A method and system for quantitatively evaluating the influence of citations on emotional semantics deconstruction

Abstract

The application discloses a kind of citation influence quantitative evaluation method and system of emotional semantic deconstruction, it is related to text analysis technical field, including: constructing feature word score table;Dependency syntax analysis is carried out to citation sentence, obtain the part-of-speech position feature set, divide citation sentence, obtain multiple citation clauses, carry out non-evaluative feature word elimination, obtain multiple optimization clauses;Multiple optimization clauses are carried out main word score identification, obtain multiple main word score identification results, auxiliary word score identification is carried out to multiple optimization clauses, obtain multiple auxiliary word score identification results;Obtain each optimization clause score, calculate citation influence quantitative score.The present application solves the technical problem that citation sentiment analysis in the prior art only stays in positive, negative, neutral coarse-grained classification, and it is difficult to quantitatively evaluate the influence degree of citation, to convert the qualitative citation influence evaluation into standardized quantitative score, improve the technical effect of evaluation objectivity and comparability.

Description

Technical Field

[0001] This invention relates to the field of text analysis technology, specifically to a method and system for quantitatively assessing the impact of citations through sentiment semantic deconstruction. Background Technology

[0002] With the development of bibliometrics and natural language processing technologies, citation sentiment analysis has gradually become an important means of identifying the evaluative tendencies of cited documents towards references. Related studies typically categorize citation sentiment into three types: positive, negative, and neutral, and use this categorization to complete sentiment identification or classification tasks. While this approach can reflect the basic sentiment polarity of citations, it lacks further differentiation of the semantic levels inherent in positive citations, such as groundbreaking contributions, original creations, in-depth advancements, and expansions. This makes it difficult to reveal the differences in the types and degrees of academic contribution among different citations, and also hinders the formation of quantitative calculations of citation impact, thus limiting the refined expression and objective comparison of citation impact evaluation. Summary of the Invention

[0003] This application provides a method and system for quantitatively assessing the impact of citations based on sentiment semantic deconstruction, which addresses the technical problem that existing citation sentiment analysis only focuses on coarse-grained classifications of positive, negative, and neutral, making it difficult to quantitatively assess the degree of citation impact.

[0004] In view of the above problems, this application provides a method and system for quantitative evaluation of citation impact based on sentiment semantic deconstruction.

[0005] The first aspect of this application provides a method for quantitatively assessing the citation impact of sentiment semantic deconstruction, the method comprising: A feature word score table is constructed; dependency parsing is performed on the cited sentences to obtain a part-of-speech position feature set; the cited sentences are segmented to obtain multiple cited clauses; non-evaluative feature word removal is performed on the multiple cited clauses to obtain multiple optimized clauses; based on the part-of-speech position feature set, the subject score of the multiple optimized clauses is identified according to the feature word score table to obtain multiple subject score identification results; based on the multiple subject score identification results, the auxiliary word score of the multiple optimized clauses is identified to obtain multiple auxiliary word score identification results; based on the multiple subject score identification results and the multiple auxiliary word score identification results, the score of each optimized clause is obtained, and the citation impact quantification score is calculated based on the scores of each optimized clause.

[0006] A second aspect of this application provides a citation impact quantification assessment system based on sentiment semantic deconstruction, the system comprising: The system includes a score table construction module for constructing a feature word score table; a dependency parsing module for performing dependency parsing on the cited sentence to obtain a part-of-speech position feature set; a segmentation module for segmenting the cited sentence to obtain multiple cited clauses; a feature word removal module for performing non-evaluative feature word removal on the multiple cited clauses to obtain multiple optimized clauses; a main word score identification module for identifying the main word score of the multiple optimized clauses based on the part-of-speech position feature set and the feature word score table to obtain multiple main word score identification results; a secondary word score identification module for identifying the secondary word score of the multiple optimized clauses based on the multiple main word score identification results to obtain multiple secondary word score identification results; and a scoring calculation module for obtaining the score of each optimized clause based on the multiple main word score identification results and the multiple secondary word score identification results, and calculating the citation impact quantification score based on the scores of each optimized clause.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application constructs a feature word score table; performs dependency parsing on the cited sentence to obtain a part-of-speech position feature set; divides the cited sentence to obtain multiple cited clauses; removes non-evaluative feature words from the multiple cited clauses to obtain multiple optimized clauses; based on the part-of-speech position feature set, identifies the main word score of the multiple optimized clauses according to the feature word score table to obtain multiple main word score identification results; identifies the auxiliary word score of the multiple optimized clauses according to the multiple main word score identification results to obtain multiple auxiliary word score identification results; obtains the score of each optimized clause according to the multiple main word score identification results and the multiple auxiliary word score identification results, and calculates the citation impact quantification score based on the scores of each optimized clause. This invention addresses the technical problem that existing citation sentiment analysis methods only offer coarse-grained classifications of positive, negative, and neutral, making it difficult to quantify the impact of citations. By deconstructing the sentiment semantics of citation sentences and combining lexical and syntactic features to quantify the impact of citations, this invention transforms qualitative citation impact assessments into standardized quantitative scores, improving the objectivity and comparability of the assessment. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart illustrating a method for quantifying the citation impact of sentiment semantic deconstruction, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a citation impact quantification assessment system based on sentiment semantic deconstruction, provided as an embodiment of this application.

[0010] Figure labeling: Score table construction module 11, dependency parsing module 12, segmentation module 13, feature word removal module 14, main word score recognition module 15, auxiliary word score recognition module 16, score calculation module 17. Detailed Implementation

[0011] This application provides a method and system for quantitatively assessing citation impact through sentiment semantic deconstruction. It addresses the technical problem that existing citation sentiment analysis methods only offer coarse-grained classifications of positive, negative, and neutral, making it difficult to quantitatively assess the degree of citation impact. By deconstructing the sentiment semantics of citation sentences and combining lexical and syntactic features, the application quantifies the degree of citation impact, transforming qualitative citation impact evaluation into standardized quantitative scoring and improving the objectivity and comparability of the assessment.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides a method for quantitatively evaluating the citation impact of sentiment semantic deconstruction, the method comprising: Step S100: Construct a feature word score table.

[0015] In this embodiment, when constructing the feature word score table, the feature words are first divided into four levels according to their influence on academic contributions: pioneering breakthrough, original first-time creation, milestone advancement, and expansion and optimization. Several feature words are included under each level, totaling 101. Based on this, an initial score corresponding to the level of contribution is assigned to each feature word. Pioneering breakthrough feature words have an initial score of 10 points, original first-time creation feature words have 7.5 points, milestone advancement feature words have 5 points, and expansion and optimization feature words have 2.5 points. Some expansion and optimization feature words are set to 2 points based on their part-of-speech and syntactic position conditions.

[0016] To ensure the operability and consistency of the scoring, the feature word scoring table specifies two conditions for each feature word to be scored: condition 1 is part of speech, and condition 2 is syntactic position. Only when a feature word meets both conditions simultaneously in the actual cited sentence can it receive an initial score; if only one condition is met, 1 point is deducted from the initial score; if neither condition is met, 2 points are deducted. Furthermore, to address word form variations in actual text, the feature word scoring table registers in-form terms and establishes a correspondence between in-form variants. For example, the in-form noun NN corresponds to variant NNS, the in-form verb VB corresponds to variants VBD, VBG, and VBN, the in-form adjective JJ corresponds to variants JJR and JJS, and the in-form adverb RB corresponds to variants RBR and RBS. This ensures that all variant forms appearing in the cited sentence can be accurately matched to the corresponding feature word terms. The feature word scoring table is shown in Table 1.

[0017] Table 1: Feature Word Score Table

[0018] Step S200: Perform dependency parsing on the quoted sentence to obtain the part-of-speech position feature set.

[0019] In this embodiment, when performing dependency parsing on the cited sentence, a pre-defined syntactic and semantic analysis tool is first invoked to perform syntactic and semantic analysis on the cited sentence. This tool parses the grammatical structure and syntactic dependency relations of the cited sentence, obtaining syntactic and semantic parsing data in a tree structure. This data includes the part-of-speech tagging, syntactic position attribution, subject-verb-object syntactic structure relationship, and modification relationships between words within the cited sentence. Then, the part-of-speech tagging information and the syntactic position information of each word are extracted from this syntactic and semantic parsing data. These two types of information are integrated to form a part-of-speech position feature set. This part-of-speech position feature set records the part-of-speech category of each word in the original sentence and its syntactic position attribution within the phrase structure.

[0020] Step S300: Divide the quoted sentence according to the part-of-speech position feature set to obtain multiple quoted clauses.

[0021] In this embodiment of the application, when dividing the cited sentence into clauses based on the syntactic and semantic analysis results, the predicate division rules, the independent connection rules of coordinating conjunctions, and the prohibition of splitting rules are first constructed. Then, the predicate division rules, the independent connection rules of coordinating conjunctions, and the prohibition of splitting rules are organized to generate a citation division model. The cited sentence is divided into clauses based on the guiding citation division model to generate multiple citation clauses.

[0022] Furthermore, in the method provided in the application embodiments, the process of dividing the cited sentence according to the part-of-speech position feature set and obtaining multiple cited clauses further includes: Construct predicate segmentation rules, independent rules for coordinating conjunctions, and rules prohibiting splitting; organize the predicate segmentation rules, independent rules for coordinating conjunctions, and rules prohibiting splitting to generate a citation segmentation model; divide the citation sentence into clauses according to the citation segmentation model to generate the multiple citation clauses.

[0023] Furthermore, the method provided in the application embodiments also includes: The predicate division rule includes that each structure containing a predicate is an independent clause; the independent connection rule of coordinating conjunctions includes that clauses connected by coordinating conjunctions are independent; the prohibition of splitting rules includes that clause structures, participial phrases and infinitive phrases are prohibited from being split into clauses.

[0024] In this embodiment, predicate segmentation rules, independent rules for coordinating conjunction connections, and rules prohibiting splitting are first constructed. The predicate segmentation rule involves scanning each syntactic node layer by layer in the syntactic tree to identify structures containing predicate verbs, and marking each complete subject-predicate structure containing a predicate as a candidate clause. The independent rules for coordinating conjunction connections involve extracting the subject-predicate parts of each coordinating structure when two or more candidate structures connected by coordinating conjunctions are identified in the syntactic tree or syntactic dependency relation, and marking each coordinating structure as an independent candidate clause. The rules prohibiting splitting involve not splitting clause structures, participial phrases, or infinitive phrases within candidate structures, but rather preserving them entirely within their respective candidate clauses. After constructing the predicate segmentation rules, independent rules for coordinating conjunction connections, and rules prohibiting splitting, these rules are organized to generate a citation segmentation model.

[0025] The citation segmentation model then processes the citation sentence in the following order: first, identifying predicate structures; second, identifying parallel structures; and finally, performing a prohibition on splitting correction. Specifically, it first determines the number of structures containing predicates in the citation sentence based on predicate segmentation rules; then, it determines whether each structure containing a predicate should be an independent sentence based on the rule of independent connection of coordinating conjunctions; and finally, it constrains the segmented results based on the prohibition on splitting rules, ensuring that clause structures, participial phrases, and infinitive phrases remain within their respective clauses. For example, in the sentence "The virus was identified, and the genome was cloned," because there are two predicate structures "was identified" and "was cloned" connected by the coordinating conjunction "and," the quoted sentence is divided into two clauses. In the sentence "The discovery led to advances, providing new insights," because there is only one predicate "led," and "providing new insights" is a participial phrase, it is not split separately according to the rule against splitting, and is therefore retained as a single clause. In the sentence "The development of HCV molecular clones that were infectious inchimpanzees was an important initial breakthrough," "that were infectious inchimpanzees" is a subordinate clause structure, and according to the rule against splitting, it is retained as a whole within its clause. Finally, based on the part-of-speech position feature set, the quoted sentences are divided into clauses, generating multiple clauses.

[0026] Step S400: Perform non-evaluative feature word removal on the multiple cited clauses to obtain multiple optimized clauses.

[0027] In this embodiment of the application, when performing non-evaluative feature word removal on multiple citation clauses, a non-evaluative feature word benchmark is first defined, and then non-evaluative feature word removal is performed on multiple citation clauses based on the non-evaluative feature word benchmark to generate multiple optimized clauses.

[0028] Furthermore, the method provided in the application embodiment, which involves removing non-evaluative feature words from the plurality of cited clauses to obtain a plurality of optimized clauses, further includes: Define a non-evaluative feature word benchmark; based on the non-evaluative feature word benchmark, remove non-evaluative feature words from the multiple citation clauses to generate the multiple optimized clauses.

[0029] In this embodiment, a non-evaluative feature word criterion is first defined. Specifically, whether a feature word demonstrates a contribution to the cited content is used as the non-evaluative feature word criterion; if a feature word appears in a citation clause but does not reflect a contribution to the cited content, then the feature word is not scored. Specifically, feature words are not scored when they appear as part of domain terminology, or when they are used only to state objective facts or general knowledge without evaluating the cited content, or when the sentence in which they appear is intended to point out the shortcomings, problems, or deficiencies of previous research rather than positively evaluating its contribution, or when they are used to describe neutral technical contexts such as experimental steps, biological processes, or mechanisms of action. Domain terminology includes, for example, professional gene names, protein names, compound names, and specific technology names. In other words, the non-evaluative feature word criterion is used to limit the scoring of feature words that appear as part of domain terminology in citation clauses.

[0030] After defining the non-evaluative feature word benchmark, non-evaluative feature words are removed from multiple citation clauses based on this benchmark. During this process, each citation clause is examined sentence by sentence to determine whether each feature word actually reflects a contribution to the cited content. If a feature word is not scored, it is excluded from the current citation clause and not retained for subsequent processing; if the feature word reflects a contribution to the cited content, it is retained. For example, the words "new," "development," "solid," "advanced," "novel," "accumulate," and "first" in expressions such as "new cell," "development of tumor," "solid tumors," "advanced melanoma," "advanced cancer," "novel viral," "Accumulate vesicle," and "First mutant" all appear as components of domain terminology in the sentences and are therefore excluded during the feature word cleaning process. After cleaning, multiple optimized clauses are generated.

[0031] Step S500: Based on the part-of-speech position feature set, perform main word score identification on the multiple optimized clauses according to the feature word score table, and obtain multiple main word score identification results.

[0032] In this embodiment, firstly, based on the part-of-speech position feature set, multiple optimized clauses are identified using a feature word score table to obtain multiple main word identification results; next, a main word score evaluation model is constructed according to the main word score evaluation rules. Then, based on the part-of-speech position feature set, the multiple main word identification results are scored according to the main word score evaluation model to generate multiple main word score identification results.

[0033] When identifying the subject score of multiple optimized clauses based on the syntactic position recognition results of each part of speech, the subject score of multiple optimized clauses is first identified to obtain multiple subject score recognition results; then, a subject score evaluation model is constructed according to the subject score evaluation rules; subsequently, based on the syntactic position recognition results of each part of speech, the multiple subject score recognition results are scored according to the subject score evaluation model to generate multiple subject score recognition results.

[0034] Furthermore, in the method provided in the application embodiment, based on the part-of-speech position feature set, the method performs main word score identification on the multiple optimized clauses according to the feature word score table to obtain multiple main word score identification results, and further includes: Based on the part-of-speech position feature set, the multiple optimized clauses are identified according to the feature word score table to obtain multiple main word identification results; a main word score evaluation model is constructed according to the main word score evaluation rules; based on the part-of-speech position feature set, the multiple main word identification results are scored according to the main word score evaluation model to generate the multiple main word score identification results.

[0035] In this embodiment, the subject word of the multiple optimized clauses is first identified based on the part-of-speech location feature set and the feature word score table. During this process, feature word degree matching is performed on the multiple optimized clauses based on the part-of-speech location feature set and the feature word score table to obtain the degree of each feature word; then, the degree of each feature word is input into the feature word score table to obtain multiple initial matching scores; finally, the subject word of the multiple optimized clauses is determined based on the multiple initial matching scores to obtain multiple subject word identification results.

[0036] Next, based on the main word score evaluation rules, a main word score evaluation model is constructed. In this process, a rule-organizing approach is used to organize the main word score evaluation rules into unified evaluation rules that can be directly used for main word scoring. The main word score evaluation rule is that a top-ranking score is obtained when both condition 1 and condition 2 are met simultaneously; condition 1 is the correct part of speech, and condition 2 is the correct syntactic position. If only one condition is met, 1 point is deducted from the top-ranking score; if neither condition is met, 2 points are deducted from the top-ranking score. When constructing the main word score evaluation model, the top-ranking score for each main word is first determined, then the correct part of speech and correct syntactic position for each main word are determined. Finally, the top-ranking score, correct part of speech, correct syntactic position, and the aforementioned deduction rules are all written into the main word score evaluation model, thus completing the construction of the main word score evaluation model.

[0037] Then, based on the part-of-speech position feature set, the main word score evaluation model scores multiple main word identification results. Specifically, the main word corresponding to each main word identification result is first input into the main word score evaluation model. Then, the main word score evaluation model extracts the actual part-of-speech and actual syntactic position of the main word from the part-of-speech position feature set, and reads the correct part-of-speech, correct syntactic position, and top-level score of the main word in the feature word score table. Subsequently, the main word score evaluation model compares the actual part-of-speech with the correct part-of-speech and compares the actual syntactic position with the correct syntactic position to obtain the corresponding comparison results. If both comparison results are satisfied, the main word score evaluation model outputs the top-level score of the main word as the score result of the main word; if only one comparison result is satisfied, the main word score evaluation model outputs the top-level score of the main word minus 1 point as the score result of the main word; if neither comparison result is satisfied, the main word score evaluation model outputs the top-level score of the main word minus 2 points as the score result of the main word. In cases where there are multiple subject identification results in the same optimized clause, each subject identification result is input into the subject score evaluation model for scoring. Based on the auxiliary word scores when different words are used as subjects, the result with the highest clause score is retained as the subject score identification result for that optimized clause. After completing the sentence-by-sentence scoring of multiple optimized clauses, multiple subject score identification results are generated.

[0038] Furthermore, in the method provided in the application embodiment, based on the part-of-speech position feature set, subject word identification is performed on the multiple optimized clauses according to the feature word score table to obtain multiple subject word identification results, which further includes: Based on the part-of-speech location feature set, feature word degree matching is performed on the multiple optimized clauses according to the feature word score table to obtain the degree of each feature word; the degree of each feature word is input into the feature word score table to obtain multiple initial matching scores; the multiple initial matching scores are used to determine the main word of the multiple optimized clauses to obtain the multiple main word recognition results.

[0039] In this embodiment, firstly, based on the part-of-speech location feature set, feature word degree matching is performed on the multiple optimized clauses according to the feature word score table. During this process, feature words in the multiple optimized clauses are read sentence by sentence. For each feature word, the contribution level of the feature word in the feature word score table is determined using the actual part-of-speech tagging and syntactic position attribution information recorded in the part-of-speech location feature set. The contribution level includes pioneering breakthrough, first-time originality, milestone advancement, and extended optimization, thereby completing the feature word degree matching and obtaining the degree of each feature word.

[0040] Next, the degree of each feature word is input into the feature word score table to obtain multiple initial matching scores. Specifically, based on the contribution level of each feature word and its corresponding initial score setting in the feature word score table, the initial feature word score corresponding to that feature word is read. When a feature word appears in plural, tense, comparative, or superlative form, it is first mapped to the infinitive word item according to the word variant correspondence, and then the initial feature word score corresponding to that infinitive word item is read. For cases where there are multiple feature words in the same optimization clause, the initial feature word score corresponding to each feature word is read separately, and the matching result corresponding to the optimization clause is formed. After executing multiple optimization clauses in sequence, multiple initial matching scores are obtained.

[0041] Finally, the subject word is determined for multiple optimized clauses based on multiple initial matching scores. Specifically, for each optimized clause, the initial matching scores corresponding to each feature word within the optimized clause are compared, and the feature word with the highest initial matching score is determined as the subject word of the optimized clause. When multiple highest initial matching scores exist in the same optimized clause, the corresponding feature words are determined as the subject words, and the clause scores based on each subject word are calculated. The group with the highest scores is taken as the subject word identification result of the optimized clause, thus generating multiple subject word identification results. For example, in the optimized clause corresponding to "The development of HCV molecular clones that were infectious inchimpanzees was an important initial breakthrough.", "breakthrough," "important," "initial," and "development" are all feature words. After reading their respective initial matching scores, the initial matching score for "breakthrough" is the highest, so "breakthrough" is identified as the main word. In the clause corresponding to "Since this breakthrough, molecular studies on autophagy have been conducted in mammalian cells as well as in yeast, and have fueled major advances in biomedical research," "major" and "advances" are both feature words, and their corresponding initial matching scores are the same. Therefore, "major" and "advances" are identified as the main words respectively, resulting in multiple main word identification results.

[0042] Step S600: Based on the multiple main word score recognition results, perform auxiliary word score recognition on the multiple optimized clauses to obtain multiple auxiliary word score recognition results.

[0043] In this embodiment of the application, when performing auxiliary word score recognition on multiple optimized clauses based on multiple main word score recognition results, the auxiliary word recognition is first performed on multiple optimized clauses based on multiple main word score recognition results to obtain auxiliary word recognition results for each clause; then, the auxiliary word recognition results for each clause are scored according to the auxiliary word score calculation rules to generate multiple auxiliary word score recognition results.

[0044] Furthermore, in the method provided in the application embodiments, the method of performing consonant word score recognition on the multiple optimized clauses based on the multiple subject word score recognition results to obtain multiple consonant word score recognition results further includes: Based on the multiple subject word score recognition results, the multiple optimized clauses are subjected to auxiliary word recognition to obtain the auxiliary word recognition results of each clause; the auxiliary word recognition results of each clause are scored according to the auxiliary word score calculation rules to generate the multiple auxiliary word score recognition results.

[0045] In this embodiment, when identifying auxiliary words in multiple optimized clauses based on multiple subject word score identification results, the multiple optimized clauses are first read one-to-one with their corresponding subject word score identification results to determine the subject word and its score in each optimized clause. Then, other feature words besides the subject word are checked one by one in the same optimized clause, and each feature word is judged based on the modification and semantic relationships between words reflected in the aforementioned syntactic and semantic analysis results. If the feature word has a direct modification relationship with the subject word, or if the feature word and the subject word together constitute the same scientific discovery, scientific progress, or scientific contribution, then the feature word is identified as an auxiliary word; if the feature word does not have a direct modification relationship with the subject word and does not form a semantic relationship with the subject word, then the feature word is not identified as an auxiliary word. For cases where multiple auxiliary words exist in the same optimized clause, the determination is based on whether there is a direct modification relationship or semantic relationship between each auxiliary word and the subject word. Feature words that have a direct modification relationship or together constitute the same scientific discovery, scientific progress, or scientific contribution expression with the subject word are identified as auxiliary words. After processing each sentence in the manner described above, the consonant recognition results for each sentence are obtained.

[0046] Next, the consonant recognition results for each sentence are scored according to the consonant scoring rules. Specifically, the part of speech and syntactic position of each consonant are read one by one. Then, conditions 1, 2, and the top-level score corresponding to the consonant are read from the feature word score table. Condition 1 is the correct part of speech, and condition 2 is the correct syntactic position. Subsequently, the actual part of speech of the consonant is compared with condition 1, and the actual syntactic position of the consonant is compared with condition 2. When both comparison results are satisfied, the top-level score of the consonant is used as the score result of the consonant. When only one comparison result is satisfied, the top-level score of the consonant is reduced by 1 point as the score result of the consonant. When neither comparison result is satisfied, the top-level score of the consonant is reduced by 2 points as the score result of the consonant. After completing the part of speech and syntactic position comparison, it is further determined whether the consonant has a direct modifying relationship with the subject in syntax. If the consonant does not have a direct modifying relationship with the subject in syntax, the score result of the consonant is determined to be 0 points. After calculating the scores for each consonant in each optimized clause according to the above steps, the scoring results corresponding to each consonant are recorded one by one, generating multiple consonant score recognition results.

[0047] For example, in the second optimized clause corresponding to "Since this breakthrough, molecular studies on autophagy have been conducted in mammalian cells as well as in yeast, and have fueled major advances in biomedical research," two cases are first processed based on the subject word score identification result. In case 1, "advances" is the subject word, so other feature words in the same sentence are checked, and "major" can be identified as a modifier; then the part-of-speech and syntactic position corresponding to "major" are read, and compared with conditions 1 and 2 corresponding to "major" in the feature word score table. Since "major" meets both conditions and has a modification relationship with the subject word, the score of "major" is 2.5. In case 2, if major is the main word, then we continue to check other feature words in the same sentence and can identify advances as an auxiliary word. Then we read the part of speech and syntactic position corresponding to advances and compare it with condition 1 and condition 2 corresponding to advances in the feature word score table. Since advances meets both conditions and has a modification relationship with the main word, the score of advances is 2.5 points. Thus, the corresponding auxiliary word score recognition results are generated respectively.

[0048] For example, in the optimized clause corresponding to "The discovery of the first autophagy mutants in yeast initiated a boom in this field," if "first" is the subject, then other feature words in the same sentence are checked, and "initiated" can be identified as an auxiliary word. Next, it is determined whether "initiated" meets the part-of-speech and syntactic position conditions, and whether there is a direct modification relationship between "initiated" and the subject. Since there is no direct modification relationship between "initiated" and the subject, the score for "initiated" is 0. Conversely, if "initiated" is the subject, then "first" is treated the same way as an auxiliary word; since there is also no direct modification relationship between "first" and the subject, the score for "first" is also 0. After completing the corresponding scoring in the above manner, the auxiliary word score recognition result corresponding to this optimized clause is generated.

[0049] Step S700: Based on the multiple main word score recognition results and the multiple auxiliary word score recognition results, obtain the score of each optimized clause, and calculate the citation impact quantification score based on the score of each optimized clause.

[0050] In this embodiment, when calculating the citation impact quantification score, firstly, a formula for calculating the total score of a clause and a formula for calculating the total score of a sentence are constructed. Then, multiple subject word score recognition results and multiple consonant word score recognition results are input into the clause total score calculation formula for calculation, outputting the score of each optimized clause. Finally, the scores of each optimized clause are input into the sentence total score calculation formula for calculation to obtain the citation impact quantification score.

[0051] Furthermore, in the method provided in the application embodiments, obtaining the score of each optimized clause based on the multiple subject word score recognition results and the multiple auxiliary word score recognition results further includes: Construct a formula for calculating the total score of a clause; input the multiple subject score recognition results and the multiple auxiliary word score recognition results into the formula for calculating the total score of a clause, and output the score of each optimized clause.

[0052] Furthermore, in the method provided in the application embodiments, calculating the citation impact quantification score based on the scores of each optimized clause further includes: Construct a formula for calculating the total sentence score; input the scores of each optimized clause into the formula for calculating the total sentence score to obtain the quantitative score of the citation impact.

[0053] In this embodiment, when calculating the citation impact quantification score based on multiple subject word score recognition results and multiple auxiliary word score recognition results, a clause total score calculation formula and a sentence total score calculation formula are first constructed. Specifically, for each optimized clause, the subject word score recognition result and auxiliary word score recognition result corresponding to the optimized clause are read, and the subject word score in the optimized clause is summed with the scores of all auxiliary words that meet the conditions to construct the clause total score calculation formula, i.e., clause total score = subject word score + all auxiliary word scores that meet the conditions; then, for all optimized clauses in the same citation sentence, the clause total score corresponding to each optimized clause is read, and the total scores of each clause are accumulated to construct the sentence total score calculation formula, i.e., sentence total score = sum of the total scores of all clauses. This process completes the construction of the clause total score calculation formula and the sentence total score calculation formula.

[0054] After constructing the formulas for calculating the total score of clauses and sentences, multiple subject score recognition results and multiple auxiliary word score recognition results are input into the formulas for calculating the total score of clauses. The total score of each optimized clause is calculated separately, and the score of each optimized clause is output. Specifically, the corresponding subject score recognition results and auxiliary word score recognition results are read sentence by sentence for each optimized clause. The subject score is added to all auxiliary word scores that meet the conditions to obtain the total score of the optimized clause. When an optimized clause has multiple sets of subject score recognition results and auxiliary word score recognition results, the total score of each set of results is calculated separately, and the set with the highest score is taken as the total score of the optimized clause.

[0055] The scores of each optimized clause are then input into the sentence total score calculation formula to obtain the citation impact quantification score. Specifically, the total scores of all optimized clauses in the same cited sentence are summed to obtain the sentence total score, which is then determined as the citation impact quantification score.

[0056] Taking the sentence "Since this breakthrough, molecular studies on autophagy have been conducted in mammalian cells as well as in yeast, and have fueled major advances in biomedical research" as an example, the subject score for clause 1 is 10 points, and there is no score for the auxiliary word. Therefore, the total score for clause 1 is 10 points. Clause 2 has two sets of results. The first set has "advances" as the subject and "major" as the auxiliary word, corresponding to a total score of 2.5 points + 2.5 points = 5 points. The second set has "major" as the subject and "advances" as the auxiliary word, corresponding to a total score of 2.5 points + 2.5 points = 5 points. Therefore, the total score for clause 2 is 5 points. Then, the total score of clause 1 (10 points) and the total score of clause 2 (5 points) are added together to get a total sentence score of 15 points, and a citation impact quantification score is generated.

[0057] In summary, the embodiments of this application have at least the following technical effects: This application constructs a feature word score table; performs dependency parsing on the cited sentence to obtain a part-of-speech position feature set; divides the cited sentence to obtain multiple cited clauses; removes non-evaluative feature words from the multiple cited clauses to obtain multiple optimized clauses; based on the part-of-speech position feature set, identifies the main word score of the multiple optimized clauses according to the feature word score table to obtain multiple main word score identification results; identifies the auxiliary word score of the multiple optimized clauses according to the multiple main word score identification results to obtain multiple auxiliary word score identification results; obtains the score of each optimized clause according to the multiple main word score identification results and the multiple auxiliary word score identification results, and calculates the citation impact quantification score based on the scores of each optimized clause. This invention addresses the technical problem that existing citation sentiment analysis methods only offer coarse-grained classifications of positive, negative, and neutral, making it difficult to quantify the impact of citations. By deconstructing the sentiment semantics of citation sentences and combining lexical and syntactic features to quantify the impact of citations, this invention transforms qualitative citation impact assessments into standardized quantitative scores, improving the objectivity and comparability of the assessment.

[0058] Example 2, based on the same inventive concept as the citation impact quantification assessment method of sentiment semantic deconstruction in the foregoing examples, such as... Figure 2 As shown, this application provides a system for quantifying the citation impact of sentiment semantic deconstruction. The system and method embodiments in this application are based on the same inventive concept. The system includes: The scoring table construction module 11 is used to construct a feature word scoring table; the dependency parsing module 12 is used to perform dependency parsing on the cited sentence to obtain a part-of-speech position feature set; the segmentation module 13 is used to segment the cited sentence to obtain multiple cited clauses; the feature word removal module 14 is used to remove non-evaluative feature words from the multiple cited clauses to obtain multiple optimized clauses; the main word score recognition module 15 is used to perform main word score recognition on the multiple optimized clauses based on the part-of-speech position feature set and the feature word scoring table to obtain multiple main word score recognition results; the auxiliary word score recognition module 16 is used to perform auxiliary word score recognition on the multiple optimized clauses based on the multiple main word score recognition results to obtain multiple auxiliary word score recognition results; and the scoring calculation module 17 is used to obtain the score of each optimized clause based on the multiple main word score recognition results and the multiple auxiliary word score recognition results, and calculate the citation impact quantification score based on the score of each optimized clause.

[0059] Furthermore, the system is also used to implement the following functions: Construct predicate segmentation rules, independent rules for coordinating conjunctions, and rules prohibiting splitting; organize the predicate segmentation rules, independent rules for coordinating conjunctions, and rules prohibiting splitting to generate a citation segmentation model; divide the citation sentence into clauses according to the citation segmentation model to generate the multiple citation clauses.

[0060] Furthermore, the system is also used to implement the following functions: The predicate division rule includes that each structure containing a predicate is an independent clause; the independent connection rule of coordinating conjunctions includes that clauses connected by coordinating conjunctions are independent; the prohibition of splitting rules includes that clause structures, participial phrases and infinitive phrases are prohibited from being split into clauses.

[0061] Furthermore, the system is also used to implement the following functions: Define a non-evaluative feature word benchmark; based on the non-evaluative feature word benchmark, remove non-evaluative feature words from the multiple citation clauses to generate the multiple optimized clauses.

[0062] Furthermore, the system is also used to implement the following functions: Based on the part-of-speech position feature set, the multiple optimized clauses are identified according to the feature word score table to obtain multiple main word identification results; a main word score evaluation model is constructed according to the main word score evaluation rules; based on the part-of-speech position feature set, the multiple main word identification results are scored according to the main word score evaluation model to generate the multiple main word score identification results.

[0063] Furthermore, the system is also used to implement the following functions: Based on the part-of-speech location feature set, feature word degree matching is performed on the multiple optimized clauses according to the feature word score table to obtain the degree of each feature word; the degree of each feature word is input into the feature word score table to obtain multiple initial matching scores; the multiple initial matching scores are used to determine the main word of the multiple optimized clauses to obtain the multiple main word recognition results.

[0064] Furthermore, the system is also used to implement the following functions: Based on the multiple subject word score recognition results, the multiple optimized clauses are subjected to auxiliary word recognition to obtain the auxiliary word recognition results of each clause; the auxiliary word recognition results of each clause are scored according to the auxiliary word score calculation rules to generate the multiple auxiliary word score recognition results.

[0065] Furthermore, the system is also used to implement the following functions: Construct a formula for calculating the total score of a clause; input the multiple subject score recognition results and the multiple auxiliary word score recognition results into the formula for calculating the total score of a clause, and output the score of each optimized clause.

[0066] Furthermore, the system is also used to implement the following functions: Construct a formula for calculating the total sentence score; input the scores of each optimized clause into the formula for calculating the total sentence score to obtain the quantitative score of the citation impact.

[0067] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for quantitatively assessing the citation impact of sentiment semantic deconstruction, characterized in that, The method includes: A feature word score table is constructed, which specifically divides the feature words into four levels according to their influence on academic contributions: pioneering breakthrough, first original, milestone advancement and expansion optimization. Several feature words are included in each level, and each feature word is assigned an initial score corresponding to its contribution level. Dependency parsing is performed on the quoted sentences to obtain a set of part-of-speech position features; The quoted sentence is divided into multiple quoted clauses; Non-evaluative feature word removal is performed on the multiple cited clauses to obtain multiple optimized clauses; Based on the part-of-speech location feature set, the main word score is identified for the multiple optimized clauses according to the feature word score table, and multiple main word score identification results are obtained. Based on the multiple subject word score recognition results, the multiple optimized clauses are subjected to auxiliary word score recognition to obtain multiple auxiliary word score recognition results; Based on the multiple subject word score recognition results and the multiple auxiliary word score recognition results, the score of each optimized clause is obtained, and the citation impact quantification score is calculated based on the score of each optimized clause; Based on the part-of-speech location feature set, the main word scores of the multiple optimized clauses are identified according to the feature word score table, and multiple main word score identification results are obtained, including: Based on the part-of-speech location feature set, the main word of the multiple optimized clauses is identified according to the feature word score table to obtain multiple main word identification results; Construct a main keyword score evaluation model based on the main keyword score evaluation rules; Based on the part-of-speech location feature set, the multiple main word identification results are scored according to the main word score evaluation model to generate the multiple main word score identification results; Based on the part-of-speech location feature set, the subject word of the multiple optimized clauses is identified according to the feature word score table, and multiple subject word identification results are obtained, including: Based on the part-of-speech location feature set, feature word degree matching is performed on the multiple optimized clauses according to the feature word score table to obtain the degree of each feature word; Input the degree of each feature word into the feature word score table to obtain multiple initial matching scores; Based on the multiple initial matching scores, the multiple optimized clauses are determined to have the multiple main word identification results. Based on the multiple subject word score recognition results, the multiple optimized clauses are subjected to consonant word score recognition to obtain multiple consonant word score recognition results, including: Based on the multiple subject word score recognition results, the multiple optimized clauses are subjected to auxiliary word recognition to obtain the auxiliary word recognition results for each clause; The consonant recognition results of each sentence are scored according to the consonant score calculation rules to generate multiple consonant score recognition results.

2. The method for quantitatively assessing the citation impact of sentiment semantic deconstruction as described in claim 1, characterized in that, The quoted sentence is divided into multiple quoted clauses, including: Construct rules for predicate segmentation, rules for independent connection of coordinating conjunctions, and rules prohibiting splitting; By organizing the predicate segmentation rules, the independent connection rules for coordinating conjunctions, and the prohibition on splitting, a citation segmentation model is generated. The citation sentence is divided into clauses according to the citation segmentation model to generate the multiple citation clauses.

3. The method for quantitatively assessing the citation impact of sentiment semantic deconstruction as described in claim 2, characterized in that, The predicate division rule includes that each structure containing a predicate is an independent clause; The rule for independent connection of coordinating conjunctions includes that the clauses connected by coordinating conjunctions are each independent; The rules prohibiting splitting include clause structures, participial phrases, and infinitive phrases from being split into clauses.

4. The method for quantitatively assessing the citation impact of sentiment semantic deconstruction as described in claim 1, characterized in that, Non-evaluative feature word removal is performed on the multiple cited clauses to obtain multiple optimized clauses, including: Define non-evaluative feature word benchmarks; Based on the non-evaluative feature word benchmark, the multiple cited clauses are subjected to non-evaluative feature word removal to generate the multiple optimized clauses.

5. The method for quantitatively assessing the citation impact of sentiment semantic deconstruction as described in claim 1, characterized in that, Based on the multiple subject score recognition results and the multiple consonant score recognition results, the scores of each optimized clause are obtained, including: Construct a formula for calculating the total score of clauses; The multiple main word score recognition results and the multiple auxiliary word score recognition results are input into the clause total score calculation formula, and the scores of each optimized clause are output.

6. The method for quantitatively assessing the citation impact of sentiment semantic deconstruction as described in claim 1, characterized in that, The citation impact quantification score is calculated based on the scores of each optimized clause, including: Construct a formula for calculating the total score of a sentence; Input the scores of each optimized clause into the sentence total score calculation formula to obtain the citation impact quantification score.

7. A citation impact quantification assessment system based on sentiment semantic deconstruction, characterized in that, The system is used to execute a citation impact quantification assessment method for sentiment semantic deconstruction as described in any one of claims 1-6, and the system comprises: The score table construction module is used to construct the feature word score table; The dependency parsing module is used to perform dependency parsing on quoted sentences and obtain part-of-speech position feature sets. The segmentation module is used to segment the quoted sentence and obtain multiple quoted clauses; The feature word removal module is used to perform non-evaluative feature word removal on the multiple cited clauses to obtain multiple optimized clauses; The main word score recognition module is used to perform main word score recognition on the multiple optimized clauses based on the part-of-speech position feature set and the feature word score table, and obtain multiple main word score recognition results. The auxiliary word score recognition module is used to perform auxiliary word score recognition on the multiple optimized clauses based on the multiple main word score recognition results, and obtain multiple auxiliary word score recognition results; The scoring calculation module is used to obtain the score of each optimized clause based on the multiple main word score recognition results and the multiple auxiliary word score recognition results, and to calculate the citation impact quantification score based on the score of each optimized clause.

Images