Bias ai text recognition system based on deep semantic contrast and recognition method thereof

Abstract

The application relates to the technical field of computer information processing, and discloses a deviation AI text recognition system based on deep semantic comparison and a recognition method thereof, which comprises a memory, a processor and a computer program stored in the memory and running on the processor. The system is started based on a duplicate record rule parameter in a duplicate record rule, after a similarity score of each candidate record in data is obtained by executing a duplicate algorithm in an asynchronous job, a classification label is generated according to a preset'recognition result level' configuration in the asynchronous job; and the classification label is referenced to a many-to-many intermediate object of the asynchronous job through a comparison result in the'recognition result level', each result is associated with the'recognition result level' and initialized with 'whether to display' a duplicate result display. The application is configurable, intelligent and platformized, has strong semantic understanding ability, is highly configurable, can realize bidirectional relationship visualization, task traceability, algorithm pluggability and state intelligent management.

Description

Technical Field

[0001] This invention relates to the field of computer information processing technology, and specifically to a deviation AI text recognition system and method based on deep semantic comparison. Background Technology

[0002] In modern enterprise information systems, with the rapid growth of business data, there are a large number of redundant records with similar but not identical content within the same type of business object (such as customers, work orders, deviation reports, etc.). Traditional deduplication mechanisms usually rely on exact field matching or simple fuzzy matching (such as LIKE queries and Levenshtein distance), which are difficult to handle situations where natural language descriptions differ greatly but semantics are similar, resulting in high false negative rates and serious misjudgments.

[0003] Existing technologies employ models such as TF-IDF, Word2Vec, or BERT for text similarity calculation, but the following problems remain: First, a lack of flexible business integration capabilities: most AI plagiarism detection functions are independent modules, unable to be deeply integrated with the object model, form layout, and lifecycle actions of business systems; Second, coarse-grained configuration: it does not support defining differentiated plagiarism detection rules at the object level, nor can it dynamically select the combination of fields to be compared; Third, unintuitive result display: plagiarism detection results are often presented in list form, without establishing a two-way association view between source records and candidate duplicate records; Fourth, lack of status management: there is a lack of mechanisms for historical tracking, conclusion confirmation, and visibility control of plagiarism detection tasks, affecting the efficiency of manual review; Fifth, poor scalability: the algorithm coupling is high, making it difficult to switch between different AI engines or local vectorized models (such as bge-m3).

[0004] Therefore, there is an urgent need for a biased AI text recognition system and method based on deep semantic comparison that can both utilize advanced AI semantic understanding capabilities and be deeply integrated into low-code / business platform architectures. Summary of the Invention

[0005] This invention aims to provide a deviation AI text recognition system and method based on deep semantic comparison to solve the problems of weak semantic understanding, difficulty in business integration, and lack of traceability of results in traditional plagiarism detection technologies. By introducing a Large Language Model (LLM) or a dedicated embedding model for semantic similarity analysis, and combining it with an object model-driven design approach, a universal plagiarism detection capability for any business object is achieved.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: 1) A biased AI text recognition system based on deep semantic comparison, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to perform the following steps: The system collects and queries whether there are recent plagiarism check rules. If the query result is "yes", it will enter the plagiarism check rule and view the parameters of the plagiarism check rule. The task can continue to perform plagiarism check with the parameters of the plagiarism check rule, or the parameters of the plagiarism check rule can be adjusted. Conversely, if the query result is "no", it will enter the plagiarism check rule settings and set the parameters of the plagiarism check rule. An asynchronous job is started based on the parameters of the deduplication record rules. After the deduplication algorithm in the asynchronous job is executed to obtain the similarity score of each candidate record in the data, a classification label is generated according to the preset "identification result level" configuration in the asynchronous job. The classification label is then referenced to the many-to-many intermediate object of the asynchronous job through the comparison results in the "identification result level". Each result is associated with the "identification result level" and the "display" setting is initialized to show the deduplication result.

[0007] This invention utilizes data collection and querying to identify recent deduplication rules, enabling traceability of task deduplication. In enterprise-level application platforms, data traceability is required for scenarios such as master data management, work order deduplication, and customer information merging. Leveraging the traceability feature of this invention, the previous deduplication data can be viewed each time a deduplication check is performed. This allows for re-checking and reproducing the deduplication data, and also enables the continuation of task deduplication or adjustment of deduplication record rule parameters. This significantly improves data management efficiency compared to resetting the deduplication record rule parameters.

[0008] Meanwhile, based on the valuable deduplication record rule parameters, asynchronous jobs are started. The deduplication algorithm in the asynchronous job can significantly improve the recall efficiency and accuracy. At the same time, the many-to-many intermediate objects in the asynchronous job are used to realize the bidirectional display of the deduplication structure, which improves the user experience. In addition, each result is associated with the "identification result level" and the "whether to display" of the deduplication result is initialized. The logic of displaying and hiding historical results is automatically handled to avoid information redundancy and interference.

[0009] 2) According to the deviation AI text recognition system based on deep semantic comparison described in 1), the deduplication record rules include the following steps: Step 1: Identify the object to be checked, its activation status, and its description, and set the data filtering range conditions, which include the time range field, time unit, and numerical value. Step 2: Based on the data filtering range conditions in Step 1, determine whether the initiated task is allowed to be initiated. If the determination is "no", then reselect the data filtering range conditions. If the determination is "yes", then save it to the database table and its related sub-tables to form structured configuration data.

[0010] 3) According to the deviation AI text recognition system based on deep semantic comparison described in 2), wherein: When the viewing record rules are designed with multiple "matching settings" groups, each group contains a set of fields for semantic comparison. The field set includes a list of display fields, a configuration of recognition result level, and an upper limit on the number of results.

[0011] 4) According to the deviation AI text recognition system based on deep semantic comparison described in 3), wherein: The identification result level configuration includes name, minimum similarity threshold, and whether to display; the deduplication record rule parameters include matching settings, data range, and comparison fields.

[0012] 5) According to the deviation AI text recognition system based on deep semantic comparison described in 2), wherein: After the deduplication record rules form structured configuration data, a corresponding self-referencing many-to-many intermediate object is automatically created for the task object to store the deduplication comparison results; The intermediate object contains the following fields: Target instance: The original record from which the plagiarism check was initiated; Comparison example: Records identified as potentially duplicates; Similarity score: a floating-point number between 0 and 1; Recognition result type: Preset option set, including "high similarity", "medium similarity" and "low similarity"; This is a record of plagiarism checks. Display Flag: Controls whether the result is visible in the relevant list on the front end.

[0013] 6) According to the deviation AI text recognition system based on deep semantic comparison described in 1), after entering the deduplication record rule setting and setting the deduplication record rule parameters, a deduplication history record is created and stored in the deduplication history object table. The metadata of the deduplication history object table includes: The following information is required: referenced plagiarism detection rules; matching settings used; ID of the instance being compared; range fields and values; JSON format of the matching fields; initiator; initiation time; and task status.

[0014] 7) According to the deviation AI text recognition system based on deep semantic comparison described in 1), the asynchronous task of executing the deduplication algorithm specifically includes: Step 1: Extract the dataset to be compared from the target objects according to the configured time range and filtering conditions; Step 2: Concatenate the current instance with the specified fields of each record in the candidate set to form a natural language text fragment; Step 3: Calculate semantic similarity using any of the following methods: AI large-scale model reasoning: The current instance content is used as the query sentence, the candidate set is used to build knowledge base documents in batches, and the large-scale model API is called to perform semantic retrieval and scoring; Vector embedding comparison: The text is converted into vectors using a locally deployed embedding model, and the cosine similarity ranking is completed through a vector database or in-memory computation; Third-party interfaces: Call external similarity service interfaces such as Baidu AI Open Platform to compare each item one by one.

[0015] 8) According to the deviation AI text recognition system based on deep semantic comparison described in 1), wherein: The display of plagiarism detection results includes the following steps: Step 1: When the user clicks the "Check for Plagiarism" button again, the system returns a summary and detailed list of the results of the latest plagiarism check task. The plagiarism check results are displayed in the form through two special types of "parts". Step 2: Users can set the "identification conclusion" for each matching result in the plagiarism detection results pop-up window, and mark it as "confirmed duplicate" or "not duplicate"; Step 3: After receiving the identification results, the system updates the visibility status of the corresponding intermediate object record: If multiple results in a deduplication check point to the same pair of records, only the latest confirmed result will be visible, and the old results will be automatically hidden; records that have been confirmed as duplicates can be processed in the business process.

[0016] This invention supports multiple languages, list layout configuration, access control and other functions to ensure that the system can adapt to complex enterprise environments.

[0017] 9) According to the deviation AI text recognition system based on deep semantic comparison described in 8), the plagiarism detection results are displayed in the form through two special types of "parts," wherein the "parts" include: The duplicate detection initiation record section displays all the duplicate detection history of this record as a "target instance" in the form of an inbound relationship list. Clicking on it will bring up the details. Duplicate identification results section: Displays all associations identified for this record as a "comparison instance" in the form of a many-to-many relationship list. Viewing is supported only; manual addition, deletion, and modification are prohibited.

[0018] 10) A biased AI text recognition method based on deep semantic contrast, applied to any one of the biased AI text recognition systems based on deep semantic contrast described in 1)-9) above, comprising the following steps: Step 1: Collect and query whether there are recent plagiarism check rules. If the query result is "yes", then enter the plagiarism check rule and view the parameters of the plagiarism check rule. Continue to perform the task plagiarism check with the parameters of the plagiarism check rule, or adjust the parameters of the plagiarism check rule. Conversely, if the query result is "no", then enter the plagiarism check rule settings and set the parameters of the plagiarism check rule. Step 2: Start the asynchronous job based on the deduplication record rule parameters in the deduplication record rules. After the deduplication algorithm in the asynchronous job obtains the similarity score of each candidate record in the data, generate classification labels according to the preset "identification result level" configuration in the asynchronous job. Then, reference the classification labels to the many-to-many intermediate object of the asynchronous job through the comparison results in the "identification result level". Associate each result with the "identification result level" and initialize the "display" setting to show the deduplication result display.

[0019] This invention also has the following technical advantages over the prior art: This invention is configurable, intelligent, and platform-based, featuring strong semantic understanding capabilities, high configurability, bidirectional relationship visualization, task traceability, pluggable algorithms, and intelligent state management. Specifically, it utilizes large models or high-quality embedded models to achieve deep semantic comparison, significantly improving recall and precision, thus achieving strong semantic understanding capabilities. It supports flexible definition of deduplication rules by object, field, and matching strategy to meet diverse business needs, thus achieving high configurability. It achieves bidirectional display of deduplication results through self-referencing many-to-many intermediate objects, enhancing user experience and thus achieving bidirectional relationship visualization. It fully records the task parameters, execution time, and comparison results for each deduplication check, facilitating auditing and review, thus achieving task traceability. It supports multiple similarity calculation methods, facilitating future upgrades or replacements of the underlying AI engine, thus achieving pluggable algorithms. It automatically handles the explicit and implicit logic of historical results, avoiding information redundancy and interference, thus achieving intelligent state management. Attached Figure Description

[0020] Figure 1 This is a framework diagram of the bias AI text recognition system based on deep semantic comparison of the present invention; Figure 2 This is a flowchart of the deviation AI text recognition method based on deep semantic comparison of the present invention. Detailed Implementation

[0021] The following detailed description illustrates the specific implementation method: See the example. Figure 1 and Figure 2 A biased AI text recognition system based on deep semantic comparison includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps: The system collects and queries whether there are recent plagiarism check rules. If the query result is "yes", it will enter the plagiarism check rule and view the parameters of the plagiarism check rule. The task can continue to perform plagiarism check with the parameters of the plagiarism check rule, or the parameters of the plagiarism check rule can be adjusted. Conversely, if the query result is "no", it will enter the plagiarism check rule settings and set the parameters of the plagiarism check rule. An asynchronous job is started based on the parameters of the deduplication record rules. After the deduplication algorithm in the asynchronous job is executed to obtain the similarity score of each candidate record in the data, a classification label is generated according to the preset "identification result level" configuration in the asynchronous job. The classification label is then referenced to the many-to-many intermediate object of the asynchronous job through the comparison results in the "identification result level". Each result is associated with the "identification result level" and the "display" setting is initialized to show the deduplication result.

[0022] More specifically, the deduplication record rules in this embodiment include the following steps: Step 1: Identify the object to be checked, its activation status, and its description, and set the data filtering range conditions. The data filtering range conditions include the time range field, the time unit (day / month / year), and the numerical value. Step 2: Based on the data filtering range conditions in Step 1, determine whether the initiated task is allowed to be initiated. If the determination is "no", then reselect the data filtering range conditions. If the determination is "yes", then save it to the database table plat_config_similarity_check and its associated sub-tables to form structured configuration data.

[0023] Meanwhile, when viewing the record rules, multiple "matching settings" groups are designed. Each group contains a set of fields for semantic comparison. The field set includes a list of fields to display, a configuration for the recognition result level, and a maximum number of results. The configuration for the recognition result level includes name, minimum similarity threshold, and whether to display. The parameters for the deduplication record rules include matching settings, data range, and comparison fields.

[0024] Meanwhile, after the deduplication record rules form structured configuration data, a corresponding self-referencing many-to-many intermediate object is automatically created for the task object to store the deduplication comparison results; The intermediate object contains the following fields: Target Instance: The original record from which the deduplication check was initiated; Compare Instance: Records identified as potentially duplicates; Similarity Score: A floating-point number between 0 and 1; Result Type: Preset option set, including "High Similarity", "Medium Similarity", and "Low Similarity"; The relevant plagiarism check history (Check History Reference); Is Displayed: Controls whether the result is visible in the relevant list on the front end.

[0025] Therefore, the above describes the deduplication rule configuration stage in the following steps of a computer program. The deduplication rule configuration stage fully records the task parameters, execution time, and comparison results for each deduplication check, which facilitates auditing and review, thereby achieving task traceability.

[0026] Simultaneously, after the processor executes the computer program, enters the plagiarism check record rule settings, and configures the plagiarism check record rule parameters, a plagiarism check history record is created and stored in the plagiarism check history object table. The metadata of the plagiarism check history object table includes: The following information is required: referenced plagiarism detection rules; matching settings used; ID of the instance being compared; range fields and values; JSON format of the matching fields; initiator; initiation time; and task status.

[0027] This invention supports flexible definition of deduplication rules by object, field, and matching strategy to meet diverse business needs, thereby achieving high configurability.

[0028] More specifically, in this embodiment, the asynchronous job execution deduplication algorithm includes: Step 1: Extract the dataset to be compared from the target objects according to the configured time range and filtering conditions; Step 2: Concatenate the current instance with the specified fields of each record in the candidate set to form a natural language text fragment; Step 3: Calculate semantic similarity using any of the following methods: AI large-scale model inference: The current instance content is used as the query sentence, candidate sets are used to build knowledge base documents in batches, and large-scale model APIs such as Qwen and ChatGLM are called to perform semantic retrieval and scoring; Vector embedding comparison: The text is converted into vectors using a locally deployed embedding model (such as bge-m3), and the cosine similarity ranking is completed through a vector database (such as Elasticsearch) or in-memory computation. Third-party interfaces: Call external similarity service interfaces such as Baidu AI Open Platform to compare each item one by one.

[0029] This invention utilizes large models or high-quality embedded models to achieve deep semantic comparison, significantly improving recall and precision, thereby achieving strong semantic understanding capabilities; it achieves bidirectional display of plagiarism detection results through self-referencing many-to-many intermediate objects, enhancing user experience and thus realizing bidirectional relationship visualization; it supports multiple similarity calculation methods, facilitating future upgrades or replacements of the underlying AI engine, thereby achieving pluggable algorithms.

[0030] Meanwhile, the display of the plagiarism detection results in this embodiment includes the following steps: Step 1: When the user clicks the "Check for Plagiarism" button again, the system returns a summary and detailed list of the results of the latest plagiarism check task. The plagiarism check results are displayed in the form through two special types of "parts". Step 2: Users can set the "identification conclusion" for each matching result in the plagiarism detection results pop-up window, and mark it as "confirmed duplicate" or "not duplicate"; Step 3: After receiving the identification results, the system updates the visibility status of the corresponding intermediate object record: If multiple results in a deduplication check point to the same pair of records, only the latest confirmed result will be visible, and the old results will be automatically hidden; records that have been confirmed as duplicates can be processed in the business process.

[0031] This embodiment supports multiple languages, list layout configuration, and access control, ensuring the system adapts to complex enterprise environments. The plagiarism detection results are displayed in the form using two special types of "parts," which include: The duplicate detection initiation record section displays all the duplicate detection history of this record as a "target instance" in the form of an inbound relationship list. Clicking on it will bring up the details. The duplicate identification results section displays all associations identified when this record is used as a "comparison instance" in the form of a many-to-many relationship list. Viewing is supported only; manual addition, deletion, and modification are prohibited. The logic for displaying / hiding historical results is automatically processed to avoid information redundancy and interference, thereby achieving intelligent state management.

[0032] See Figure 2 As shown, the deviation AI text recognition method based on deep semantic comparison in this embodiment includes the following steps: Step 1: Collect and query whether there are recent plagiarism check rules. If the query result is "yes", then enter the plagiarism check rule and view the parameters of the plagiarism check rule. Continue to perform the task plagiarism check with the parameters of the plagiarism check rule, or adjust the parameters of the plagiarism check rule. Conversely, if the query result is "no", then enter the plagiarism check rule settings and set the parameters of the plagiarism check rule. Step 2: Start the asynchronous job based on the deduplication record rule parameters in the deduplication record rules. After the deduplication algorithm in the asynchronous job obtains the similarity score of each candidate record in the data, generate classification labels according to the preset "identification result level" configuration in the asynchronous job. Then, reference the classification labels to the many-to-many intermediate object of the asynchronous job through the comparison results in the "identification result level". Associate each result with the "identification result level" and initialize the "display" setting to show the deduplication result display.

[0033] This embodiment is configurable, intelligent, and platform-based, featuring strong semantic understanding capabilities, high configurability, bidirectional relationship visualization, task traceability, pluggable algorithms, and intelligent state management. It addresses the problems of weak semantic understanding, difficult business integration, and untraceable results in traditional plagiarism detection technologies. By introducing a Large Language Model (LLM) or a dedicated embedding model for semantic similarity analysis, combined with an object model-driven design approach, it achieves universal plagiarism detection capabilities for any business object.

[0034] The above are merely embodiments of this solution. Commonly known technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solution of this solution. These modifications and improvements should also be considered within the scope of protection of this solution, and will not affect the effectiveness of the implementation of this solution or the practicality of the patent. The scope of protection claimed in this application shall be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A biased AI text recognition system based on deep semantic comparison, characterized in that, Includes a memory, a processor, and a computer program stored in the memory and executed by the processor, wherein the processor, when executing the computer program, performs the following steps: If the query returns "yes" to indicate whether there are any recent plagiarism check rules, the user will be taken into the plagiarism check rule settings and will be able to view the parameters of the plagiarism check rule. The user can then continue the task plagiarism check using the parameters of the plagiarism check rule or adjust the parameters of the plagiarism check rule. Conversely, if the query returns "no", the user will be taken into the plagiarism check rule settings and will be able to set the parameters of the plagiarism check rule. An asynchronous job is started based on the parameters of the deduplication record rules. After the deduplication algorithm in the asynchronous job is executed to obtain the similarity score of each candidate record in the data, a classification label is generated according to the preset "identification result level" configuration in the asynchronous job. The classification tags are then referenced to the many-to-many intermediate object of the asynchronous job through the comparison results in "Identification Result Level". Each result is associated with "Identification Result Level" and "Display" is initialized to show the plagiarism detection results.

2. The deviation AI text recognition system based on deep semantic comparison according to claim 1, characterized in that: The rules for recording plagiarism detection include the following steps: Step 1: Identify the object to be checked, its activation status, and its description, and set the data filtering range conditions, which include the time range field, time unit, and numerical value. Step 2: Based on the data filtering range conditions in Step 1, determine whether the initiated task is allowed to be initiated. If the determination is "no", then reselect the data filtering range conditions. If the determination is "yes", then save it to the database table and its related sub-tables to form structured configuration data.

3. The deviation AI text recognition system based on deep semantic comparison according to claim 2, characterized in that: When the viewing record rule is designed with multiple "matching settings" groups, each group contains a set of fields for semantic comparison. The field set includes a list of display fields, configuration of recognition result level, and upper limit of the number of results.

4. The deviation AI text recognition system based on deep semantic comparison according to claim 3, characterized in that: The identification result level configuration includes name, minimum similarity threshold, and whether to display; the deduplication record rule parameters include matching settings, data range, and comparison fields.

5. The deviation AI text recognition system based on deep semantic comparison according to claim 2, characterized in that: After the deduplication record rules form structured configuration data, a corresponding self-referencing many-to-many intermediate object is automatically created for the task object to store the deduplication comparison results; The intermediate object contains the following fields: Target instance: The original record from which the plagiarism check was initiated; Comparison example: Records identified as potentially duplicates; Similarity score: a floating-point number between 0 and 1; Recognition result types: Preset option set, including "high similarity", "medium similarity", and "low similarity"; This is a record of plagiarism checks. Display Flag: Controls whether the result is visible in the relevant list on the front end.

6. The deviation AI text recognition system based on deep semantic comparison according to claim 1, characterized in that: After entering the plagiarism check record rule settings and configuring the plagiarism check record rule parameters, a plagiarism check history record is created and stored in the plagiarism check history object table. The metadata of the plagiarism check history object table includes: The following information is required: referenced plagiarism detection rules; matching settings used; ID of the instance being compared; range fields and values; JSON format of the matching fields; initiator; initiation time; and task status.

7. The deviation AI text recognition system based on deep semantic comparison according to claim 1, characterized in that: The asynchronous job, which executes a deduplication algorithm, specifically includes: Step 1: Extract the dataset to be compared from the target objects according to the configured time range and filtering conditions; Step 2: Concatenate the current instance with the specified fields of each record in the candidate set to form a natural language text fragment; Step 3: Calculate semantic similarity using any of the following methods: AI large-scale model reasoning: The current instance content is used as the query sentence, the candidate set is used to build knowledge base documents in batches, and the large-scale model API is called to perform semantic retrieval and scoring; Vector embedding comparison: The text is converted into vectors using a locally deployed embedding model, and the cosine similarity ranking is completed through a vector database or in-memory computation; Third-party interfaces: Call external similarity service interfaces such as Baidu AI Open Platform to compare each item one by one.

8. The deviation AI text recognition system based on deep semantic comparison according to claim 1, characterized in that: The display of plagiarism detection results includes the following steps: Step 1: When the user clicks the "Check for Plagiarism" button again, the system returns a summary and detailed list of the results of the latest plagiarism check task. The plagiarism check results are displayed in the form through two special types of "parts". Step 2: Users can set the "identification conclusion" for each matching result in the plagiarism detection results pop-up window, and mark it as "confirmed duplicate" or "not duplicate"; Step 3: After receiving the identification results, the system updates the visibility status of the corresponding intermediate object record: If multiple results in a deduplication check point to the same pair of records, only the latest confirmed result will be visible, and the old results will be automatically hidden; records that have been confirmed as duplicates can be processed in the business process.

9. The deviation AI text recognition system based on deep semantic comparison according to claim 8, characterized in that: The plagiarism detection results are displayed in the form using two special types of "parts," which include: The duplicate detection initiation record section displays all the duplicate detection history of this record as a "target instance" in the form of an inbound relationship list. Clicking on it will bring up the details. The duplicate identification results section displays all associations identified for this record as a "comparison instance" in the form of a many-to-many relationship list. It only supports viewing and does not allow manual addition, deletion or modification.

10. A biased AI text recognition method based on deep semantic comparison, applied to the biased AI text recognition system based on deep semantic comparison as described in any one of claims 1-9, characterized in that: Includes the following steps: Step 1: Collect and query whether there are recent plagiarism check rules. If the query result is "yes", then enter the plagiarism check rule and view the parameters of the plagiarism check rule. Continue to perform the task plagiarism check with the parameters of the plagiarism check rule, or adjust the parameters of the plagiarism check rule. Conversely, if the query result is "no", then enter the plagiarism check rule settings and set the parameters of the plagiarism check rule. Step 2: Start an asynchronous job based on the deduplication record rule parameters within the deduplication record rules. After executing the deduplication algorithm in the asynchronous job to obtain the similarity score of each candidate record in the data, generate classification labels according to the preset "identification result level" configuration within the asynchronous job. The classification tags are then referenced to the many-to-many intermediate object of the asynchronous job through the comparison results in "Identification Result Level". Each result is associated with "Identification Result Level" and "Display" is initialized to show the plagiarism detection results.

Images