Semantic slot iteration method and device, computer equipment and readable storage medium

Abstract

The invention provides a semantic slot iteration method and device, computer equipment and a readable storage medium. The method is applied to the question-answering system, the question-answering system is provided with a synonym table, the synonym table comprises a plurality of semantic slot records, each semantic slot record comprises a semantic slot and synonyms of a plurality of semantic slots, and the question-answering system is used for extracting semantic slot original information from a user question. When the semantic slot original information is matched with a first synonym in the first semantic slot record, determining a calling parameter by utilizing a first semantic slot in the first semantic slot record so as to call an API (Application Program Interface) to generate a question and answer result; the semantic slot iteration method comprises the steps of obtaining user feedback for a question and answer result; when the user feedback is negative feedback, adding the semantic slot original information as candidate words into a synonym candidate table; when the synonym table is updated, determining a second semantic slot record to which the candidate word belongs; and adding the candidate words into the second semantic slot record. According to the invention, the iteration efficiency can be improved, and the hit rate of answers can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a semantic slot iteration method, device, computer equipment and readable storage medium. Background Art

[0002] A question-answering system is a human-computer interaction system based on natural language processing technology and knowledge acquisition mechanism. It can receive natural language questions raised by users, and generate and output structured or natural language answers through intent recognition, semantic slot filling and external data interface (API) calls.

[0003] Among them, intent recognition, also known as intent detection, is a task in natural language processing. The system needs to understand the intention or purpose behind the text entered by the user. For example, the user may want to book a restaurant, check the weather, or get an answer to a question. The accuracy of intent recognition is crucial to providing relevant and useful answers. Semantic slot filling refers to identifying and extracting key information from user input, such as time, place, name, etc. This information is usually used to fill predefined "slots" so that the system can understand the user's specific needs and respond accordingly. After the semantic slot is filled, the call parameters are determined according to the semantic slot-parameter mapping table, and the API is further called by calling the parameters to generate answers to the user's questions.

[0004] Therefore, how to better improve the matching rate of semantic slot information in question-answering systems has become a technical problem that needs to be urgently solved in this field. Summary of the Invention

[0005] The purpose of the present invention is to provide a semantic slot iteration method, apparatus, computer equipment and readable storage medium for solving the above-mentioned technical problems in the prior art.

[0006] On the one hand, to achieve the above objectives, the present invention provides a semantic slot iteration method.

[0007] This semantic slot iteration method is applied to a question-and-answer system. The question-and-answer system is provided with a synonym table, which includes multiple semantic slot records. Each semantic slot record includes a semantic slot and several synonyms of the semantic slot. The question-and-answer system is used to extract the original semantic slot information from the user question. When the original semantic slot information matches the first synonym in the first semantic slot record, the system uses the first semantic slot in the first semantic slot record to determine the call parameters to call the API to generate the question-and-answer result. The semantic slot iteration method includes: obtaining user feedback on the question-and-answer result; when the user feedback is negative feedback, adding the original semantic slot information as a candidate word to the synonym candidate table; when the synonym table is updated, determining the second semantic slot record to which the candidate word belongs; and adding the candidate word to the second semantic slot record.

[0008] Furthermore, the semantic slot iteration method further includes: when the user feedback is positive feedback and the original information of the semantic slot is different from the first synonym, adding the original information of the semantic slot to the first semantic slot record.

[0009] Furthermore, when the synonym table is updated, the step of determining the second semantic slot record to which the candidate word belongs includes: monitoring a positive feedback update event in which user feedback is positive feedback; and when a positive feedback update event is monitored, determining the second semantic slot record to which the candidate word belongs.

[0010] Furthermore, the step of determining the second semantic slot record to which the candidate word belongs includes: using each word in the synonym table as a classification center point, classifying the words in the synonym candidate table, and determining the classification center point corresponding to the candidate word; and determining the semantic slot record to which the classification center point corresponding to the candidate word belongs as the second semantic slot record.

[0011] Furthermore, the synonym candidate table includes multiple candidate word records, and the candidate word records include synonyms and several semantic slot original information matching the synonyms; the step of adding the semantic slot original information as a candidate word to the synonym candidate table includes: searching for a first synonym in the synonym candidate table, and when the first synonym is found, adding the semantic slot original information to the candidate word record to which the first synonym belongs, and when the first synonym cannot be found, creating a new candidate word record including the first synonym and the semantic slot original information; the step of determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record includes: marking the correspondence between the candidate word record where the candidate word is located and the semantic slot record to which the classification center point corresponding to the candidate word belongs; when confirmation information of the correspondence is received, determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record.

[0012] Furthermore, the question-answering system adopts the following steps to determine synonyms that match the original information of the semantic slot: the original information of the semantic slot is subjected to a first matching process with the synonym table, wherein the first matching process is used to determine a first word in the synonym table that meets a first matching requirement with the original information of the semantic slot; when the first word is included in the synonym table, the first word is determined to be a synonym that matches the original information of the semantic slot; when the first word is not included in the synonym table, the original information of the semantic slot is subjected to a second matching process with the synonym table, wherein the second matching process is used to determine a second word in the synonym table that meets a second matching requirement with the original information of the semantic slot, wherein the first matching requirement is higher than the second matching requirement; and the second word is determined to be a synonym that matches the original information of the semantic slot.

[0013] Furthermore, the step of performing a second matching process on the original information of the semantic slot and the synonym table includes: converting the original information of the semantic slot and the synonym table into text vectors through word2vec technology according to the trained Bert model; calculating the cosine similarity between the text vector of the original information of the semantic slot and the text vector of the word in the synonym table; and determining the second word according to the cosine similarity.

[0014] On the other hand, to achieve the above-mentioned purpose, the present invention provides a semantic slot iteration device.

[0015] The semantic slot iteration device is applied to a question-answering system. The question-answering system is provided with a synonym table, which includes multiple semantic slot records. Each semantic slot record includes a semantic slot and several synonyms of the semantic slot. The question-answering system is used to extract original semantic slot information from user questions, and when the original semantic slot information matches the first synonym in the first semantic slot record, the first semantic slot in the first semantic slot record is used to determine the call parameters to call the API to generate a question-answering result; the semantic slot iteration device includes: an acquisition module, which is used to obtain user feedback on the question-answering result; a first processing module, which is used to add the original semantic slot information as a candidate word to the synonym candidate table when the user feedback is negative feedback; a second processing module, which is used to determine the second semantic slot record to which the candidate word belongs when the synonym table is updated; and a third processing module, which is used to add the candidate word to the second semantic slot record.

[0016] On the other hand, to achieve the above objectives, the present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0017] On the other hand, to achieve the above object, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0018] The semantic slot iteration method, device, computer equipment and readable storage medium provided by the present invention are used for semantic slot normalization identification by setting a synonym table, wherein a plurality of semantic slot records are stored in the synonym table, each record including a standard semantic slot and its corresponding multiple synonyms. Based on the synonym table, the system maps the user's natural language input into a structured semantic slot name, which is used to construct call parameters, initiate API requests, and generate question-and-answer results. After obtaining the question-and-answer results, a negative feedback-driven semantic slot self-evolution mechanism is introduced. When the user provides negative feedback on a certain question-and-answer result, the system automatically identifies the original expression of the semantic slot that was not correctly matched in this round, records it as a candidate word and adds it to the synonym candidate table. When the synonym table is updated, the semantic slot record to which the candidate word should belong is re-determined based on the updated synonym table, and the candidate word is formally added to the corresponding semantic slot record, realizing a closed-loop iteration from feedback, re-determination to update, realizing the correction of matching errors of the candidate word in the synonym table, and realizing the effective and dynamic expansion of the synonym table. Through the present invention, with the guidance of the synonym table, the parsing ability of non-standard expressions such as colloquialisms, dialects, and aliases is improved, and unified processing of multiple expressions of the same semantic slot is supported, thereby improving the generalization ability of the dialogue system. Low matching situations in actual user use are fed back to the system through a negative feedback mechanism, forming a data-driven self-optimization chain and building a user feedback closed loop. By using the candidate word selection table and the judgment mechanism based on the updated synonym table, incorrect matches are dynamically corrected and slot synonyms are expanded, realizing automatic iterative upgrades of the semantic system, reducing manual maintenance costs, and enhancing the system's adaptability to real user corpus. Users can express themselves more freely in natural language without having to deliberately cater to the system vocabulary. During the iteration process, the system continuously learns new expressions, so that user satisfaction and accuracy are simultaneously improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings: Figure 1 A schematic diagram of the workflow of the question-answering system provided in an embodiment of the present invention; Figure 2 A flowchart for determining parameters in a semantic slot iteration method provided by an embodiment of the present invention; Figure 3 A flowchart of a semantic slot iteration method provided by an embodiment of the present invention; Figure 4 A flowchart for determining parameters in another semantic slot iteration method provided by an embodiment of the present invention; Figure 5A flowchart of another semantic slot iteration method provided by an embodiment of the present invention; Figure 6 A schematic diagram of another question-answering system workflow provided by an embodiment of the present invention; Figure 7 A block diagram of a semantic slot iteration device provided by an embodiment of the present invention; Figure 8 A hardware structure diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] Figure 1 A schematic diagram of the workflow of the question-answering system provided in an embodiment of the present invention is shown in FIG. Figure 1 As shown in the figure, after receiving a user question, the question-answering system extracts the semantic slot information (i.e., the original semantic slot information) and first matches it against the semantic slot-parameter mapping table. If the match is successful, the extracted semantic slot information is converted into parameters matching the API interface. If the match fails in the semantic slot-parameter mapping table, further matching is performed using the synonym table. Similarly, if the match is successful, the extracted semantic slot information is converted into parameters matching the API interface. If the original semantic slot information still does not match successfully in the synonym table, the API call will fail due to incomplete parameter extraction.

[0022] To mitigate the problem of parameter mapping failures and subsequent scenario question-answering failures caused by a lack of synonym tables, the only way to improve matching rates is to manually update and maintain the semantic slot synonym table. However, this method of manually updating the semantic slot synonym table is inefficient, labor-intensive, unstable, and unsustainable.

[0023] In order to further solve the above technical problems, the present invention provides a semantic slot iteration method, apparatus, computer equipment and readable storage medium. The specific embodiments of the semantic slot iteration method, apparatus, computer equipment and readable storage medium provided by the present invention will be described in detail below.

[0024] Example 1 An embodiment of the present invention provides a semantic slot iteration method, through which the efficiency of semantic slot iteration can be improved, the matching rate of semantic slot information in a question-and-answer system can be better improved, and the hit rate of answers can be improved. Specifically, the semantic slot iteration method is applied to a question-and-answer system, and the question-and-answer system is provided with a synonym table, which includes multiple semantic slot records, each semantic slot record includes a semantic slot and several synonyms of the semantic slot. The question-and-answer system is used to extract the original semantic slot information from the user's question, match the original semantic slot information with the synonym table, and when the original semantic slot information matches a synonym (defined as the first synonym) in a semantic slot record (defined as the first semantic slot record), the semantic slot in the first semantic slot record (defined as the first semantic slot) is used to determine the call parameters to call the API to generate a question-and-answer result. Optionally, a fuzzy matching function can be used when matching the original semantic slot information with the synonym table. Specifically, the extracted semantic slot original information is vectorized and the similarity is calculated with all the words in the synonym table. After obtaining the word with the highest similarity, the threshold is determined. If it is higher than the threshold, it is considered a hit, and the word is mapped to the parameter value in the parameter mapping table through the synonym table to participate in subsequent API calls. Figure 2 A flowchart of calling parameters in a semantic slot iteration method provided by an embodiment of the present invention is as follows: Figure 2 As shown, the user question extracts semantic slot information words B2 and C8. B2 finds a confirmed correspondence in the semantic slot synonym table and can be located with Key=B. However, C8 is not in the synonym table. Fuzzy matching is then used to find the closest C2 word, which is then located with Key=C. Key=B and Key=C are then used to find param_value_b and param_value_c through the parameter mapping table. Finally, the two parameter values ​​are used to construct API parameters and execute subsequent rules. Different solutions can be used in the fuzzy matching stage, such as vector similarity calculation, edit distance, and neural network training classifiers. Compared with full matching, fuzzy matching increases generalization and parameter hit rate. On this basis, Figure 3 This is a flowchart of the semantic slot iteration method provided in the first embodiment of the present invention, such as Figure 3 As shown, the semantic slot iteration method provided by this embodiment includes the following steps S101 to S104.

[0025] Step S101: Obtain user feedback on the question and answer results.

[0026] Specifically, the question-answering system calls the API, uses the return value of the API to generate question-answering results, and presents the question-answering results for the user's questions to the user.

[0027] Optionally, in one scenario, the Q&A system provides a channel for users to provide feedback on the Q&A results and obtain user feedback on the results. This is also known as explicit user feedback, where users explicitly express satisfaction or dissatisfaction, typically through actions within the interactive interface. For example, after returning the Q&A results, a "Helpful?" button is displayed, and the user clicks "Satisfied" or "Dissatisfied" under the Q&A results, whereupon the Q&A system obtains user feedback. Another example is a rating mechanism where users rate the Q&A results from 1 to 5 stars, with a set threshold (e.g., 3 stars or less is considered negative feedback). Alternatively, a text comment box can be provided where users can actively enter feedback text such as "This is wrong" or "That's not what I meant."

[0028] Alternatively, in another scenario, the Q&A system indirectly determines whether the Q&A result is acceptable based on the user's behavior patterns. For example, this can be determined by whether the user clicks on subsequent recommendations. If the user immediately clicks on other candidate Q&A candidates or related resources, it indicates that the current result may be unsatisfactory. Another example is whether the user continues to ask questions or corrects the question. If the user immediately asks the same question again or replies with the same intention, it is likely that the user is dissatisfied with the current answer. Another example is that the user's reading time for the Q&A result is extremely short, including quickly closing the result panel, directly jumping to a new search, or canceling the operation, indicating that the user has no interest in the result and may be a mismatch.

[0029] Regardless of the method used to obtain user feedback, the user feedback in the present invention includes positive feedback and negative feedback, wherein positive feedback indicates that the current question and answer result hits the user's question and meets the user's needs, and negative feedback indicates that the user is dissatisfied with the current question and answer result.

[0030] Step S102: When the user feedback is negative feedback, the original information of the semantic slot is added as a candidate word to the synonym selection table.

[0031] Optionally, negative feedback in the present invention refers to the user's lack of clear affirmative expression of the question and answer result, including clear negative expression and the situation where it is impossible to clearly infer whether the user is sure of the question and answer result.

[0032] When the system generates question-answer results that do not meet user expectations based on incorrect synonyms or slot matching, users provide negative feedback, indicating that the system has misunderstood the semantic slots in the input.

[0033] The present invention sets up a synonym candidate table. When the user feedback is negative feedback, the original information of the semantic slot triggered in the current question (that is, the expression actually input by the user) is traced back and added as a candidate word to an independent synonym candidate table for subsequent review and expansion, so as to quickly archive the synonyms missed in the user's actual expression, provide a structured input source for the dynamic iteration of the synonym system, and realize a semantic adaptation mechanism based on a feedback closed loop.

[0034] Step S103: When the synonym table is updated, the second semantic slot record to which the candidate word belongs is determined.

[0035] After the original information of the semantic slot is added to the synonym candidate table as a candidate word, if the state of the synonym table is updated relative to the state when the original information of the semantic slot is matched with the synonym table, for example, the synonym table is incrementally updated, then in the updated synonym table, the semantic slot record to which the candidate word belongs is found again, that is, the semantic slot classification error of the original information of the semantic slot is corrected based on the updated synonym table.

[0036] When the synonym candidate list includes multiple candidate words, each candidate word is processed in turn.

[0037] Step S104: adding the candidate word to the second semantic slot record.

[0038] When the new semantic slot record to which the candidate word belongs, i.e., the second semantic slot record, is determined, the candidate word is added to the second semantic slot record, further updating the synonym table and completing the iteration of the semantic slot. It should be noted that in step S103, "updating the synonym table" refers to updates other than adding the candidate word to the semantic slot record, to avoid the step logic from entering an improper loop.

[0039] In this embodiment, a semantic slot iteration method suitable for a question-answering system is provided. A synonym table is set for semantic slot normalization identification. A number of semantic slot records are stored in the synonym table. Each record includes a standard semantic slot and its corresponding multiple synonyms. Based on the synonym table, the system maps the user's natural language input into a structured semantic slot name, which is used to construct call parameters, initiate API requests, and generate question-answering results. After obtaining the question-answering results, a negative feedback-driven semantic slot self-evolution mechanism is introduced. When the user provides negative feedback on a certain question-answering result, the system automatically identifies the original expression of the semantic slot that was not correctly matched in this round, records it as a candidate word, and adds it to the synonym candidate table. When the synonym table is updated, the semantic slot record to which the candidate word should belong is re-determined based on the updated synonym table, and the candidate word is formally added to the corresponding semantic slot record, realizing a closed-loop iteration from feedback, re-determination to update, realizing the correction of matching errors of candidate words in the synonym table, and realizing the effective and dynamic expansion of the synonym table. The semantic slot iteration method provided in this embodiment, guided by a synonym table, improves the parsing capabilities of non-standard expressions (such as colloquialisms, dialects, and aliases), supports unified processing of multiple expressions for the same semantic slot, and enhances the generalization capabilities of the dialogue system. Low matches in actual user usage are fed back to the system through a negative feedback mechanism, forming a data-driven self-optimization chain and a user feedback closed loop. Using a candidate word list and a judgment mechanism based on an updated synonym table, incorrect matches are dynamically corrected and slot synonyms are expanded. This enables automatic iterative upgrades of the semantic system, reduces manual maintenance costs, and enhances the system's adaptability to real user corpus. Users can express themselves more freely in natural language without having to deliberately conform to the system vocabulary. During the iteration process, the system continuously learns new expressions, resulting in simultaneous improvements in user satisfaction and accuracy.

[0040] Optionally, in one embodiment, the semantic slot iteration method further includes: when the user feedback is positive feedback and the original semantic slot information is different from the first synonym, adding the original semantic slot information to the first semantic slot record.

[0041] In this embodiment, in order to further improve the adaptability of the question-answering system to natural language expressions, the semantic slot iteration method not only responds to negative feedback for correction learning, but also supports vocabulary expansion based on positive feedback. Specifically, when the user feedback is positive feedback, for example, the user clicks the "Satisfied" button, or the user actively replies with confirmation statements such as "Yes" or "Exactly what I want", the system traces back the original information of the semantic slot used to construct the API parameters in this round of question-answering, that is, the expression actually entered by the user, and compares it with the first synonym used for its mapping. If the two are not equal, it means that although the original expression was not included as a synonym, the match is still successful and the user confirms that the result is correct, then it can be regarded as a valid new synonym expression, and the original information of the semantic slot is appended to the synonym list recorded in the first semantic slot, so that when other users use similar expressions in the future, the system can directly hit it. For example, the current semantic slot is: date, the first synonym is: "today", the user input sentence is: "How is the weather in Beijing today", the system parses it as: date = today → match successful → generate a reply, the user clicks "satisfied", the system detects that "today" ≠ "today", and adds "today" to the synonym list of the date semantic slot.

[0042] This embodiment provides a semantic slot iteration method, realizes an automatic expansion mechanism of the synonym list based on positive feedback, captures expressions in the user's real natural language that have not yet been included in the system but have been successfully parsed, enriches the semantic slot vocabulary, does not need to rely on manual review, uses positive feedback as a semantic credibility signal, directly promotes the growth of the vocabulary, realizes low-cost self-learning based on user confirmation, has the ability to quickly respond and converge to regional spoken language, personalized expressions, and popular sayings, improves the system's language adaptability in unsupervised scenarios, improves the hit rate of the next recognition, avoids falling into low-confidence fuzzy channels again, and improves performance and user experience.

[0043] Furthermore, the semantic slot iteration method provided in this embodiment not only supports a synonym matching correction and completion mechanism based on negative user feedback, but also introduces a synonym expansion strategy based on positive feedback, establishing a bidirectional semantic learning mechanism that covers both error correction and correct reinforcement. This iterative mechanism, combining positive and negative feedback, establishes a self-learning, self-correcting, and self-expanding semantic slot management loop for the question-and-answer system. This significantly improves the system's ability to understand non-standard, personalized, and diverse expressions, reduces the cost of manual intervention, and enhances the system's intelligence and user experience.

[0044] Optionally, in one embodiment, when the synonym table is updated, the step of determining the second semantic slot record to which the candidate word belongs includes: monitoring a positive feedback update event in which user feedback is positive feedback; and when a positive feedback update event is monitored, determining the second semantic slot record to which the candidate word belongs.

[0045] In this embodiment, not only is candidate word attribution judgment performed when the synonym table is updated, but a real-time monitoring mechanism for positive feedback update events is further introduced. When the user expresses approval of the current question and answer results through positive feedback and the synonym table is updated based on the positive feedback, the original information of the current semantic slot can be used as a candidate word and an attempt can be made to judge the slot attribution, thereby improving the adaptive and automatic iteration capabilities of the semantic slot iteration method.

[0046] Specifically, after the question-and-answer process is completed, the system continues to monitor the user's interactive behavior on the current question-and-answer results. If a positive feedback event is detected and the synonym table is updated based on the positive feedback, the step of automatically determining the slot attribution of the candidate words is triggered.

[0047] This embodiment provides a semantic slot iteration method. By setting up a semantic slot iteration mechanism driven by user feedback and combining positive and negative feedback signals, a corresponding mechanism is set for the generation of candidate words, attribution judgment, and the triggering time of attribution judgment, thereby improving the effectiveness and rationality of the evolution of the semantic synonym system. When the user provides negative feedback, the system believes that there is a semantic understanding error in the current question and answer result. Therefore, the original information of the current semantic slot is directly added to the synonym candidate list as a candidate word, but the attribution judgment is not immediately performed. This avoids semantic drift caused by expressions that the user does not agree with being incorrectly attached to the semantic slot, ensuring that the processing of the candidate words remains cautious.

[0048] Conversely, when users provide positive feedback, if the original information in the current semantic slot is not included in the hit synonym list, the system not only directly adds the original information to the synonym list to enhance the expression, but also simultaneously triggers the candidate word attribution judgment logic. In this case, the candidate word attribution judgment is based on the interaction results that have been approved by the user, with a higher confidence level. The system can immediately infer the target semantic slot record to which it should belong and update its slot attachment relationship.

[0049] Through the above-mentioned mechanism design, this embodiment constructs a dual-path judgment system with positive feedback - judgment and inclusion, and negative feedback - observation first and then attachment, to achieve dynamic control of the processing rhythm of candidate words under different semantic confidence backgrounds, and use positive feedback to strengthen expression and negative feedback to carefully fill in the gaps, realizing two-way system optimization from recognition errors to recognition blind spots, and ultimately improving the question-answering system's comprehensive understanding of user expressions.

[0050] Optionally, in one embodiment, the step of determining the second semantic slot record to which the candidate word belongs includes: using each word in the synonym table as a classification center point, classifying the words in the synonym candidate table, and determining the classification center point corresponding to the candidate word; and determining the semantic slot record to which the classification center point corresponding to the candidate word belongs as the second semantic slot record.

[0051] In this embodiment, in order to realize the automatic attribution of candidate words to semantic slot records, a clustering attribution method based on synonym classification center points is adopted to realize intelligent classification of candidate words.

[0052] Specifically, all semantic slot records in the current synonym table are traversed, and each term under it is extracted. Each term is treated as a classification center point, and a set of centers distributed across the semantic expression space is constructed. Each center point naturally belongs to the semantic slot record to which it belongs. Each candidate word in the synonym candidate table is converted into a vector representation. This can be constructed using static word embedding models (such as Word2Vec and GloVe) or by extracting the semantic embedding of the candidate word using a contextual semantic model (such as BERT and SimCSE). For each candidate word to be assigned, the similarity (such as cosine similarity) between its vector and the vectors of all classification centers is calculated, and the center point with the highest similarity is selected. This center point is considered the candidate word's nearest semantic neighbor, representing the closest semantic slot assignment direction. The semantic slot record to which this classification center point belongs is found and identified as the target semantic slot record to which the current candidate word should be attached. This is recorded as the second semantic slot record. Subsequently, the candidate word can be added to this semantic slot record as a new synonym, achieving self-evolutionary updates of the semantic slot vocabulary.

[0053] This embodiment provides a semantic slot iteration method that constructs a semantic vector space of synonym terms and uses each term as a clustering center to complete the classification and attribution of synonym candidate terms. The attribution of candidate terms does not rely on manual slot labels. Instead, it automatically infers their attribution through the nearest neighbor principle in the semantic space, improving attribution accuracy and avoiding reliance on direct matching of slot names. This method is suitable for determining polysemous expressions. The attribution of each candidate term can be traced back to the nearest neighbor center point, and supports similarity thresholds to control the error attribution rate, improving the robustness and explainability of attribution decisions. The system can complete attribution judgments under unsupervised or weakly supervised conditions, making it suitable for the continuous maintenance and evolution of large-scale semantic slot systems, reducing manual classification costs, and adapting to large-scale multi-slot systems.

[0054] Optionally, in one embodiment, the synonym candidate table includes multiple candidate word records, and the candidate word records include synonyms and several semantic slot original information that match the synonyms. The step of adding the semantic slot original information as a candidate word to the synonym candidate table includes: searching for a first synonym in the synonym candidate table, and when the first synonym is found, adding the semantic slot original information to the candidate word record to which the first synonym belongs, and when the first synonym is not found, creating a new candidate word record including the first synonym and the semantic slot original information. The step of determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record includes: marking the correspondence between the candidate word record where the candidate word is located and the semantic slot record to which the classification center point corresponding to the candidate word belongs; when confirmation information of the correspondence is received, determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record.

[0055] In this embodiment, synonyms and original information are managed in the synonym candidate table based on candidate word records. In combination with the correspondence between candidate words and classification center points, the semantic slot attribution judgment scheme is completed through a confirmation mechanism, thereby improving the efficiency of candidate word organization and management and the controllability of attribution judgment. Specifically, a synonym candidate table is maintained to store multiple candidate word records. The candidate word record includes the original semantic slot information and the synonyms matched by the original semantic slot information in the synonym table. Among them, a synonym may be matched by multiple different original semantic slot information. Therefore, a candidate word record corresponds to a synonym, including one or more original semantic slot information, forming an original expression list. When user feedback is negative feedback, the original semantic slot information is obtained, and the synonym matched by the original semantic slot information in the synonym table is defined as the first synonym. Then, the synonym candidate table is checked to see whether there is a candidate word record including the first synonym. If so, the original semantic slot information is directly added to the original expression list of the corresponding record. If not, a new candidate word record is created and the first synonym and the original semantic slot information are written together.

[0056] A semantic similarity calculation can be used to establish an attribution tendency relationship between candidate words and terms in the synonym table, i.e., classification center points. In this embodiment, a confirmation mechanism is introduced as a manual or weakly supervised confirmation method to avoid misattribution. Optionally, after the preliminary correspondence between the candidate word and the classification center point is determined, the attribution direction is temporarily stored in a pending confirmation state. When the backend management interface manually reviews or the semantic consistency score of the large model meets the requirements, the candidate word is bound as a synonym for the slot to which the center point belongs.

[0057] This embodiment is used to provide a semantic slot iteration method, and a high-confidence semantic slot attribution logic based on candidate word records, classification center point classification, and confirmation mechanism is constructed. Each synonym candidate can carry multiple real expression examples, which is convenient for attribution judgment, manual review, and model assistance, and enhances the context carrying capacity of candidate word records. The system adopts a mechanism of first establishing attribution tendency and then confirming the attachment to improve the accuracy of the vocabulary, prevent misclassification, realize separate decision-making of semantic slot attribution judgment, and maintain the openness and controllability of the system. It adapts to intelligent operation and controllable vocabulary strategies, supports manual and intelligent dual-channel confirmation mechanisms, and each word attribution is supported by semantic paths and user feedback, which enhances the traceability and optimizability of the system and constructs an explainability mechanism for self-learning word performance evolution paths.

[0058] Optionally, in one embodiment, the question-answering system uses the following steps to determine synonyms that match the original information of the semantic slot: performing a first matching process on the original information of the semantic slot and the synonym table, wherein the first matching process is used to determine a first word in the synonym table that meets a first matching requirement with the original information of the semantic slot; when the first word is included in the synonym table, determining the first word as a synonym that matches the original information of the semantic slot; when the first word is not included in the synonym table, performing a second matching process on the original information of the semantic slot and the synonym table, wherein the second matching process is used to determine a second word in the synonym table that meets a second matching requirement with the original information of the semantic slot, wherein the first matching requirement is higher than the second matching requirement; determining the second word as a synonym that matches the original information of the semantic slot.

[0059] In this embodiment, a two-layer matching mechanism is proposed to hierarchically search for synonyms that match the original information of the semantic slot from the synonym table, so as to improve the question-answering system's ability to understand the user's natural language expression, while avoiding slot recognition errors caused by mismatching.

[0060] Specifically, the first matching step is performed. This step searches for terms that are highly consistent with the original semantic slot information. Exact matching, such as direct index matching, can be used. If the first matching fails, the system enters the second matching stage, which matches candidates with similar but not identical expressions. The second matching stage uses fuzzy matching methods, such as edit distance and Jaro-Winkler distance. The first matching step requires exact or nearly exact content consistency, while the second matching step allows for a certain degree of semantic fuzziness, favoring error tolerance.

[0061] Figure 4 A flowchart of calling parameters in another semantic slot iteration method provided by an embodiment of the present invention is shown as follows: Figure 4As shown, a user asks, "How many moderately dangerous alerts have been handled today?" The parameter extraction model first extracts the semantic slot raw information terms "moderately dangerous" and "handled." The two information terms are then matched against a synonym table, and the key value is found based on the search and index. "Moderately dangerous" can be accurately matched against the synonym table, and its key value, "medium-risk," is found. However, "handled" does not exist in the synonym table. A second matching process is then performed. Using the trained BERT model, the word and synonym table are converted into text vectors using word2vec technology. Cosine similarity is used to find the closest synonym, "handled," and its corresponding key value, "handled." Finally, the parameter mapping table maps these values ​​to the API's required parameter values ​​{severity: 003, status: 005}. The API is then called to retrieve the results and construct and generate the answer output.

[0062] This embodiment provides a semantic slot iteration method, which combines a two-layer matching strategy, that is, precision combined with fuzziness, with a semantic slot self-learning mechanism based on user feedback, to build a semantic understanding system with high-confidence matching priority, fault-tolerant matching as a backup, and feedback-driven evolution. By setting a first matching process with a higher priority, it is possible to quickly locate standard expressions, achieve direct placement of semantic slots, ensure recognition efficiency and accuracy, give priority to the use of high-confidence corpus, and avoid over-reliance on fuzzy processes with high computational overhead; when the first matching process fails, it enters the second matching process under the fault-tolerant mechanism, intelligently covers colloquial, regionalized, and non-standard inputs, and reduces the problem of limited user expression. On this basis, the semantic slot self-learning mechanism combined with user feedback strengthens and corrects the fuzzy matching results. On the basis of ensuring a high confidence hit rate, the fault tolerance and learning ability of the question-answering system for non-standard expressions are enhanced, achieving a balance between recognition robustness, semantic coverage, and dynamic optimizability.

[0063] Optionally, in one embodiment, the step of performing a second matching process on the original information of the semantic slot and the synonym table includes: converting the original information of the semantic slot and the synonym table into text vectors through word2vec technology according to the trained Bert model; calculating the cosine similarity between the text vector of the original information of the semantic slot and the text vector of the word in the synonym table; and determining the second word based on the cosine similarity.

[0064] The semantic slot iteration method provided in this embodiment is used to perform a "second matching process" based on the fuzzy synonym recognition method using the Bert model and Word2Vec vector matching, thereby improving the semantic matching capability between the original semantic slot information and the expressions in the synonym table and effectively identifying deformed, colloquial, and diversified expressions.

[0065] Optionally, in one embodiment, Figure 5A flowchart of another semantic slot iteration method provided by an embodiment of the present invention, Figure 6 Another question-answering system workflow diagram provided by an embodiment of the present invention is as follows: Figure 5 and Figure 6 As shown, the question-answering system first receives a user question, extracts the original semantic slot information from the user question, and performs a first matching process on the original semantic slot information and the synonym table, wherein the first matching process is used to determine a first word in the synonym table that meets a first matching requirement with the original semantic slot information. When the first word is included in the synonym table, the first word is determined to be a synonym that matches the original semantic slot information, and a first synonym is obtained. When the first word is not included in the synonym table, the original semantic slot information is subjected to a second matching process with the synonym table, wherein the second matching process is used to determine a second word in the synonym table that meets a second matching requirement with the original semantic slot information, wherein the first matching requirement is higher than the second matching requirement. The second word is determined to be a synonym that matches the original semantic slot information, and a first synonym is obtained. The first semantic slot in the first semantic slot record where the first synonym is located is used to determine the call parameters; the determined call parameters are used to call the API to generate the question-answering result.

[0066] Obtain user feedback on the question-and-answer results. If the user feedback is negative, add the original semantic slot information as a candidate word to the synonym candidate table. If the user feedback is positive, add the original semantic slot information to the first semantic slot record. Monitor for positive feedback update events. If a positive feedback update event is detected, determine the second semantic slot record to which the candidate word belongs and add the candidate word to the second semantic slot record.

[0067] In this embodiment, when the original information of the semantic slot does not exist in the synonym table, a second matching process is adopted to find the parameter key with the highest similarity by means of vector matching. If the similarity is higher than the threshold, the result is regarded as a successful match. Secondly, the synonym table is updated and the synonym candidate table is collected by adaptive learning through analysis of user feedback and behavior. Finally, the words in the candidate table are synchronized to the synonym table through regular maintenance in the form of manual review, which can greatly improve the iteration efficiency and reduce manpower expenditure. In addition, because there is a mechanism for fuzzy matching parameters, the answer hit rate can be improved and the user experience can be enhanced.

[0068] By analyzing user feedback on question-and-answer results, the synonym table is updated and the synonym candidate table is collected in an adaptive learning manner. User feedback is divided into positive and negative. Positive feedback indicates that the user approves of the results hit by the question-and-answer model, which indirectly confirms that the intent recognition is accurate and the semantic slot approximate conversion mapping is correct. The model then automatically enters the user's original question into the question template library to expand the training set and increase intent recognition capabilities. At the same time, the extracted semantic slot information words are entered into the synonym table to expand the synonym table and increase the parameter conversion success rate. Negative feedback indicates that the user does not approve of the results returned by the question-and-answer model. The reason may be that a certain parameter conversion is inaccurate, causing the result to deviate from the expectation. In this case, the corresponding relationship between the extracted semantic slot information words and the calculated relationship in the question-and-answer stage is automatically stored in the candidate table for manual review by operation and maintenance personnel to update the new words in the candidate table to the official synonym table. At the same time, positive and negative feedback events are monitored. Once a new synonym enters the synonym table, the classifier is used to reclassify all the words in the candidate table to increase the classification accuracy and improve the review efficiency of operation and maintenance personnel.

[0069] refer to Figure 4 In the illustrated embodiment, if the user feedback is positive, the user's original question, "How many moderately dangerous alarms have been handled today?", is automatically entered into the question template dataset for better training of subsequent intent recognition. The extracted semantic slot information words "moderately dangerous" and "handled" are automatically entered into the synonym table. If "moderately dangerous" already exists, it is skipped, and "handled" is added to the list with the key value of "handled" to increase the parameter mapping hit rate in the future. If the user feedback is negative or there is no feedback, the extracted semantic slot information word "handled" is entered into the candidate table for regular review and addition of new words to the synonym table by operation and maintenance personnel. Since the question-and-answer service is running, the synonym table will automatically update new words at any time. If a word that better fits the candidate information appears, the corresponding relationship matching degree in the candidate table may decrease over time. Therefore, here, a KNN classification model is run to monitor the synonym entry event, classify all candidate words in the candidate table, and combine new similar word correspondences. In this way, the corresponding relationship matching degree in the candidate table can always be maintained at the highest state, thereby reducing manpower expenditure.

[0070] Example 2 Corresponding to the above-mentioned embodiment 1, embodiment 2 of the present invention provides a semantic slot iteration device. The corresponding technical feature details and corresponding technical effects can be referred to the above-mentioned embodiment 1, and will not be repeated in this embodiment. The semantic slot iteration device is applied to a question-answering system, and the question-answering system is provided with a synonym table, and the synonym table includes a plurality of semantic slot records, each of the semantic slot records includes a semantic slot and several synonyms of the semantic slot, and the question-answering system is used to extract the original information of the semantic slot from the user question, and when the original information of the semantic slot matches the first synonym in the first semantic slot record, the first semantic slot in the first semantic slot record is used to determine the call parameters to call the API to generate the question-answering result. Figure 7 This is a block diagram of the semantic slot iteration device provided in the second embodiment of the present invention, such as Figure 7 As shown, the device includes: an acquisition module 301 , a first processing module 302 , a second processing module 303 and a third processing module 304 .

[0071] Among them, the acquisition module 301 is used to obtain user feedback on the question and answer results; the first processing module 302 is used to add the original information of the semantic slot as a candidate word to the synonym candidate table when the user feedback is negative feedback; the second processing module 303 is used to determine the second semantic slot record to which the candidate word belongs when the synonym table is updated; and the third processing module 304 is used to add the candidate word to the second semantic slot record.

[0072] Optionally, in one embodiment, the semantic slot iteration device further includes: a fourth processing module, configured to add the semantic slot original information to the first semantic slot record when the user feedback is positive feedback and the semantic slot original information is different from the first synonym.

[0073] Optionally, in one embodiment, the second processing module includes: a monitoring unit for monitoring a positive feedback update event in which the user feedback is positive feedback; and a first processing unit for determining a second semantic slot record to which the candidate word belongs when the positive feedback update event is monitored.

[0074] Optionally, in one embodiment, the second processing module includes: a classification unit, used to classify the words in the synonym candidate table using each word in the synonym table as a classification center point, and determine the classification center point corresponding to the candidate word; and a determination unit, used to determine that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record.

[0075] Optionally, in one embodiment, the synonym candidate table includes multiple candidate word records, each of which includes the synonym and a plurality of semantic slot original information matching the synonym; the first processing module includes: a query unit for querying the first synonym in the synonym candidate table; a second processing unit for adding the semantic slot original information to the candidate word record to which the first synonym belongs when the first synonym is found; and a third processing unit for creating a new candidate word record including the first synonym and the semantic slot original information when the first synonym is not found. When the determination unit determines that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record, the specific steps performed include: marking the correspondence between the candidate word record to which the candidate word belongs and the semantic slot record to which the classification center point corresponding to the candidate word belongs; and upon receiving confirmation information of the correspondence, determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record.

[0076] Optionally, in one embodiment, the question-and-answer system uses the following steps to determine synonyms that match the original information of the semantic slot: performing a first matching process on the original information of the semantic slot and the synonym table, wherein the first matching process is used to determine a first word in the synonym table that meets a first matching requirement with the original information of the semantic slot; when the first word is included in the synonym table, determining that the first word is a synonym that matches the original information of the semantic slot; when the first word is not included in the synonym table, performing a second matching process on the original information of the semantic slot and the synonym table, wherein the second matching process is used to determine a second word in the synonym table that meets a second matching requirement with the original information of the semantic slot, wherein the first matching requirement is higher than the second matching requirement; determining that the second word is a synonym that matches the original information of the semantic slot.

[0077] Optionally, in one embodiment, the step of performing a second matching process on the original information of the semantic slot and the synonym table includes: converting the original information of the semantic slot and the synonym table into text vectors through word2vec technology according to a trained Bert model; calculating the cosine similarity between the text vector of the original information of the semantic slot and the text vector of the word in the synonym table; and determining the second word based on the cosine similarity.

[0078] Example 3 This embodiment also provides a computer device, such as a smart phone, tablet computer, laptop computer, desktop computer, rack server, blade server, tower server or cabinet server (including an independent server or a server cluster composed of multiple servers) that can execute programs. Figure 8 As shown, the computer device 01 of this embodiment includes at least but not limited to: a memory 012 and a processor 011 which can be interconnected via a system bus. Figure 8 It should be pointed out that Figure 8 The computer device 01 is shown only with components memory 012 and processor 011 , but it should be understood that implementation of all of the components shown is not a requirement and more or fewer components may alternatively be implemented.

[0079] In this embodiment, memory 012 (i.e., readable storage medium) includes flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, a magnetic disk, an optical disk, etc. In some embodiments, memory 012 may be an internal storage unit of computer device 01, such as the hard disk or internal memory of computer device 01. In other embodiments, memory 012 may also be an external storage device of computer device 01, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. equipped on computer device 01. Of course, memory 012 may also include both internal storage units and external storage devices of computer device 01. In this embodiment, memory 012 is generally used to store the operating system installed on computer device 01 and various reference software, such as the program code of the semantic slot iteration device in Example 2. In addition, the memory 012 can also be used to temporarily store various types of data that have been output or are to be output.

[0080] In some embodiments, processor 011 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. Processor 011 is typically used to control the overall operation of computer device 01. In this embodiment, processor 011 is used to execute program code stored in memory 012 or process data, such as the semantic slot iteration method.

[0081] Example 4 This embodiment also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a disk, an optical disk, a server, an app reference store, etc., on which a computer program is stored. When the program is executed by a processor, the corresponding function is implemented. The computer-readable storage medium of this embodiment is used to store a semantic slot iteration device, and when executed by a processor, the semantic slot iteration method of Example 1 is implemented.

[0082] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0083] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0084] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method.

[0085] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A semantic slot iteration method, characterized in that: The semantic slot iteration method is applied to a question-answering system. The question-answering system is provided with a synonym table, the synonym table including multiple semantic slot records, each of which includes a semantic slot and a plurality of synonyms of the semantic slot. The question-answering system is used to extract original semantic slot information from a user question, and when the original semantic slot information matches a first synonym in a first semantic slot record, determine a call parameter using the first semantic slot in the first semantic slot record to call an API to generate a question-answering result. The semantic slot iteration method includes: Obtaining user feedback on the question-and-answer results; When the user feedback is negative feedback, the original information of the semantic slot is added as a candidate word to the synonym selection table; When the synonym table is updated, determining the second semantic slot record to which the candidate word belongs; and The candidate word is added to the second semantic slot record.

2. The semantic slot iteration method according to claim 1, characterized in that: Also includes: When the user feedback is positive feedback, and the original information of the semantic slot is different from the first synonym, the original information of the semantic slot is added to the first semantic slot record.

3. The semantic slot iteration method according to claim 2, characterized in that: When the synonym table is updated, the step of determining the second semantic slot record to which the candidate word belongs includes: Monitoring positive feedback update events in which the user feedback is positive feedback; When the positive feedback update event is monitored, the second semantic slot record to which the candidate word belongs is determined.

4. The semantic slot iteration method according to claim 1, wherein: The step of determining the second semantic slot record to which the candidate word belongs includes: Using each word in the synonym table as a classification center point, classify the words in the synonym candidate table to determine the classification center point corresponding to the candidate word; and Determine the semantic slot record to which the classification center point corresponding to the candidate word belongs as the second semantic slot record.

5. The semantic slot iteration method according to claim 4, characterized in that: The synonym candidate table includes a plurality of candidate word records, each of which includes the synonym and a plurality of semantic slot original information matching the synonym; The step of adding the original semantic slot information as a candidate word to a synonym candidate table includes: searching the synonym candidate table for the first synonym, and when the first synonym is found, adding the original semantic slot information to a candidate word record to which the first synonym belongs; and when the first synonym is not found, creating a new candidate word record including the first synonym and the original semantic slot information; The step of determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record includes: marking the correspondence between the candidate word record where the candidate word is located and the semantic slot record to which the classification center point corresponding to the candidate word belongs; when confirmation information of the correspondence is received, determining that the semantic slot record to which the classification center point corresponding to the candidate word belongs is the second semantic slot record.

6. The semantic slot iteration method according to claim 1, characterized in that: The question-answering system uses the following steps to determine synonyms that match the original information of the semantic slot: Performing a first matching process on the original information of the semantic slot and the synonym table, wherein the first matching process is used to determine a first word in the synonym table that meets a first matching requirement with the original information of the semantic slot; When the first word is included in the synonym table, determining that the first word is a synonym that matches the original information of the semantic slot; When the first word is not included in the synonym table, performing a second matching process on the original semantic slot information and the synonym table, wherein the second matching process is used to determine a second word in the synonym table that meets a second matching requirement with the original semantic slot information, wherein the first matching requirement is higher than the second matching requirement; Determine that the second word is a synonym that matches the original information of the semantic slot.

7. The semantic slot iteration method according to claim 6, characterized in that: The step of performing a second matching process on the original semantic slot information and the synonym table includes: According to the trained Bert model, the original information of the semantic slot and the synonym table are converted into text vectors through word2vec technology; Calculating the cosine similarity between the text vector of the original information of the semantic slot and the text vector of the word in the synonym table; and The second word is determined according to the cosine similarity.

8. A semantic slot iteration device, characterized in that: include: The semantic slot iteration device is applied to a question-answering system, wherein the question-answering system is provided with a synonym table, wherein the synonym table includes a plurality of semantic slot records, each of which includes a semantic slot and a plurality of synonyms of the semantic slot. The question-answering system is used to extract original semantic slot information from a user question, and when the original semantic slot information matches a first synonym in a first semantic slot record, determine a call parameter using the first semantic slot in the first semantic slot record to call an API to generate a question-answering result; The semantic slot iteration device comprises: An acquisition module, configured to obtain user feedback on the question-and-answer results; A first processing module is configured to add the original information of the semantic slot as a candidate word to a synonym selection table when the user feedback is negative feedback; A second processing module is configured to determine a second semantic slot record to which the candidate word belongs when the synonym table is updated; and The third processing module is configured to add the candidate word to the second semantic slot record.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images