་སྐྱོབ (protect); བྱུང་བ (has happened); ཡོང་བ (came); སྤེལ (held); ཐོན (come); ཐོབ (get) ལྡན། (have); ལེན། (take); ཆེ། (big).

[0166] (III) Results and Discussion of Divalent Verb Extraction According to the automatic extraction results, a total of 1,879 divalent verbs were extracted from the corpus, including 779 monosyllabic divalent verbs, 1,013 disyllabic divalent verbs, 64 trisyllabic divalent verbs, and 23 quadryllabic divalent verbs.

[0167] The results of divalent verbs extracted from the corpus are shown in Table 9.

[0168] Table 9 According to corpus statistics, the top ten most frequent divalent verbs are, in order: "ཡིན་_vj is", "ཡོད་_vc (exist)", "མེད་_vj (do not)", "གྱུར་_vi (become)", "སོང་_vi (go)", "བྱུང་_vi (happen)", "བྱེད་_ The frequencies of "vt (do)," "བྱས་_vt (do)," "ཐོབ་_vi (get)," and "ཞུགས་_vi (participate)" are "548," "193," "72," "57," "56," "53," "51," "48," "45," and "41," respectively.

[0169] The frequency distribution of the highest frequency of divalent verbs in Tibetan is shown in Table 10.

[0170] Table 10 The Chinese equivalents of the Tibetan words in the sixth column (example) of Table 10 are: 1. The person I walked with that day was my lover.

[0171] 2. Major agricultural and sideline products exported to China include beef and wine.

[0172] 3. No money or support.

[0173] 4. Wang Zixiwu Chi Xi became the prescriber of medicine.

[0174] 5. You can go wherever you want.

[0175] 6. I suddenly had an idea when I was in high school.

[0176] 7. The streets of the village and town were bustling.

[0177] 8. The original version was engraved with Zhuoni woodcuts.

[0178] Figure 30 The first 30 Tibetan bivalent verb word cloud diagrams provided for this invention. Figure 30 The Tibetan verbs in it include: ཡིན‌‌ (is); ཡོད‌ (has, exists); མེད‌ (doesn't have); ཞུགས‌ (participates); ཐོབ།‌ (gets); དགའ‌ (happy); གྱུར‌ (changes); བཟུང‌ (takes); བཟོས‌ (makes); བླངས‌ (takes, gets); ཤེས‌ (knows); བཞག‌ (puts); སྤེལ‌ (holds); བྱུང‌ (has); གཏོང‌ (distributes); འདུག‌ (is at); བཤད‌ (says); ཤོར‌ (loses); མིན‌ (is not); སོང་‌ (goes); སྐྱེས‌ (gives birth); བྱས‌ (has done); འགྲོ (walks); བཏང (distributes); ལྟ (looks); རྒྱག (sends out); ཆེ (big); ཟེར (it is said); བྱེད (does).

[0179] (IV) Results and Discussion of Trivalent Verb Extraction A total of 309 trivalent verbs were extracted from the corpus, among which there were 215 monosyllabic trivalent verbs, 90 disyllabic trivalent verbs, 4 trisyllabic trivalent verbs, and 1 tetra-syllabic trivalent verb.

[0180] The following are the results of trivalent verb extraction from the corpus as shown in Table 11.

[0181] Table 11 According to the corpus statistics, the top ten trivalent verbs with the highest frequencies are "རེད་_vj (is)", "བཤད་_vt (says)", "ཡིན་_vj (is)", "སྟོན་_vt (points)", "ཡོད་_vc (exists)", "བཏང་_vt (releases)", "བཞག_vt" (puts), "རྒྱག_vt (hits)", "སྐྱེས་_vi (gives birth)" and "བྱེད་_vt (does)", and their frequencies are "217", "21", "20", "16", "15", "11", "11", "8", "7" and "7", etc.

[0182] The frequency distribution of the highest frequency of trivalent verbs in Tibetan is shown in Table 12.

[0183] Table 12 The Chinese translation for the sixth column (example column) in Table 12 is: 1. I think snow is a delicious gift from heaven.

[0184] 2. Can you tell us about your most memorable experience in the past? 3. Some are like foxes, cowardly flatterers.

[0185] 4. Parents should welcome their children's friends.

[0186] 5. Ghor province is more than 300 kilometers away from Kabul.

[0187] 6. Parents should not argue in places where there are children.

[0188] 7. I often feel depressed about this vast county town.

[0189] 8. What kind of person does your child want to be? Why do you want their names? Figure 31 The word cloud diagram of the first 30 Tibetan trivalent verbs provided for this invention. Figure 31 The Tibetan verbs in this context correspond to the following Chinese characters: རེད (is); བཤད (speak); ཡིན (is); སྟོན (guide); བཏང (issue); གཏོང (issue); བཙུགས (build); རྒྱག (issue); སྐྱེས (give birth); བཏགས (falsely establish); བཏོན (read); གྱུར (already changed); བཞག (put down); ཁུར (carry); ཡོད་པ (exist); འདོན (recite; བསྟན (speak); སོང་ (go); ཡོད (have, exist); བྱེད (do); མེད (do not); ཉན (listen); བྱས (do); ཁྱེར (take); སྤེལ (hold); བྱིན (give); སྟེར (send); ཕུལ (gift); ཆེ (great).

[0190] The "complement" in valency grammar is related to the "entity" in entity relation extraction. In domination theory, monovalent, divalent, and trivalent verbs correspond to binary, triple, and quadruple relations in entity relations. Therefore, when automatically extracting verb valency information from a syntactically annotated treebank corpus, this invention uses dependency relationships between entity words and relation words, identifying verbs with serial verb, adverbial-head, and verb-complement relationships that can syntactically be classified as zero-valency verbs. From a valency grammar perspective, an action element without any nominal component is a zero-valency verb; a verb that can be dominated by one NP is called a monovalent verb. In this invention's method combining dependency relations and triple entity relations, the subject-verb, object-verb, and object-verb relationships within the dependency relation set are placed in monovalent verbs, corresponding to binary relations in relation extraction; a verb that can be dominated by two NPs becomes a divalent verb. Generally, sentences have a subject and object opposition, and its action element is composed of agent and patient components, corresponding to the... A triple relation, namely subject, object, and relation, refers to a verb that connects three actors: the agent (subject), the receiver (object), and the other actor (participant). This corresponds to a quadruple relation (subject, object, relation attribute value) in entity relation extraction, which can be represented by P(x, y, z), where x represents the subject, y represents the direct object, z represents the indirect object, and P represents the predicate relation.

[0191] Based on the need for automatic extraction of valence information from Tibetan verbs, this invention characterizes a multi-entity group automatic extraction rule model, namely, the valence extraction rule for zero-valence verbs: hed + predicate (VV, Verb-verb), hed + adv / Predicate + adv (ADV, adverbial), and hed + com / predicate + com (CMP, complement); Vesting rules for monovalent verbs: sub + hed and sub + bo + hed (SBV), sub / obj1 + hed (object subject VOB), obj2 + ls + hed (IOB); Vesting rules for divalent verbs: sub + obj2 + ls + hed (Subject, Indirect-object, Verb), sub + obj + hed (Subject, object, Verb), and sub + ནན + obj + hed (topic, Comment, Verb); Vesting rules for trivalent verbs: sub + obj1 + ls + obj2 + hed (Subject, object1, object2, Verb).

[0192] Analysis and statistics based on the corpus extraction results reveal several problems with the above rules: First, the frequency of repeated terms is high because manual part-of-speech tagging or dependency syntax tagging did not perform full tagging of the terms. The following issues arise: First, repetition can be caused by selection, marking, or section breaks. Second, a verb with two or more markings, or several different markings, can lead to different valence information for the verb. Third, a verb can have different valence information due to the influence of context. Fourth, the same verb in a sentence must have different markings for its action and state elements, leading to different valence information.

[0193] According to corpus analysis and statistics, a total of 4,553 valence information entries were extracted from the corpus, including 375 zero-valence verbs, 1,990 monovalent verbs, 1,879 divalent verbs, and 309 trivalent verbs.

[0194] When applying the Tibetan verb valence information extraction method based on dependency syntax treebank provided by this invention, it is not necessary to rely on... Figure 2 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0195] The above describes a method for extracting Tibetan verb valence information based on a dependency syntax treebank, provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding apparatus for extracting Tibetan verb valence information based on a dependency syntax treebank, such as... Figure 32 As shown.

[0196] Figure 32 A schematic diagram of a Tibetan verb valence information extraction device based on a dependency syntax treebank provided by the present invention includes: Module 3201 is used to acquire the Tibetan language corpus.

[0197] The preprocessing module 3202 is used to preprocess the Tibetan language corpus, generate a dependency syntax tree database, and calibrate the dependency syntax tree database to obtain a set of sentences with natural language processing tags. The sources of the Tibetan language corpus include literature, news, history, novels, textbooks, and the Yunnan Tibetan BART syntax library tool.

[0198] The extraction module 3203 is used to extract Tibetan verb valency information from sentences in a set of sentences with natural language processing tags based on the automatic verb valency extraction rules; the Tibetan verb valency information includes zero-valence verbs, monovalent verbs, divalent verbs and trivalent verbs.

[0199] Specific limitations regarding the Tibetan verb valence information extraction device based on dependency syntax treebanks can be found in the limitations of the Tibetan verb valence information extraction method based on dependency syntax treebanks mentioned above, and will not be repeated here. Each module in the aforementioned Tibetan verb valence information extraction device based on dependency syntax treebanks can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0200] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 2 The method provided is for extracting Tibetan verb valence information based on a dependency syntax treebank.

[0201] The present invention also provides Figure 33 The schematic diagram of the computer device shown is as follows: Figure 33 As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 2 The method provided is for extracting Tibetan verb valence information based on a dependency syntax treebank.

[0202] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0203] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A method for extracting valence information of Tibetan verbs based on a dependency syntax treebank, characterized in that, include: Obtain a Tibetan language corpus; the sources of the Tibetan language corpus include literature, news, history, novels, textbooks, and the Yunnan Tibetan Brat syntax library tool; The Tibetan corpus is preprocessed to generate a dependency syntax tree library, and the dependency syntax tree library is analyzed to determine the rules for automatic extraction of verb valence information. The automatic extraction rules for verb valence information include rules for zero-valence verbs, rules for monovalent verbs, rules for divalent verbs, and rules for trivalent verbs. Obtain the Tibetan sentence from which verb valence information is to be extracted, and execute the automatic verb valence information extraction rules to obtain the Tibetan verb valence information in the Tibetan sentence; The Tibetan verb valency information includes zero-valent verbs, monovalent verbs, divalent verbs, and trivalent verbs.

2. The method as described in claim 1, characterized in that, The process of obtaining Tibetan sentences from which verb valency information is to be extracted, and executing the automatic verb valency information extraction rules to obtain Tibetan verb valency information in the Tibetan sentences, specifically includes: When the entity component of a Tibetan sentence is a preposition and the syntactic relationship of the Tibetan sentence is a serial verb relationship, an adverbial-head relationship, a verb-complement relationship, a predicate-subject relationship, or a tense relationship, the verb in the Tibetan sentence is determined to be a zero-valence verb. When the entity component of a Tibetan sentence is a subject component and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship, or when the entity component of a Tibetan sentence is an object component and the syntactic relationship of the Tibetan sentence is an object-verb relationship or an object-verb relationship, the verb in the Tibetan sentence is determined to be a monovalent verb. When the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-verb-object, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-related, or when the subject component of a Tibetan sentence is the main component and the object component is the object component, and the syntactic relationship of the Tibetan sentence is topic and description, the verb in the Tibetan sentence is determined to be a divalent verb. When a Tibetan sentence has a subject-verb-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

3. The method as described in claim 1, characterized in that, The preprocessing includes sentence segmentation, word segmentation, part-of-speech tagging, and dependency parsing; the preprocessing of the Tibetan corpus to generate a dependency parsing treebase specifically includes: The Tibetan language corpus was then processed sequentially into sentence segmentation and word segmentation. Part-of-speech tagging is performed on the results of the word segmentation process; Based on the part-of-speech tagging results, dependency syntax relation tagging is performed to generate the dependency syntax tree library.

4. The method as described in claim 1, characterized in that, The rule for extracting zero-valent verbs is as follows: when the entity component of a Tibetan sentence is a preposition and the syntactic relationship of the Tibetan sentence is a serial verb relationship, an adverbial-head relationship, a verb-complement relationship, a predicate-subject relationship, or a tense relationship, the verb in the Tibetan sentence is determined to be a zero-valent verb.

5. The method as described in claim 1, characterized in that, The rule for extracting monovalent verbs is as follows: when the entity component of a Tibetan sentence is a subject component and the syntactic relationship of the Tibetan sentence is a subject-predicate relationship, or when the entity component of a Tibetan sentence is an object component and the syntactic relationship of the Tibetan sentence is an object-verb relationship or an object-verb relationship, the verb in the Tibetan sentence is determined to be a monovalent verb.

6. The method as described in claim 1, characterized in that, The rule for extracting divalent verbs is as follows: when the subject component of a Tibetan sentence is the main component and the object component is the secondary component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-verb-object, or when the subject component of a Tibetan sentence is the main component and the object component is the secondary component, and the syntactic relationship of the Tibetan sentence is subject-predicate or object-related, or when the subject component of a Tibetan sentence is the main component and the object component is the secondary component, and the syntactic relationship of the Tibetan sentence is topic-explanation, the verb in the Tibetan sentence is determined to be a divalent verb.

7. The method as described in claim 1, characterized in that, The rule for extracting trivalent verbs is as follows: when a Tibetan sentence has a subject-verb-object structure and two objects, the verb in the Tibetan sentence is determined to be a trivalent verb.

Images