A brand AI expressiveness evaluation system and method based on multi-source data fusion

Abstract

The application belongs to the technical field of semantic processing, and discloses a brand AI expressiveness evaluation system and method based on multi-source data fusion, which comprises a brand semantic construction module, which is used for obtaining brand related data from different data sources and performing semantic disassembly to generate a semantic primitive set; different semantic primitives are matched and uniformly identified to form a brand semantic reference package; a semantic relationship construction module is used for constructing the association structure between semantic primitives according to the brand semantic reference package, the preset semantic pointing relationship and the interaction time sequence relationship, and forming a semantic coupling relationship set; a semantic conflict judgment module is used for introducing a conflict judgment function to identify the conflicts of different semantic primitives when different semantic primitives act simultaneously according to the semantic coupling relationship set; and the semantic primitives with conflicts are inhibited to obtain an effective semantic set after inhibition, thereby significantly improving the accuracy and interpretability of brand AI expressiveness evaluation.

Description

Technical Field

[0001] This invention relates to the field of semantic processing technology, and more specifically, to a brand AI performance evaluation system and method based on multi-source data fusion. Background Technology

[0002] In existing technologies related to brand AI performance evaluation, multi-source data fusion is usually a basic capability, but the overall processing method is relatively crude. Most existing solutions simply summarize brand-related information from different channels and then fuse it through weighted or statistical methods. While this approach can produce a comprehensive result, it does not perform in-depth decomposition and structured expression at the semantic level, especially lacking fine-grained modeling of the relationships between semantics. Therefore, the combination logic between different semantics is often "black box," resulting in weak interpretability of the overall result.

[0003] First, in the specific semantic processing, different semantic primitives may be activated simultaneously. For example, positive and negative evaluations may coexist, or brand descriptions from different perspectives may appear in parallel. However, existing technologies usually only focus on whether they appear simultaneously, without further analyzing the consistency or conflict of these semantics in terms of expression goals. Therefore, semantic superposition is very likely to occur, ultimately leading to confusion or even semantic aliasing and distortion in the expression results.

[0004] Second, existing methods for semantic relationship modeling mostly rely on co-occurrence frequency, statistical correlation, or simple association rules to determine the connections between semantics. This approach remains at the level of surface structural analysis and does not introduce a deeper mechanism for judging "consistency of expression goals." Therefore, it cannot accurately identify whether there are essential conflicts between semantics, resulting in insufficient conflict identification capabilities and a tendency to miss key semantic contradictions. Even in some schemes that can detect semantic conflicts, these conflicts are usually simply marked and do not participate in subsequent processing. In other words, there is a lack of a true conflict resolution mechanism, which cannot automatically remove or suppress conflicting semantics during the fusion process. As a result, the final output may still be affected by conflicting information, leading to poor stability.

[0005] Third, in the process of characterizing brand AI performance, existing technologies often focus on static semantic matching or analysis methods based on statistical changes, lacking the ability to model the continuous evolution of semantics over time. In particular, the changing relationships between adjacent semantic units are not uniformly incorporated into the analysis framework, making the semantic change analysis dimension relatively singular and difficult to fully reflect the real changes in brand performance in the dynamic process.

[0006] Fourth, existing methods for updating performance status often rely on continuous numerical adjustments or empirical rules, lacking clear state trigger boundaries and a unified state switching mechanism. This results in an unclear process of performance status changes, and even introduces a degree of randomness and instability. Furthermore, these methods typically only output the final evaluation result without providing a structured record of how brand performance changes throughout the process. Consequently, it is difficult to establish a traceable state evolution path, leading to a weak process-oriented analysis of the overall results.

[0007] In view of this, the present invention proposes a brand AI performance evaluation system and method based on multi-source data fusion to solve the above problems. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the existing technology and to achieve the above objectives, the present invention provides the following technical solution: a brand AI performance evaluation system based on multi-source data fusion, comprising: The brand semantic construction module is used to acquire brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; it then matches and unifies the identification of different semantic primitives to form a brand semantic baseline package. The semantic relationship construction module is used to construct the association structure between semantic primitives based on the brand semantic benchmark package and according to the preset semantic pointing relationship and interaction timing relationship, forming a set of semantic coupling relationships; The semantic conflict determination module is used to identify conflicts between different semantic primitives when they act simultaneously, based on the set of semantic coupling relationships; and to suppress the semantic primitives that have conflicts, so as to obtain the effective semantic set after suppression. The expressive state determination module is used to construct an expressive state sequence based on the effective semantic set and to determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, the expressive state is triggered to flip and a state flip sequence is formed. The performance evaluation generation module is used to comprehensively analyze the semantic performance characteristics of brand AI in the performance process based on the set of semantic coupling relationships, the effective semantic set, and the state reversal sequence, and output the corresponding brand AI performance evaluation results.

[0009] Preferably, the method for generating the semantic primitive set includes: Brand-related data from different data sources is obtained through a preset data access interface, and the format of the brand-related data obtained from different data sources is normalized; data cleaning is performed on the normalized brand-related data to obtain original semantic data with a unified structure; based on preset semantic parsing rules, the original semantic data is semantically decomposed and divided into different semantic fragments. During the semantic decomposition process, based on the syntactic structure and semantic role labeling rules, the expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are extracted from each semantic segment. The expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are then combined and encapsulated to form a set of semantic primitives.

[0010] Preferably, the method for forming a brand semantic benchmark package includes: The semantic primitives in the semantic primitive set are matched and judged. Based on the expression subject information, expression behavior information and expression target information in the semantic primitives, the semantic consistency between different semantic primitives is compared. The semantic primitives with consistent semantic orientation are identified according to the preset keyword matching rules. Semantic primitives that meet the keyword matching rules are merged, and semantic primitives with consistent semantic orientations are merged into the same semantic unit, while retaining the corresponding expressive feedback information. The merged semantic units are then uniformly identified, with each semantic unit assigned a unique identifier. The semantic units that have completed the uniform identification process are then aggregated to form a brand semantic baseline package.

[0011] Preferably, the method for forming a set of semantically coupled relationships includes: Based on the brand semantic baseline package, the semantic primitives contained in each semantic unit are extracted, and the semantic primitives are sorted according to the time identifier to form a semantic sequence. Based on the preset semantic pointing relationship, the expression subject information, expression behavior information and expression target information of adjacent semantic primitives in the semantic sequence are compared to identify the semantic response relationship between adjacent semantic primitives. The association between semantic primitives is constrained by the preset interaction time sequence relationship. Only semantic primitive pairs that satisfy the sequential order are retained, and the association direction of semantic primitive pairs is determined according to the semantic response relationship. Semantic primitive pairs that satisfy the semantic response relationship and the interaction time sequence relationship are summarized to construct the association structure between semantic primitives. The association structure is then standardized and aggregated to form a set of semantic coupling relationships.

[0012] Preferably, the method for conflict identification of different semantic primitives includes: Extract semantic primitives that are simultaneously occupied from the set of semantic coupling relationships and characterize the occupancy state of each semantic primitive; construct a semantic conflict determination function to combine and determine any two semantic primitives that are in an active state. When two semantic primitives are simultaneously occupied and their respective expression target states are different, it is determined that there is a semantic conflict relationship between the two semantic primitives. When two semantic primitives are not simultaneously occupied, or when two semantic primitives are simultaneously occupied and their respective expression target states are consistent, it is determined that there is no semantic conflict relationship between the two semantic primitives; the semantic primitive pairs that meet the conflict determination conditions are marked to form a conflict semantic set.

[0013] Preferably, the method for obtaining the suppressed effective semantic set includes: Based on the conflict semantic set, for each semantic primitive, the conflict relationship between the semantic primitive and other semantic primitives is obtained, and the corresponding maximum conflict value is determined; the occupancy state of each semantic primitive is suppressed. When a semantic primitive has a semantic conflict relationship with any other semantic primitive, the occupancy state of the semantic primitive is suppressed to an invalid state. When a semantic primitive does not have a semantic conflict with other semantic primitives, the occupancy state of the semantic primitive remains unchanged. The semantic primitives after suppression are screened, and semantic primitives with valid occupancy state are retained, while semantic primitives with invalid occupancy state are removed, thus forming a set of valid semantic primitives after suppression.

[0014] Preferably, the method for constructing an expressive state sequence based on an effective semantic set includes: Based on the effective semantic set after conflict suppression, the time markers corresponding to each semantic primitive are extracted, and the semantic primitives are sorted according to the time markers to form an ordered semantic sequence. The state of each semantic primitive in the ordered semantic sequence is represented. Based on the occupancy state, expression behavior information and expression target state of the semantic primitive, each semantic primitive is mapped to a corresponding expressive state unit. The expressive state units are then organized in series according to the time order to form an expressive state sequence.

[0015] Preferably, the method for forming the state reversal sequence includes: In the expressive state sequence, the expressive state units corresponding to adjacent effective semantic primitives are compared. Based on the occupancy state, expressive behavior information and expressive target state of the semantic primitives, the degree of semantic change between adjacent effective semantic primitives is determined. The degree of semantic change between adjacent valid semantic primitives is compared with a preset change threshold. When the degree of semantic change meets the preset change threshold, a state flip trigger function is constructed based on the degree of semantic change to trigger the expressive state flip, so that the current expressive state is switched from the first state to the second state or from the second state to the first state. When the degree of semantic change does not meet the preset conditions, the current expressive state remains unchanged; the expressive state at each moment is recorded sequentially to form a state reversal sequence that reflects the process of changes in the brand's AI expressiveness.

[0016] Preferably, the method for outputting the corresponding brand AI performance evaluation result includes: Based on the set of semantic coupling relationships, statistical analysis is performed on the structural features of associations between semantic primitives. By statistically processing the number and density of connections between semantic primitive pairs, the degree of connectivity in the semantic expression process is determined. Based on the effective semantic set, the semantic primitives involved in semantic expression are statistically analyzed, and the ratio of the number of semantic primitives in the effective semantic set to the total number of semantic primitives is calculated to determine the proportion of effective information in the semantic expression content. Based on state flip sequences, the state change characteristics in the semantic expression process are analyzed by serialization. By statistically analyzing the number of state changes, the duration of the state, and the frequency of state switching in the state flip sequences, the dynamic change characteristics of the semantic expression process are determined. The connectivity of the semantic expression process, the proportion of effective information in the semantic expression content, and the dynamic change characteristics are normalized and then weighted and fused based on preset weights to form the corresponding brand AI semantic performance evaluation results.

[0017] A brand AI performance evaluation method based on multi-source data fusion includes: S1. Obtain brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; perform matching and unified identification processing on different semantic primitives to form a brand semantic baseline package; S2. Based on the brand semantic baseline package, construct the association structure between semantic primitives according to the preset semantic pointing relationship and interaction timing relationship to form a set of semantic coupling relationships; S3. Based on the set of semantic coupling relationships, when different semantic primitives act simultaneously, a conflict determination function is introduced to identify conflicts between different semantic primitives; and the semantic primitives with conflicts are suppressed to obtain the effective semantic set after suppression. S4. Construct an expressive state sequence based on the effective semantic set, and determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, trigger the expressive state flip and form a state flip sequence. S5. Based on the semantic coupling relationship set, effective semantic set and state reversal sequence, comprehensively analyze the semantic performance characteristics of brand AI in the performance process, and output the corresponding brand AI performance evaluation results.

[0018] Compared with the prior art, the present invention has the following beneficial effects: By introducing an occupancy state representation mechanism and constructing a conflict determination function based on the target state, this method explicitly identifies and structures the conflict relationships between semantic primitives, addressing the problems of traditional methods relying on empirical judgment and lacking unified standards. An XOR determination logic is introduced to accurately distinguish conflicts at the semantic intent level, improving the accuracy and semantic rationality of conflict identification. Global maximum value aggregation processing is applied to the conflict determination results to unify the constraints on the conflict relationships of semantic primitives, enhancing the global consistency of the results. An occupancy state suppression mechanism is constructed to automatically eliminate or invalidate conflict semantics, preventing conflict semantics from entering subsequent fusion, preventing expression bias and information interference, accurately identifying and automatically resolving conflict semantics, and improving the reliability and accuracy of brand AI performance evaluation results.

[0019] By introducing a unified quantification mechanism, the degree of multidimensional semantic change of adjacent semantic primitives is comprehensively characterized, expanding the dimensions of semantic change characterization and improving the completeness and expressive power of the analysis. A preset change threshold is set to determine the degree of semantic change, and a state reversal trigger function is constructed to map semantic changes into a discrete state switching mechanism, improving the clarity and interpretability of state evolution. XOR operations are used to control the state reversal of performance, avoiding the instability problem of state updates in traditional methods and improving the stability and consistency of state evolution. The performance state at each moment is recorded to form a state reversal sequence, transforming the process of brand AI performance change into a dynamic expression, enhancing the traceability and structured expression capabilities of the evaluation results. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of a brand AI performance evaluation system based on multi-source data fusion according to the present invention; Figure 2 This is a schematic diagram of a brand AI performance evaluation method based on multi-source data fusion according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0022] Please see Figure 1 As shown, this embodiment provides a brand AI performance evaluation system based on multi-source data fusion, specifically including the following steps: The brand semantic construction module is used to acquire brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; it then matches and unifies the identification of different semantic primitives to form a brand semantic baseline package. The semantic relationship construction module is used to construct the association structure between semantic primitives based on the brand semantic benchmark package and according to the preset semantic pointing relationship and interaction timing relationship, forming a set of semantic coupling relationships; The semantic conflict determination module is used to identify conflicts between different semantic primitives when they act simultaneously, based on the set of semantic coupling relationships; and to suppress the semantic primitives that have conflicts, so as to obtain the effective semantic set after suppression. The expressive state determination module is used to construct an expressive state sequence based on the effective semantic set and to determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, the expressive state is triggered to flip and a state flip sequence is formed. The performance evaluation generation module is used to comprehensively analyze the semantic performance characteristics of brand AI in the performance process based on the set of semantic coupling relationships, the effective semantic set, and the state reversal sequence, and output the corresponding brand AI performance evaluation results.

[0023] Methods for generating semantic primitive sets include: Brand-related data from different data sources is obtained through a preset data access interface, and the format of the brand-related data obtained from different data sources is normalized; data cleaning is performed on the normalized brand-related data to obtain original semantic data with a unified structure; based on preset semantic parsing rules, the original semantic data is semantically decomposed and divided into different semantic fragments. In this embodiment, it should be noted that brand-related data from different data sources is obtained through preset data access interfaces. Specifically, the data access interfaces include interfaces for structured data, semi-structured data, and unstructured data, used to access database data, log data, text data, and interaction record data, respectively. During the access process, various types of data are retrieved or pushed according to preset communication protocols. At the access end, data from different sources is uniformly encapsulated and processed, converting the raw data into a unified data description format. The data description format includes a unified field structure and a unified encoding method, thereby completing the data format normalization process between different data sources.

[0024] After data access and format normalization are completed, data cleaning is performed on the normalized brand-related data. Specifically, the data is filtered and processed, including removing duplicate data, eliminating data with missing key fields, filtering data with abnormal formats, and removing noise. Simultaneously, text-based data undergoes symbol normalization, such as standardizing capitalization, removing invalid punctuation and semantically meaningless characters, and performing consistency correction on time and identifier fields. Through these processes, data from different sources are made structurally consistent, resulting in original semantically unified data.

[0025] After obtaining the original semantic data, the original semantic data is semantically decomposed based on preset semantic parsing rules. The semantic parsing rules include sentence segmentation rules based on syntactic analysis and segmentation rules based on semantic boundaries. By parsing the syntactic structure of the text content, continuous data is divided into several semantic segments; for non-text data, semantic units are segmented according to the semantic definitions of its fields, thereby achieving a unified semantic segmentation of the original semantic data.

[0026] During the semantic decomposition process, based on the syntactic structure and semantic role labeling rules, the expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are extracted from each semantic segment. The expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are then combined and encapsulated to form a set of semantic primitives.

[0027] In the semantic decomposition process, semantic elements are extracted from each semantic segment based on syntactic structure and semantic role labeling rules. Specifically, the subject-verb-object structure and modification relationships in the semantic segment are identified to determine the subject information; behavioral information is extracted through verbs or action descriptors; target information is extracted through objects or referents; and feedback information is extracted by combining contextual information or feedback identifiers, thus forming a complete set of semantic elements.

[0028] For example, in a specific application scenario, data from social media comments, brand customer service conversations, and user review logs are obtained through a data access interface. Social media comments are unstructured text data, customer service conversations are semi-structured data, and user review logs are structured data. This data from different sources is then uniformly encapsulated and converted into a unified data description format containing "data source identifier, time identifier, text content field, and user identifier field," thus completing the format normalization process.

[0029] After normalization, the data is cleaned, such as removing duplicate comments, filtering data records containing invalid characters, and standardizing special symbols in the text content to obtain original semantic data with a unified structure.

[0030] Furthermore, the original semantic data is semantically decomposed. For example, for a user review "The brand's customer service responds promptly, but the efficiency in solving problems is average", it is divided into two semantic segments using semantic parsing rules: "customer service responds promptly" and "the efficiency in solving problems is average".

[0031] Semantic elements are extracted based on syntactic structure and semantic role labeling rules: For the semantic fragment "customer service replies promptly", the subject of expression is "customer service", the behavior is "reply", the target is "user problem", and the feedback is "promptly"; For the semantic fragment "problem-solving efficiency is average", the subject of expression is "customer service", the behavior is "solve", the target is "problem", and the feedback is "average efficiency".

[0032] The extracted information on the subject, behavior, target, and feedback of expression are combined and encapsulated to form corresponding semantic primitives, for example, by constructing semantic primitives. =<Customer service, response, user issues, timeliness>, semantic primitive =<Customer service, problem resolution, average efficiency>, and the semantic primitives are uniformly numbered and indexed to build a semantic primitive set.

[0033] Methods for creating a brand semantic benchmark package include: The semantic primitives in the semantic primitive set are matched and judged. Based on the expression subject information, expression behavior information and expression target information in the semantic primitives, the semantic consistency between different semantic primitives is compared. The semantic primitives with consistent semantic orientation are identified according to the preset keyword matching rules. In this embodiment, it should be noted that the preset keyword matching rules are used to determine the consistency of the subject information, behavior information, and target information in the semantic primitives. Specifically, by extracting keywords from each semantic element, the extracted keywords are used as the basic unit for semantic comparison. When performing matching judgment, the keywords are first standardized, including unifying character format and removing invalid symbols, to eliminate the impact of expression differences on the matching results.

[0034] Based on this, keywords in different semantic primitives are matched and compared. The matching comparison includes keyword exact match determination and equivalent match determination. Among them, keyword exact match determination is used to identify the cases where keywords are consistent at the character level; equivalent match determination is used to identify keywords with the same semantic meaning. By using a preset thesaurus or semantic mapping relationship to make corresponding judgments on keywords, keywords with different semantic expressions but the same meaning are judged as matches.

[0035] Furthermore, when the keywords in the expressing subject information, expressing behavior information, and expressing target information respectively meet the matching conditions, it is determined that the corresponding semantic primitives have semantic consistency; if any of the semantic elements does not meet the matching conditions, it is determined that the semantic primitives are semantically inconsistent. Through the above keyword matching rules, the determination of semantic consistency between semantic primitives is achieved.

[0036] For example, for semantic primitives =<Customer service, response, user inquiry, timely> and semantic primitives =<Customer service personnel, response, inquiry, promptness>, first extract the keywords, among which Keywords include "customer service", "response", and "user inquiry". Keywords include "customer service personnel", "response", and "inquiry".

[0037] When performing matching, the keywords are standardized, "customer service personnel" is standardized to "customer service", and "reply" and "response" are determined to be equivalent expressions by using a preset thesaurus, and "user consultation" and "consultation question" are determined to be semantically consistent.

[0038] In the above matching process, since the subject "customer service" matches "customer service personnel", the action "reply" matches "respond", and the target "user inquiry" matches "inquiry question", the semantic primitives are determined. and If the keyword matching rules are met, it can be determined that the two have the same semantic meaning and can be grouped into the same semantic unit.

[0039] Semantic primitives that meet the keyword matching rules are merged, and semantic primitives with consistent semantic orientations are merged into the same semantic unit, while retaining the corresponding expressive feedback information. The merged semantic units are then uniformly identified, with each semantic unit assigned a unique identifier. The semantic units that have completed the uniform identification process are then aggregated to form a brand semantic baseline package.

[0040] A unified identifier is a combination of a data source identifier, a time identifier, and a sequence number. It is used to uniquely identify semantic units and distinguish the order and source differences between different semantic primitives. For example, a semantic unit from social media comment data can be represented by the unified identifier "SM_20260423_001", where "SM" indicates that the data source is social media, "20260423" indicates the generation date, and "001" indicates the sequence number within that time series. This identifier can uniquely determine the source and generation order of the semantic unit.

[0041] Methods for forming a set of semantically coupled relationships include: Based on the brand semantic baseline package, the semantic primitives contained in each semantic unit are extracted, and the semantic primitives are sorted according to the time identifier to form a semantic sequence. Based on the preset semantic pointing relationship, the expression subject information, expression behavior information and expression target information of adjacent semantic primitives in the semantic sequence are compared to identify the semantic response relationship between adjacent semantic primitives. In this embodiment, it should be noted that for any adjacent semantic primitive pair in the semantic sequence, its expressing subject information, expressing behavior information, and expressing target information are extracted respectively, and the above information is compared item by item. During the comparison process, based on the preset semantic pointing relationship, it is determined whether the expressing subjects are consistent or related, whether the expressing behaviors constitute a continuous or responsive relationship, and whether there is a pointing continuation or pointing correspondence relationship between the expressing targets. When adjacent semantic primitives satisfy at least the preset association conditions in the expressing subject, expressing behavior, and expressing target, it is determined that there is a semantic response relationship between the semantic primitive pair.

[0042] The predefined semantic pointing relationship is used to characterize the pointing association rules between different semantic primitives in semantic expression, and is used to determine whether the expression content of one semantic primitive semantically points to or is related to another semantic primitive. Specifically, the semantic pointing relationship includes at least the consistency or correspondence between the expressing subjects, the succession or response between the expressing behaviors, and the pointing continuation or object correspondence between the expressing targets; when at least one of the above relationships meets the predefined conditions, it is determined that there is a semantic pointing relationship between the two semantic primitives.

[0043] For example, semantic primitives =<User, Asks, Question, No Response> and Semantic Primitives =<Customer Service, Reply, Problem, Resolved> in progress The target of expression, "problem", corresponds to The expression target is clear, and the expression behavior "respond" forms a response relationship with "propose," therefore it is determined that... and There is a semantic relationship between them.

[0044] The association between semantic primitives is constrained by the preset interaction sequence relationship, only semantic primitive pairs that satisfy the sequential order are retained, and the association direction of semantic primitive pairs is determined according to the semantic response relationship. It should be noted that the preset interaction temporal relationship is used to characterize the sequential constraints of semantic primitives during the interaction process, in order to determine whether different semantic primitives satisfy a temporal dependency relationship. Specifically, the preset interaction temporal relationship is established based on the time identifier or sequence position information corresponding to the semantic primitives, and is used to limit the association between semantic primitives to conform to the preset first-occurrence-later response sequence logic; when the time identifier of one semantic primitive is earlier than that of another semantic primitive, and the two form a continuous or responsive relationship during the interaction process, it is determined that the two satisfy the interaction temporal relationship.

[0045] For example, semantic primitives =<User, Question, No Response> time marker is Semantic primitives =<Customer Service, Reply, Problem, Resolved> time markers are ,when Earlier ,and right When a semantic response relationship is formed, the determination is made. and The interaction sequence must be satisfied; if the time order is reversed, the preset interaction sequence must not be satisfied.

[0046] Semantic primitive pairs that satisfy semantic response relationships and interaction temporal relationships are summarized, and the association structure between semantic primitives is constructed. The association structure is then standardized and aggregated to form a set of semantic coupling relationships.

[0047] The semantic primitives in the association structure are processed in a unified relational expression manner. Each pair of semantic primitives with an association relationship is encapsulated according to a preset relational description format, so that it contains the semantic primitive of the association start point, the semantic primitive of the association end point, and the corresponding association type information, thereby forming a standardized semantic association unit.

[0048] After unifying the relational expressions, the semantic association units are standardized. Specifically, duplicate semantic association units are merged, and redundant semantic association units are filtered and simplified, retaining only those that satisfy the constraints of semantic response relationships and interaction timing relationships to ensure the consistency and effectiveness of semantic relational expressions. Furthermore, the standardized semantic association units are organized according to a unified data structure and aggregated to form a set of semantically coupled relationships.

[0049] Methods for conflict identification of different semantic primitives include: Extract semantic primitives that are simultaneously occupied from the set of semantic coupling relationships, and characterize the occupancy state of each semantic primitive. Occupancy status representation of each semantic primitive: ;in, express The occupancy status is used to characterize the role of the semantic primitive in the current set of semantic coupling relationships; when the semantic primitive is in the occupancy status, the value is 1; when the semantic primitive does not participate in the current semantic action or is in the non-occupancy status, the value is 0. Represents the first in the set of semantic primitives A semantic primitive; Index identifiers representing semantic primitives; Construct a semantic conflict determination function to combine and determine any two semantic primitives that are in an active state. When two semantic primitives are simultaneously in an occupied state and their respective expression target states are different, it is determined that there is a semantic conflict relationship between the two semantic primitives. Semantic conflict determination function: ;in, Representing semantic primitives With semantic primitives The conflict determination result between them is used to characterize whether there is a semantic conflict relationship between them; when When, it indicates the existence of a semantic conflict; when When there is no semantic conflict, it indicates that there is no semantic conflict. Represents the first in the set of semantic primitives A semantic primitive; The AND operation represents a relationship where multiple conditions must be met simultaneously; in this embodiment, it is used to limit the two semantic primitives to be in an occupied state at the same time. The XOR operation is used to determine the difference between two semantic primitives in terms of their expression goals. When the expression goals of the two semantic primitives are the same, the result of the operation is 0; when the expression goals are different, the result of the operation is 1. express The expression target state represents the semantic state of the semantic primitive on its expression target, and the expression target state is obtained by mapping the corresponding semantic target content; Representing semantic primitives The target state is expressed.

[0050] When two semantic primitives are not simultaneously occupied, or when two semantic primitives are simultaneously occupied and their respective expression target states are consistent, it is determined that there is no semantic conflict relationship between the two semantic primitives; the semantic primitive pairs that meet the conflict determination conditions are marked to form a conflict semantic set.

[0051] Methods for obtaining the suppressed effective semantic set include: Based on the conflict semantic set, for each semantic primitive, the conflict relationship between the semantic primitive and other semantic primitives is obtained, and the corresponding maximum conflict value is determined; the occupancy state of each semantic primitive is suppressed. When a semantic primitive has a semantic conflict relationship with any other semantic primitive, the occupancy state of the semantic primitive is suppressed to an invalid state. Suppress the occupancy state of each semantic primitive: ;in, Representing semantic primitives The occupancy state after conflict suppression is used to characterize the validity result after conflict resolution; when the semantic primitive remains valid after suppression, its value is 1, otherwise it is 0; Representing semantic primitives The maximum value of the conflict determination results among all related semantic primitives is used to characterize whether the semantic primitive has a conflict relationship with any other semantic primitive; the maximum value is 1 when there is at least one conflict relationship, and 0 otherwise; When a semantic primitive does not have a semantic conflict with other semantic primitives, the occupancy state of the semantic primitive remains unchanged. The semantic primitives after suppression are screened, and semantic primitives with valid occupancy state are retained, while semantic primitives with invalid occupancy state are removed, thus forming a set of valid semantic primitives after suppression.

[0052] Methods for constructing expressive state sequences based on effective semantic sets include: Based on the effective semantic set after conflict suppression, the time markers corresponding to each semantic primitive are extracted, and the semantic primitives are sorted according to the time markers to form an ordered semantic sequence. The state of each semantic primitive in the ordered semantic sequence is represented. Based on the occupancy state, expression behavior information and expression target state of the semantic primitive, each semantic primitive is mapped to a corresponding expressive state unit. The expressive state units are then organized in series according to the time order to form an expressive state sequence.

[0053] In this embodiment, it should be noted that after obtaining the ordered semantic sequence, state mapping processing is performed on each semantic primitive in the sequence to generate corresponding expressive state units. Specifically, for any semantic primitive, its occupancy state, expressive behavior information, and expressive target state are extracted as the basic elements for constructing the expressive state unit. During the state mapping process, it is first determined whether the semantic primitive participates in the construction of the current expressive state based on its occupancy state; when the semantic primitive is in a valid occupancy state, it is included in the state mapping process; when the semantic primitive is in a non-occupancy state, it can be marked as an invalid state or ignored during the mapping process.

[0054] By combining the expressive behavior information and the target state of semantic primitives, state encoding is performed on their semantic expression features. Specifically, the expressive behavior information is used as the behavioral feature representing the semantic action mode, and the target state is used as the target feature representing the semantic orientation. The two are combined to form a state representation that reflects the expressive features of the semantic primitive. The state representation is used to reflect the expression direction and action form of the semantic primitive at the current moment. The occupancy state and the state representation are fused to generate a corresponding expressive state unit. The expressive state unit is used to uniformly represent the expressive behavior features and target orientation features of the semantic primitive under the occupancy condition, thereby forming a structured state description result.

[0055] Methods for forming state-flipping sequences include: In the expressive state sequence, the expressive state units corresponding to adjacent effective semantic primitives are compared. Based on the occupancy state, expressive behavior information and expressive target state of the semantic primitives, the degree of semantic change between adjacent effective semantic primitives is determined. The degree of semantic change between adjacent semantic primitives: ;in, Indicates at time The degree of semantic change between adjacent effective semantic primitives is used to characterize the intensity of changes in semantic expression over continuous time; Indicates time Semantic primitives The occupancy status; Indicates at time Semantic primitives The occupancy status is used to characterize the occupancy status of the semantic primitive in the current semantic process, where the value is 1 when it is in the occupancy status and 0 when it is in the non-occupancy status; This represents the absolute value operation, used to represent two occupancy states. and The degree of difference between them; A function representing the difference between expressive behavioral information, used to characterize the degree of change in behavioral features between adjacent semantic primitives; This represents a difference measurement function between target states, used to characterize the degree of change in target orientation between adjacent semantic primitives; Indicates time Semantic primitives Corresponding expressive behavioral information; Indicates at time Semantic primitives Corresponding expressive behavioral information; Indicates time Semantic primitives The corresponding target state; Indicates at time Semantic primitives The expression of the target state; An index identifier indicating the time; In this embodiment, it should be noted that the degree of semantic change... This represents the dimensionless representation result after uniformly quantizing semantic features across different dimensions. Specifically, it represents the difference in behavioral information. and the difference in the target state All calculations employ a normalized difference measure to ensure that the outputs are constrained to a uniform numerical scale; simultaneously, the state difference term is used. Since its value is only 0 or 1, it is a dimensionless discrete difference quantity. Therefore, the three difference terms have a consistent dimensional basis at the numerical level, thus avoiding the incomparability problem caused by the inconsistency of dimensions between different dimensions.

[0056] It is used to characterize the degree of difference between adjacent semantic primitives in expressing behavioral information. The method of obtaining it is: after processing the semantic feature vectorization of the behavioral information, it is obtained by using the vector space distance calculation method, such as calculating the degree of offset between the behavioral feature vectors at two time points based on cosine distance or Euclidean distance, so as to obtain the corresponding difference measurement result.

[0057] It is used to characterize the degree of difference between adjacent semantic primitives in expressing target states. The method of obtaining it is as follows: after semantic encoding or category mapping of the target state, it is obtained by category consistency comparison or semantic distance calculation, such as consistency judgment based on encoding results or distance calculation based on semantic embedding space, to obtain the degree of deviation between different target states, thereby forming the corresponding difference measurement result.

[0058] By using the above methods, the degree of semantic change can be comprehensively represented under the premise that all components are dimensionless and uniform in scale, thereby ensuring that different semantic dimensions can be compared and calculated under the same evaluation system.

[0059] The semantic change degree between adjacent valid semantic primitives is compared with a preset change threshold. When the semantic change degree meets the preset change threshold, a state reversal trigger function is constructed based on the semantic change degree to trigger the expressive state reversal, thus reversing the current expressive state. Switching from the first state to the second state or from the second state to the first state; State toggle trigger function: ;in, This represents the state reversal trigger function, used to generate a trigger signal based on the degree of semantic change; when When the value exceeds the preset change threshold, the state reversal trigger function takes a value of 1; otherwise, it takes a value of 0. This indicates a preset change threshold, used to determine the degree of semantic change; Triggering an expression state flip: ;in, Indicates at time The expressive state is used to characterize the state outcome at the next moment after the current semantic change. Indicates at time The expressive state is used to characterize the expressive state of the brand AI in the current semantic expression stage. Its value is a binary state, corresponding to the first state or the second state respectively. This represents the XOR operation, used to implement state flip control; when the state flip trigger function is set to 1, the current state flips; when the state flip trigger function is set to 0, the current state remains unchanged.

[0060] When the degree of semantic change does not meet the preset conditions, the current expressive state remains unchanged; the expressive state at each moment is recorded sequentially to form a state reversal sequence that reflects the process of changes in the brand's AI expressiveness.

[0061] Methods for outputting corresponding brand AI performance evaluation results include: Based on the set of semantic coupling relationships, statistical analysis is performed on the structural features of associations between semantic primitives. By statistically processing the number and density of connections between semantic primitive pairs, the degree of connectivity in the semantic expression process is determined. Based on the effective semantic set, the semantic primitives involved in semantic expression are statistically analyzed, and the ratio of the number of semantic primitives in the effective semantic set to the total number of semantic primitives is calculated to determine the proportion of effective information in the semantic expression content. Based on state flip sequences, the state change characteristics in the semantic expression process are analyzed by serialization. By statistically analyzing the number of state changes, the duration of the state, and the frequency of state switching in the state flip sequences, the dynamic change characteristics of the semantic expression process are determined. The connectivity of the semantic expression process, the proportion of effective information in the semantic expression content, and the dynamic change characteristics are normalized and then weighted and fused based on preset weights to form the corresponding brand AI semantic performance evaluation results.

[0062] In this embodiment, it should be noted that the evaluation process of the brand AI semantic performance characteristics is a multi-dimensional comprehensive analysis process based on the constructed set of semantic coupling relationships, effective semantic set and state reversal sequence.

[0063] Specifically, based on the semantic coupling relationship set, statistical analysis is performed on the structural features of the associations between semantic primitives to determine the connectivity of the semantic expression process. The semantic coupling relationship set is used to characterize the association paths and connection directions between semantic primitives. By statistically processing the number and density of connections between semantic primitive pairs, a structural index reflecting the tightness of the overall semantic organization structure is obtained, and this structural index is used as the characterization result of the connectivity of the semantic expression process.

[0064] Furthermore, based on the effective semantic set, statistical analysis is performed on the semantic primitives involved in semantic expression to determine the proportion of effective information in the semantic expression content. Specifically, by calculating the ratio between the number of semantic primitives in the effective semantic set and the total number of semantic primitives, the degree of effective semantic coverage is obtained, thereby characterizing the proportion of effective information in the overall semantic expression content.

[0065] Furthermore, based on the state flip sequence, a serialization analysis is performed on the state change characteristics in the semantic expression process to determine the dynamic change characteristics of the semantic expression process. By statistically analyzing the number of state changes, state duration, and state switching frequency in the state flip sequence, dynamic characteristic indicators reflecting the intensity and stability of dynamic changes in the semantic expression process are formed.

[0066] After obtaining the connectivity, effective information ratio and dynamic change characteristics mentioned above, each feature index is normalized to map it to a unified numerical scale range to eliminate the dimensional differences between different feature dimensions. Then, based on preset weights, the normalized feature indexes are weighted and fused to generate a comprehensive evaluation result, which serves as the corresponding brand AI semantic performance evaluation result. This result is used to characterize the overall performance capability and comprehensive semantic performance level of brand AI in the semantic expression process.

[0067] For example, suppose a brand's AI generates three semantic expressions when answering a user's question: Semantic primitive sequence: Introduce product features; Additional use cases; : Responding to user questions; Connectivity (set of semantic coupling relationships): Related (functional) Scene); Related (Scene) Response); Number of connections = 2; Connection density = "Almost every step is related"; Result: Connectivity = High (indicating that the expression is coherent).

[0068] Percentage of effective information (effective semantic set): Original semantic primitives: 3; Valid semantic primitives: 3; Ratio: Effective percentage = 3 / 3 = 100%; Result: The percentage of effective information is high (indicating the absence of noise).

[0069] Dynamic change characteristics (state flip sequence): The state changes are as follows: :introduce; Additional notes (flip once); : Continue explaining (stable); Statistics show that: the number of state transitions is 1; the stable state interval is relatively long; and the switching frequency is moderate. Result: Dynamic characteristics = "variable but not excessively abrupt" (reasonable rhythm).

[0070] Based on the preset weights, the normalized feature indicators are weighted and fused to generate a comprehensive evaluation result: The system unifies the three indicators to between 0 and 1: connectivity 0.9, effective proportion 1.0, dynamic feature: 0.7; preset weights: connectivity 0.4, effectiveness 0.4, dynamic feature 0.2; brand AI performance evaluation result: 0.9×0.4+1.0×0.4+0.7×0.2=0.88.

[0071] The preset change threshold is set by staff based on the results of historical data analysis. This historical analysis process includes the system collecting the degree of semantic change between multiple adjacent valid semantic primitives and calculating their average value as a reference to obtain the preset change threshold. During system operation, the staff adjusts the threshold according to the actual situation.

[0072] This embodiment introduces an occupancy state representation mechanism and constructs a conflict determination function based on the expressed target state. This explicitly identifies and structures the semantic primitive conflict relationships, addressing the problems of traditional methods relying on empirical judgment and lacking unified standards. An XOR determination logic is introduced to accurately distinguish conflicts at the semantic intent level, improving the accuracy and semantic rationality of conflict identification. Global maximum value aggregation processing is applied to the conflict determination results to unify the constraints on the semantic primitive conflict relationships, enhancing the global consistency of the results. An occupancy state suppression mechanism is constructed to automatically eliminate or invalidate conflict semantics, preventing conflict semantics from entering subsequent fusion, preventing expression bias and information interference, accurately identifying and automatically resolving conflict semantics, and improving the reliability and accuracy of brand AI performance evaluation results.

[0073] By introducing a unified quantification mechanism, the degree of multidimensional semantic change of adjacent semantic primitives is comprehensively characterized, expanding the dimensions of semantic change characterization and improving the completeness and expressive power of the analysis. A preset change threshold is set to determine the degree of semantic change, and a state reversal trigger function is constructed to map semantic changes into a discrete state switching mechanism, improving the clarity and interpretability of state evolution. XOR operations are used to control the state reversal of performance, avoiding the instability problem of state updates in traditional methods and improving the stability and consistency of state evolution. The performance state at each moment is recorded to form a state reversal sequence, transforming the process of brand AI performance change into a dynamic expression, enhancing the traceability and structured expression capabilities of the evaluation results. Example

[0074] Please see Figure 2 As shown, the parts not described in detail in this embodiment are described in Embodiment 1. A brand AI performance evaluation method based on multi-source data fusion is provided, including: S1. Obtain brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; perform matching and unified identification processing on different semantic primitives to form a brand semantic baseline package; S2. Based on the brand semantic baseline package, construct the association structure between semantic primitives according to the preset semantic pointing relationship and interaction timing relationship to form a set of semantic coupling relationships; S3. Based on the set of semantic coupling relationships, when different semantic primitives act simultaneously, a conflict determination function is introduced to identify conflicts between different semantic primitives; and the semantic primitives with conflicts are suppressed to obtain the effective semantic set after suppression. S4. Construct an expressive state sequence based on the effective semantic set, and determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, trigger the expressive state flip and form a state flip sequence. S5. Based on the semantic coupling relationship set, effective semantic set and state reversal sequence, comprehensively analyze the semantic performance characteristics of brand AI in the performance process, and output the corresponding brand AI performance evaluation results. Example

[0075] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the brand AI performance evaluation system based on multi-source data fusion described above.

[0076] Since the electronic device described in this embodiment is used to implement the brand AI performance evaluation system and method based on multi-source data fusion described in this application, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the brand AI performance evaluation system and method based on multi-source data fusion described in this application. Therefore, how the electronic device implements the method in this application will not be described in detail here. Any electronic device used by those skilled in the art to implement the brand AI performance evaluation system and method based on multi-source data fusion described in this application falls within the scope of protection of this application.

[0077] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0078] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A brand AI performance evaluation system based on multi-source data fusion, characterized in that, include: The brand semantic construction module is used to acquire brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; Different semantic primitives are matched and uniformly identified to form a brand semantic benchmark package; The semantic relationship construction module is used to construct the association structure between semantic primitives based on the brand semantic benchmark package and according to the preset semantic pointing relationship and interaction timing relationship, forming a set of semantic coupling relationships; The semantic conflict determination module is used to identify conflicts between different semantic primitives when they act simultaneously, based on the set of semantic coupling relationships and by introducing a conflict determination function. The method for conflict identification of different semantic primitives includes: Extract semantic primitives that are simultaneously occupied from the set of semantic coupling relationships and characterize the occupancy state of each semantic primitive; construct a semantic conflict determination function to combine and determine any two semantic primitives that are in an active state. When two semantic primitives are simultaneously occupied and their respective expression target states are different, it is determined that there is a semantic conflict relationship between the two semantic primitives. Semantic conflict determination function: ;in, Representing semantic primitives With semantic primitives The conflict determination results between them; express The occupancy status; Represents the first in the set of semantic primitives A semantic primitive; This represents the logical AND operation; Represents the XOR operation; express The expression of the target state; Representing semantic primitives The expression of the target state; When two semantic primitives are not simultaneously occupied, or when two semantic primitives are simultaneously occupied and their respective expression target states are consistent, it is determined that there is no semantic conflict relationship between the two semantic primitives; the semantic primitive pairs that meet the conflict determination conditions are marked to form a conflict semantic set. Then, the semantic primitives that conflict are suppressed to obtain the effective semantic set after suppression; The method for obtaining the suppressed effective semantic set includes: Based on the conflict semantic set, for each semantic primitive, the conflict relationship between the semantic primitive and other semantic primitives is obtained, and the corresponding maximum conflict value is determined; the occupancy state of each semantic primitive is suppressed. When a semantic primitive has a semantic conflict relationship with any other semantic primitive, the occupancy state of the semantic primitive is suppressed to an invalid state. Suppress the occupancy state of each semantic primitive: ;in, Representing semantic primitives Occupancy status after conflict suppression; Representing semantic primitives The maximum value of the conflict determination result among all related semantic primitives; When a semantic primitive does not have a semantic conflict with other semantic primitives, the occupancy state of the semantic primitive remains unchanged; the semantic primitives after suppression are screened, and semantic primitives with valid occupancy state are retained, while semantic primitives with invalid occupancy state are removed, thus forming a set of valid semantic primitives after suppression. The expressive state determination module is used to construct an expressive state sequence based on the effective semantic set and to determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, the expressive state is triggered to flip and a state flip sequence is formed. The performance evaluation generation module is used to comprehensively analyze the semantic performance characteristics of brand AI in the performance process based on the set of semantic coupling relationships, the effective semantic set, and the state reversal sequence, and output the corresponding brand AI performance evaluation results.

2. The brand AI performance evaluation system based on multi-source data fusion according to claim 1, characterized in that, The method for generating the set of semantic primitives includes: Brand-related data from different data sources is obtained through a preset data access interface, and the format of the brand-related data obtained from different data sources is normalized; data cleaning is performed on the normalized brand-related data to obtain original semantic data with a unified structure; based on preset semantic parsing rules, the original semantic data is semantically decomposed and divided into different semantic fragments. During the semantic decomposition process, based on the syntactic structure and semantic role labeling rules, the expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are extracted from each semantic segment. The expressive subject information, expressive behavior information, expressive target information, and expressive feedback information are then combined and encapsulated to form a set of semantic primitives.

3. The brand AI performance evaluation system based on multi-source data fusion according to claim 2, characterized in that, The method for forming the brand semantic benchmark package includes: The semantic primitives in the semantic primitive set are matched and judged. Based on the expression subject information, expression behavior information and expression target information in the semantic primitives, the semantic consistency between different semantic primitives is compared. The semantic primitives with consistent semantic orientation are identified according to the preset keyword matching rules. Semantic primitives that meet the keyword matching rules are merged, and semantic primitives with consistent semantic orientations are merged into the same semantic unit, while retaining the corresponding expressive feedback information. The merged semantic units are then uniformly identified, with each semantic unit assigned a unique identifier. The semantic units that have completed the uniform identification process are then aggregated to form a brand semantic baseline package.

4. The brand AI performance evaluation system based on multi-source data fusion according to claim 3, characterized in that, The method for forming a set of semantic coupling relationships includes: Based on the brand semantic baseline package, the semantic primitives contained in each semantic unit are extracted, and the semantic primitives are sorted according to the time identifier to form a semantic sequence. Based on the preset semantic pointing relationship, the expression subject information, expression behavior information and expression target information of adjacent semantic primitives in the semantic sequence are compared to identify the semantic response relationship between adjacent semantic primitives. The association between semantic primitives is constrained by the preset interaction time sequence relationship. Only semantic primitive pairs that satisfy the sequential order are retained, and the association direction of semantic primitive pairs is determined according to the semantic response relationship. Semantic primitive pairs that satisfy the semantic response relationship and the interaction time sequence relationship are summarized to construct the association structure between semantic primitives. The association structure is then standardized and aggregated to form a set of semantic coupling relationships.

5. A brand AI performance evaluation system based on multi-source data fusion as described in claim 4, characterized in that, The method for constructing expressive state sequences based on effective semantic sets includes: Based on the effective semantic set after conflict suppression, the time markers corresponding to each semantic primitive are extracted, and the semantic primitives are sorted according to the time markers to form an ordered semantic sequence. The state of each semantic primitive in the ordered semantic sequence is represented. Based on the occupancy state, expression behavior information and expression target state of the semantic primitive, each semantic primitive is mapped to a corresponding expressive state unit. The expressive state units are then organized in series according to the time order to form an expressive state sequence.

6. A brand AI performance evaluation system based on multi-source data fusion as described in claim 5, characterized in that, The method for forming the state-flipping sequence includes: In the expressive state sequence, the expressive state units corresponding to adjacent effective semantic primitives are compared. Based on the occupancy state, expressive behavior information and expressive target state of the semantic primitives, the degree of semantic change between adjacent effective semantic primitives is determined. The degree of semantic change between adjacent valid semantic primitives is compared with a preset change threshold. When the degree of semantic change meets the preset change threshold, a state flip trigger function is constructed based on the degree of semantic change to trigger the expressive state flip, so that the current expressive state is switched from the first state to the second state or from the second state to the first state. When the degree of semantic change does not meet the preset conditions, the current expressive state remains unchanged; the expressive state at each moment is recorded sequentially to form a state reversal sequence that reflects the process of changes in the brand's AI expressiveness.

7. A brand AI performance evaluation system based on multi-source data fusion as described in claim 6, characterized in that, The method for outputting the corresponding brand AI performance evaluation results includes: Based on the set of semantic coupling relationships, statistical analysis is performed on the structural features of associations between semantic primitives. By statistically processing the number and density of connections between semantic primitive pairs, the degree of connectivity in the semantic expression process is determined. Based on the effective semantic set, the semantic primitives involved in semantic expression are statistically analyzed, and the ratio of the number of semantic primitives in the effective semantic set to the total number of semantic primitives is calculated to determine the proportion of effective information in the semantic expression content. Based on state flip sequences, the state change characteristics in the semantic expression process are analyzed by serialization. By statistically analyzing the number of state changes, the duration of the state, and the frequency of state switching in the state flip sequences, the dynamic change characteristics of the semantic expression process are determined. The connectivity of the semantic expression process, the proportion of effective information in the semantic expression content, and the dynamic change characteristics are normalized and then weighted and fused based on preset weights to form the corresponding brand AI semantic performance evaluation results.

8. A brand AI performance evaluation method based on multi-source data fusion, implemented by a brand AI performance evaluation system based on multi-source data fusion as described in any one of claims 1 to 7, characterized in that, include: S1. Obtain brand-related data from different data sources and perform semantic decomposition to generate a set of semantic primitives; Different semantic primitives are matched and uniformly identified to form a brand semantic benchmark package; S2. Based on the brand semantic baseline package, construct the association structure between semantic primitives according to the preset semantic pointing relationship and interaction timing relationship to form a set of semantic coupling relationships; S3. Based on the set of semantic coupling relationships, when different semantic primitives act simultaneously, a conflict determination function is introduced to identify conflicts between different semantic primitives; and the semantic primitives with conflicts are suppressed to obtain the effective semantic set after suppression. S4. Construct an expressive state sequence based on the effective semantic set, and determine the degree of semantic change between adjacent effective semantic primitives. When the degree of semantic change meets the preset change threshold, trigger the expressive state flip and form a state flip sequence. S5. Based on the semantic coupling relationship set, effective semantic set and state reversal sequence, comprehensively analyze the semantic performance characteristics of brand AI in the performance process, and output the corresponding brand AI performance evaluation results.

Images