Intelligent inspection method and system for inspection work order based on semantic features

Abstract

The application discloses a semantic feature-based intelligent inspection method and system for an inspection work order, and relates to the technical field of inspection work order inspection. The method comprises the following steps: reading historical inspection work order data and identifying a set of identified noun candidates; calculating a set of synonymous expression update frequencies of the set of identified noun candidates; dividing the set of identified noun candidates into multiple sets of identified noun candidates according to the sizes of the set of synonymous expression update frequencies; connecting an inspection work order inspection system; performing priority rule inspection on a set of identified nouns of a current inspection work order; and outputting an identified noun inspection return result. The application solves the technical problem that a single inspection rule is used for identified nouns of an inspection work order in the prior art, which cannot adapt to the expression changes of nouns with different activity levels, resulting in low inspection efficiency and insufficient accuracy. The application achieves precise adaptation to nouns with different activity levels, and improves the inspection efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of inspection of work orders, and particularly relates to an intelligent inspection method and system for inspection of work orders based on semantic features. BACKGROUND

[0002] In the field of inspection of work orders, work order inspection is a key link for guaranteeing the execution of business specifications and improving service quality, and the core thereof needs to check the expression of identified terms in the work order. At present, traditional inspection methods mostly rely on manual checking one by one, which not only consumes a large amount of labor cost, but also easily leads to low inspection efficiency and missed inspection due to subjective judgment differences and fatigue of the workers when facing a large number of work orders. Some automatic inspection tools are introduced, but most of them use a single keyword matching rule, and the active differences of identified terms in actual business are not considered. The single rule cannot adapt to the checking needs of terms with different activity levels, and either leads to misjudgment of highly active terms due to rigid rules, or leads to missed inspection of low active terms due to insufficient coverage, so it is difficult to balance the efficiency and accuracy of inspection, and cannot meet the inspection needs of large-scale and refined inspection of inspection work orders.

[0003] The prior art uses a single inspection rule for identified terms of inspection work orders, which cannot adapt to the expression changes of terms with different activity levels, resulting in the technical problems of low inspection efficiency and insufficient accuracy. SUMMARY

[0004] The present application provides an intelligent inspection method and system for inspection of work orders based on semantic features, which is used to solve the technical problems that the prior art uses a single inspection rule for identified terms of inspection work orders, which cannot adapt to the expression changes of terms with different activity levels, resulting in low inspection efficiency and insufficient accuracy.

[0005] In view of the above problems, the present application provides an intelligent inspection method and system for inspection of work orders based on semantic features.

[0006] In a first aspect of the present application, an intelligent inspection method for inspection of work orders based on semantic features is provided, which comprises:

[0007] reading historical inspection work order data, identifying a set of identification noun candidates of the historical inspection work order data, the identification noun being a non-professional term noun of the work order text; calculating a set of synonymous expression update frequencies of the set of identification noun candidates according to semantic features; dividing the set of identification noun candidates into a plurality of sets of identification noun candidates according to the sizes of the set of synonymous expression update frequencies, the plurality of sets of identification noun candidates including at least a set of stable identification noun candidates, a set of low-active identification noun candidates, and a set of high-active identification noun candidates; connecting an inspection work order quality inspection system, the inspection work order quality inspection system performing priority rule quality inspection on a set of identification nouns of a current inspection work order according to the plurality of sets of identification noun candidates, and outputting an identification noun quality inspection return result, the rule quality inspection including a static keyword matching rule, a dynamic keyword matching rule, and a semantic model matching rule.

[0008] In a second aspect of the present application, a semantic feature-based intelligent inspection work order quality inspection system is provided, and the system includes:

[0009] a candidate set identification module configured to read historical inspection work order data, identify a set of identification noun candidates of the historical inspection work order data, the identification noun being a non-professional term noun of the work order text; an update frequency set calculation module configured to calculate a set of synonymous expression update frequencies of the set of identification noun candidates according to semantic features; a candidate set division module configured to divide the set of identification noun candidates into a plurality of sets of identification noun candidates according to the sizes of the set of synonymous expression update frequencies, the plurality of sets of identification noun candidates including at least a set of stable identification noun candidates, a set of low-active identification noun candidates, and a set of high-active identification noun candidates; and a rule quality inspection module configured to connect an inspection work order quality inspection system, the inspection work order quality inspection system performing priority rule quality inspection on a set of identification nouns of a current inspection work order according to the plurality of sets of identification noun candidates, and outputting an identification noun quality inspection return result, the rule quality inspection including a static keyword matching rule, a dynamic keyword matching rule, and a semantic model matching rule.

[0010] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0011] reading historical inspection work order data, identifying a set of identification noun candidates of the historical inspection work order data, the identification noun being a non-professional term noun of the work order text; calculating a set of synonymous expression update frequencies of the set of identification noun candidates according to semantic features; dividing the set of identification noun candidates into a plurality of sets of identification noun candidates according to the sizes of the set of synonymous expression update frequencies; connecting an inspection work order quality inspection system, the inspection work order quality inspection system performing priority rule quality inspection on a set of identification nouns of a current inspection work order according to the plurality of sets of identification noun candidates, and outputting an identification noun quality inspection return result. The technical effect of accurately adapting to nouns of different activity levels is achieved, and the quality inspection efficiency and accuracy are improved. BRIEF DESCRIPTION OF DRAWINGS

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic diagram of the intelligent quality inspection method for audit work orders based on semantic features provided in this application embodiment;

[0014] Figure 2 This is a schematic diagram of the structure of an intelligent quality inspection system for audit work orders based on semantic features, provided in an embodiment of this application.

[0015] Figure labeling: Candidate set identification module 10, update frequency set calculation module 20, candidate set partitioning module 30, rule quality inspection module 40. Detailed Implementation

[0016] This application provides an intelligent quality inspection method and system for audit work orders based on semantic features. It addresses the technical problem that existing technologies use a single quality inspection rule for audit work order identifiers, which cannot adapt to the changing expressions of terms with different activity levels, resulting in low quality inspection efficiency and insufficient accuracy.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides an intelligent quality inspection method for audit work orders based on semantic features, the method comprising:

[0019] Step S100: Read historical audit work order data and identify the candidate set of identifier terms in the historical audit work order data. The identifier terms are non-technical terms in the work order text.

[0020] Specifically, first, the historical inspection work order data stored in the database is called through the data interface, which covers the text information of the work order generated in the past inspection business, including the business description, processing object information, feedback content record and other text content of the work order. Then, the historical inspection work order text is preprocessed, and the work order text is disassembled and analyzed through natural language processing techniques such as text cleaning, word segmentation processing, and part-of-speech tagging. Subsequently, according to the definition that the identified noun is a non-professional term noun of the work order text, the professional term nouns related to the exclusive field of the inspection business are excluded, such as "inspection process code", "violation judgment standard" and other professional expressions directly related to the inspection business, and the nouns used to describe general information, non-business exclusive nouns in the work order text are retained, such as "customer name", "feedback date", "contact address", "submission channel" and the like. Finally, the filtered non-professional term nouns are integrated and summarized to form a candidate set of identified nouns corresponding to the historical inspection work order data, providing basic data support for subsequent semantic feature-based synonym expression update frequency calculation and other links.

[0021] Step S200: Calculate the synonym expression update frequency set of the candidate set of identified nouns according to the semantic features.

[0022] Specifically, the obtained candidate set of identified nouns is time-series data sliced according to the time dimension, and a plurality of continuous time-series slices are divided according to the generation period of the inspection work order, such as monthly, quarterly, each slice corresponding to a candidate subset of identified nouns in a specific time period, ensuring that the time variation rule of the synonym expression can be captured. Then, the semantic vectorization processing is performed on the candidate subset of identified nouns of each time-series slice, and each identified noun is converted into a high-dimensional semantic feature vector by means of a pre-trained semantic model such as a word embedding model, forming a semantic feature vector set to represent the semantic connotation of the noun in numerical form, providing a basis for synonym judgment. Subsequently, the semantic similarity of the same identified noun in different time-series slices is calculated based on the semantic feature vector set, and the nouns with a similarity reaching a preset threshold are classified into the same synonym expression set to determine the composition of the synonym expression corresponding to each time-series slice. Finally, the appearance frequency of each synonym expression in each time-series slice is counted, the frequency proportion is calculated to determine the dominant synonym expression of each slice, and whether the dominant synonym expression changes between adjacent time-series slices is compared, and the frequency proportion of the changed dominant synonym expression exceeds the preset threshold, the expression update event is recorded, and finally the update frequency of each identified noun synonym expression is calculated according to the number of update events, and the synonym expression update frequency set is integrated.

[0023] Step S300: Divide the candidate set of identified nouns into a plurality of candidate sets of identified nouns according to the size of the synonym expression update frequency set, the plurality of candidate sets of identified nouns at least including a stable type candidate set of identified nouns, a low active candidate set of identified nouns and a high active candidate set of identified nouns.

[0024] Specifically, in combination with the business scene characteristics of the inspection work order, the change rule of the historical identification noun synonymous expression and the quality inspection accuracy requirement, two frequency division thresholds, i.e., a first threshold and a second threshold, are preset. For example, according to the past data, the first threshold is set to 1 time per month and the second threshold is set to 3 times per month, and the specific values can be adjusted according to the actual business. Then, the calculated synonymous expression update frequency set is retrieved, and the synonymous expression update frequency value corresponding to each identification noun candidate is extracted one by one, and the value is sequentially compared with the preset first threshold and second threshold. Subsequently, the identification noun candidate set is classified according to the comparison result: if the synonymous expression update frequency of an identification noun is less than the first threshold, it means that its synonymous expression is stable and rarely changes, and it is classified into the stable identification noun candidate set, such as “customer name”, which is used as the only expression for a long time without other synonymous expression updates; if the update frequency is greater than or equal to the first threshold and less than the second threshold, it means that its synonymous expression occasionally appears adjustment, and the change frequency is low, and it is classified into the low-active identification noun candidate set, such as “business number”, which occasionally appears temporary replacement of synonymous expressions such as “work order number”; if the update frequency is greater than or equal to the second threshold, it means that its synonymous expression frequently changes due to business optimization, expression adjustment, etc., and it is classified into the high-active identification noun candidate set, such as “feedback type”, which frequently appears synonymous expression updates such as “problem feedback” and “complaint feedback” due to business classification addition or merger. Through the classification process, clear classification basis is provided for triggering differentiated quality inspection rules for different types of identification nouns in the subsequent steps, ensuring the pertinence and efficiency of the quality inspection process.

[0025] Step S400: Connect the inspection work order quality inspection system, which performs priority rule quality inspection on the identification noun set of the current inspection work order according to the multi-class identification noun candidate set, outputs the identification noun quality inspection return result, and the rule quality inspection includes static keyword matching rule, dynamic keyword matching rule and semantic model matching rule.

[0026] Specifically, a connection with the inspection system of the inspection work order is established, and the divided stable, low-active, and high-active three types of identification noun candidate sets are synchronized to the inspection system, thereby providing classification basis for subsequent priority rule matching. Then, the inspection system of the inspection work order processes the current inspection work order text to be inspected first, identifies the identification noun set in the text, i.e., the non-professional term noun in the work order, and then traverses each identification noun in the set to determine the type of the current noun, such as stable, low-active, or high-active, by comparing the noun with the three types of identification noun candidate sets. If the current noun belongs to the stable identification noun candidate set, a static keyword matching rule is triggered, a preset static keyword library is called, a standard expression of the stable noun is included, the current noun is accurately matched with the content of the keyword library, and the inspection is completed. If the current noun belongs to the low-active identification noun candidate set, a dynamic keyword matching rule is triggered, a dynamic keyword library including the static keyword library and the synonymous expansion keywords within a preset time window is used to match and inspect the current noun, thereby covering the occasional synonymous expressions of the low-active noun. If the current noun belongs to the high-active identification noun candidate set, a semantic model matching rule is triggered, a semantic model pre-trained based on a preset semantic sample is called, the similarity between the current noun and the preset semantic sample is calculated, and the inspection is completed through flexible matching at the semantic level. After all the identification nouns are traversed and matched, the inspection system of the inspection work order integrates the inspection results of each type of noun, such as matching qualified, matching unqualified, and unqualified reasons, and finally outputs the complete identification noun inspection return result, thereby realizing intelligent and differentiated inspection of the identification nouns of the current inspection work order.

[0027] In one possible implementation manner, step S200 further includes:

[0028] Step S210: performing time sequence data slicing on the identification noun candidate set to obtain a plurality of identification noun candidate subsets corresponding to a plurality of time sequence slices.

[0029] Step S220: performing semantic vectorization processing on the plurality of identification noun candidate sets to obtain a semantic feature vector set.

[0030] Step S230: performing similarity calculation on the identification noun candidate subset corresponding to each time sequence slice according to the semantic feature vector set, and obtaining a synonymous expression set of each time sequence slice according to the similarity calculation result.

[0031] Step S240: performing synonymous expression update frequency calculation according to the synonymous expression set of each time sequence slice, and outputting a synonymous expression update frequency set.

[0032] Specifically, the time division standard of the time series slice is determined in combination with the generation period of historical inspection work orders and the characteristics of business data, such as selecting division according to natural months, quarters or fixed time windows according to the work order generation density, to ensure that each slice can cover sufficient and time-representative work order data; then, the identified historical inspection work order identification noun candidate set composed of non-professional terms in the work order text is called, and the candidate set is split into multiple continuous and non-overlapping time series slices according to the preset time division standard, and each time series slice corresponds to a specific time period; finally, the identification nouns appearing in all historical inspection work orders in the time period are extracted from each time series slice to form an identification noun candidate subset corresponding to each time series slice.

[0033] First, a semantic encoding model pre-trained on a large-scale text corpus is called, such as a word embedding model or a context semantic model, which can capture the semantic association and context features of words; then, for the obtained multiple identification noun candidate subsets, all identification nouns in each subset are traversed, and each noun is input into the semantic encoding model for processing, and the model outputs corresponding high-dimensional numerical vectors, which represent the semantic connotation of different identification nouns and the semantic distance between them through the difference and association of numerical dimensions; finally, all identification noun corresponding high-dimensional vectors are stored in association with their belonging time series slice and the noun itself, and integrated to form a semantic feature vector set covering all identification nouns, realizing the conversion from text form identification nouns to numerical form semantic feature vectors.

[0034] For each time series slice corresponding identification noun candidate subset, the semantic feature vectors corresponding to all identification nouns in the subset are extracted from the obtained semantic feature vector set to form a vector subset matching the time series slice; then, a vector similarity calculation algorithm such as cosine similarity algorithm is used to calculate each two identification noun semantic feature vectors in the vector subset in pairs to obtain a similarity value representing the degree of semantic association between them; then, according to the definition standard of "synonyms" in the inspection work order business scenario, a similarity threshold is preset, such as 0.85, which can be adjusted according to actual semantic matching requirements, and identification nouns with a similarity value greater than or equal to the threshold are determined as synonyms; finally, after completing the pair-wise similarity calculation and synonym judgment for all identification nouns in each time series slice, the identification nouns determined as synonyms are grouped, and all groups are integrated to form the synonym expression set of the time series slice, realizing the clustering and integration of identification nouns with the same or similar semantics in the same time period.

[0035] For each synonym expression set of each time slice, the frequency of each group of synonym expressions in the corresponding historical inspection work order of the slice is counted, the frequency proportion of each group of synonym expressions is calculated, that is, the proportion of the frequency of the group to the total frequency of all identified nouns in the slice, and the dominant synonym expression frequency proportion of each group is determined according to the frequency proportion. If there is a same proportion, further determination is made in combination with the business commonly used expression. Then, the dominant synonym expressions of the same type of identified nouns in adjacent two time slices are compared in time sequence: if the dominant synonym expressions of adjacent slices change, and the frequency proportion of the new dominant synonym expression after the change exceeds a preset threshold, such as 0.5, the change is ensured to have effectiveness rather than accidental fluctuation, and then an expression update event for the identified noun is recorded. Subsequently, all time slices are traversed, the total number of expression update events of each identified noun in the entire time range is counted, and the synonym expression update frequency of the identified noun is obtained by dividing the total number by the total number of time slices or the total time span. Finally, the synonym expression update frequencies of all identified nouns are integrated in a unified format to form and output the synonym expression update frequency set.

[0036] In a possible implementation manner, step S240 further includes:

[0037] Step S241: calculating the frequency proportion of each synonym expression according to the synonym expression set of each time slice, and determining the dominant synonym expression based on the frequency proportion of each synonym expression.

[0038] Step S242: comparing the dominant synonym expressions of adjacent time slices in the plurality of time slices, and recording an expression update event.

[0039] Step S243: calculating the synonym expression update frequency of each synonym expression according to the number of expression update events, and outputting the synonym expression update frequency set.

[0040] Specifically, for the synonym expression set corresponding to each time slice, each group of synonym expressions in the set is traversed, the frequency of each synonym expression in the group in the historical inspection work order covered by the time slice, that is, the number of times the expression is used in the work order text, is counted. Then, the frequency proportion of each synonym expression is calculated, that is, the ratio of the frequency of the synonym expression to the total frequency of all synonym expressions in the group, so as to quantify the use proportion of the expression in the same group of synonym expressions. Finally, the expression with the highest frequency proportion in each group of synonym expressions is determined as the dominant synonym expression of the group. If there are multiple expressions with the same frequency proportion and all are the highest, a secondary determination is made in combination with the standard expression or commonly used expression in the inspection business scenario, so as to determine the main expression form of each type of identified noun in each time slice, and lay a foundation for the expression change comparison between adjacent slices.

[0041] The dynamic change of the synonymous expression is captured by time sequence comparison. Firstly, the multiple time sequence slices are sorted in chronological order to form a continuous time sequence. Then, for each synonymous expression group corresponding to the identified noun, the dominant synonymous expression of the group in the adjacent two time sequence slices is extracted in turn, i.e. the expression with the highest frequency ratio is determined. Then, for the case that the expression forms are different but the semantics are consistent, it is judged whether the dominant synonymous expressions of the adjacent time sequence slices are different through string exact matching or semantic similarity verification. If they are different, and the frequency ratio of the dominant synonymous expression in the new time sequence slice exceeds the preset validity threshold, such as 50%, it is ensured that the change has stability rather than accidental fluctuation, it is determined that the synonymous expression of the identified noun is updated, an expression update event is recorded, and the corresponding time sequence slice time information is associated. In this way, the change node of the synonymous expression over time is accurately captured, and the event basis is provided for subsequent update frequency calculation.

[0042] All identified nouns are traversed, and the total number of expression update events related to each identified noun is recorded. The number directly reflects the change frequency of the synonymous expression of the identified noun in the entire time sequence range. Then, the total number of time sequence slices or the total time span, such as the total number of months, is combined to calculate the synonymous expression update frequency of each identified noun, i.e. the ratio of the total number of expression update events to the total number of time sequence slices or the total time span. In this way, the change rate of the synonymous expression over time is quantified. Finally, the synonymous expression update frequencies of all identified nouns are integrated in a unified data format to form and output a synonymous expression update frequency set. The set clearly presents the dynamic change characteristics of the synonymous expression of each identified noun.

[0043] In one possible implementation manner, step S240 further includes:

[0044] The dominant synonymous expressions of adjacent time sequence slices in the multiple time sequence slices are compared, and expression update events are recorded.

[0045] The expression update event is the change of the dominant synonymous expressions of adjacent time sequence slices, and the frequency ratio of the updated dominant synonymous expression exceeds the preset threshold.

[0046] Specifically, in the intelligent quality inspection process of the inspection work order, the accurate capture of synonymous expression changes is realized through dynamic tracking of time sequence slices. First, multiple time sequence slices are arranged in chronological order to form a continuous time sequence. Then, for each synonymous expression group corresponding to an identified noun, the dominant synonymous expression in the group, i.e., the expression with the highest frequency ratio, is extracted from the adjacent two time sequence slices. Whether the two expressions are consistent is determined through string comparison or semantic similarity calculation. If there is a difference between the dominant synonymous expressions of adjacent slices, and the frequency ratio of the updated dominant synonymous expression in the new slice exceeds a preset threshold, such as 50%, ensuring that the change is stable, an expression update event is recorded, and the corresponding identified noun and time sequence information are associated. In this way, the substantial changes in synonymous expressions over time are captured, providing reliable event basis for subsequent update frequency calculation.

[0047] The definition of expression update events needs to meet two core conditions to ensure the effectiveness and stability of the recorded changes at the business level. The first condition is that the dominant synonymous expressions of adjacent time sequence slices change, i.e., for the synonymous expression group of the same identified noun, the dominant synonymous expression in the previous time sequence slice is different from the dominant synonymous expression in the next adjacent time sequence slice in terms of expression form or semantic connotation. String matching or semantic similarity verification is needed to exclude non-substantial changes where the expressions are different but the semantics are consistent. The second condition is that the frequency ratio of the updated dominant synonymous expression exceeds a preset threshold. The updated dominant synonymous expression here refers to the new dominant synonymous expression in the next time sequence slice. The frequency ratio data of this expression in the slice, i.e., the proportion of the frequency of this expression to the total frequency of the synonymous expression group, is retrieved and compared with the pre-set threshold, such as 50%. The specific value can be adjusted according to the demand of the inspection business for expression stability. Only when the frequency ratio exceeds the threshold can it be considered that the new dominant synonymous expression has formed a stable usage trend in this time period, rather than a few expressions that appear accidentally. Only when both conditions are met can an effective expression update event be determined, avoiding the misjudgment of accidental expression fluctuations as effective updates, and providing a reliable basis for the accurate calculation of synonymous expression update frequency.

[0048] In one possible implementation manner, step S300 further includes:

[0049] Step S310: setting a first threshold and a second threshold.

[0050] Step S320: dividing the identified noun candidate set smaller than the first threshold into a stable identified noun candidate set, dividing the identified noun candidate set greater than or equal to the first threshold and smaller than the second threshold into a low-active identified noun candidate set, and dividing the identified noun candidate set greater than or equal to the second threshold into a high-active identified noun candidate set according to the size of the synonymous expression update frequency set.

[0051] Specifically, the setting of the focus threshold, combined with the update rule of identifying the synonym expression of the historical inspection work order, the demand of the inspection business for the stability of expression, such as the stable noun needs long-term and infrequent change, the high active noun is related to business adjustment, and the numerical distribution characteristics of the synonym expression update frequency set, such as the median and quartile of the statistical frequency data, two key division thresholds, the first threshold and the second threshold, are determined. For example, if the annual update frequency of most identifying nouns in the historical data is less than 2 times, the first threshold can be set to "2 times / year"; if the high-frequency update nouns are mostly concentrated in more than 5 times per year, the second threshold can be set to "5 times / year". The specific value can be dynamically adjusted according to the business scene to ensure that the threshold can effectively distinguish identifying nouns with different update activity levels.

[0052] Based on the set threshold and the synonym expression update frequency set, the synonym expression update frequency set is first called, and the update frequency value corresponding to each identifying noun candidate is extracted. Then, the update frequency of each identifying noun is compared with the first threshold and the second threshold in turn: if the synonym expression update frequency of a certain identifying noun is less than the first threshold, it means that the synonym expression remains stable and rarely changes for a long time, and only a single expression exists for a long time, which is classified into the stable identifying noun candidate set; if the update frequency is greater than or equal to the first threshold and less than the second threshold, it means that the synonym expression has a small amount of update but low frequency, and occasionally appears synonym expression change due to record specification fine-tuning, which is classified into the low active identifying noun candidate set; if the update frequency is greater than or equal to the second threshold, it means that the synonym expression frequently changes due to business optimization, classification adjustment, etc., and the synonym expression update frequently appears due to the addition of feedback scenarios, which is classified into the high active identifying noun candidate set.

[0053] In one possible implementation manner, step S400 further includes:

[0054] Step S410: The inspection work order quality inspection system identifies the set of identifying nouns of the current inspection work order, iterates through the set of identifying nouns of the current inspection work order, and judges the type to which the current noun belongs.

[0055] Step S420: If the type to which the current noun belongs is the stable identifying noun candidate set, the static keyword matching rule is triggered for quality inspection; if the type to which the current noun belongs is the low active identifying noun candidate set, the dynamic keyword matching rule is triggered for quality inspection; if the type to which the current noun belongs is the high active identifying noun candidate set, the semantic model matching rule is triggered for quality inspection.

[0056] Step S430: Until the iteration is completed, the identifying noun quality inspection return result is output.

[0057] Specifically, the current to-be-inspected audit work order text is processed, and an identification noun recognition logic is adopted to extract all identification nouns from the work order content based on text segmentation and non-professional term screening to form an identification noun set of the current audit work order. Then, each identification noun in the set is sequentially traversed and compared with the divided stable identification noun candidate set, low-active identification noun candidate set, and high-active identification noun candidate set, respectively, to determine to which candidate set the identification noun belongs through accurate name matching or semantic similarity secondary verification, so as to determine the type of the identification noun.

[0058] Differentiated inspection rules are executed for different types of identification nouns. If the current noun belongs to the stable identification noun candidate set, a static keyword matching rule is triggered to perform accurate string comparison between the noun and a preset static keyword library containing standard expressions and fixed synonymous expressions verified for a long time to complete the inspection. If the noun belongs to the low-active identification noun candidate set, a dynamic keyword matching rule is triggered to call a dynamic library integrating static keywords and newly added synonymous expressions in recent period to cover a small amount of expression changes through flexible matching. If the noun belongs to the high-active identification noun candidate set, a semantic model matching rule is triggered by the system. The model is constructed based on a Transformer architecture, takes multiple versions of synonymous expressions of historical high-active nouns as training samples, learns semantic correlation features through pre-training, and adapts to the audit work order business scenario through transfer learning. Finally, the noun can be converted into a semantic vector and the similarity can be calculated to achieve flexible matching inspection of frequently changing synonymous expressions.

[0059] After the inspection of all identification nouns of the current audit work order is completed, the results are output. The identification noun set of the current work order is sequentially and continuously traversed, and the type determination and corresponding rule inspection operation are sequentially performed on each noun. The inspection states of the nouns, such as qualified, unqualified, type, and specific abnormal information, such as static matching inconsistency and insufficient semantic similarity, are recorded synchronously. After all identification nouns in the set are processed, the scattered inspection records are integrated into structured data to form complete inspection details of each identification noun, and the identification noun inspection return result is finally output to provide specific and traceable basis for overall quality evaluation of the audit work order.

[0060] In one possible implementation manner, step S420 further includes:

[0061] Step S421: If the type of the current noun is the stable identification noun candidate set, a static keyword matching rule is triggered, and a static keyword library of the static keyword matching rule is used to read the stable identification noun candidate set for matching inspection.

[0062] Specifically, when it is determined that the current noun belongs to the stable identification noun candidate set, a static keyword matching rule is triggered immediately, which calls a specially constructed static keyword library storing standard expressions of the stable identification noun confirmed by business specifications and long-term synonymous expressions, and the content is only updated when major adjustments are made to the business bottom specification, maintaining high stability. Through a data interaction interface, all keyword entries corresponding to the stable identification noun candidate set are extracted from the static keyword library, and then the current noun to be inspected is compared with the entries in the library one by one by using string exact matching. If there is a completely consistent matching item, it is determined that the noun is qualified; if no item is matched, it is marked as unqualified, and the specific information that does not conform to the static keyword library expression is recorded synchronously, ensuring that the expression specification of the stable noun is strictly checked.

[0063] In one possible implementation manner, step S420 further includes:

[0064] Step S422: wherein the dynamic keyword matching rule includes a dynamic keyword library, and the dynamic keyword library includes a static keyword library and a synonymous extended keyword library within a preset time window.

[0065] Step S423: reading the low-active identification noun candidate set for matching inspection by using the dynamic keyword library of the dynamic keyword matching rule.

[0066] Specifically, it is clarified that the core of the dynamic keyword matching rule is the dynamic keyword library, which adopts a double-layer integrated architecture: on the one hand, the static keyword library is completely included to ensure that the basic standard expression of the low-active identification noun is stably covered; on the other hand, the synonymous extended keyword library within a preset time window is included, and the extended library dynamically introduces high-frequency synonyms in the past 3 months in the inspection work order through a system automatic mechanism. These synonyms need to meet the preset threshold of the appearance frequency and be confirmed as valid synonymous expressions through semantic verification, so as to form a dynamically updated extended word set. Through joint calling of the static basic word library and the dynamic extended word library, the dynamic keyword library can not only guarantee the basic specification of the expression of the low-active identification noun, but also cover reasonable expression changes within a preset time range, providing comprehensive and time-effective word source support for subsequent matching inspection.

[0067] When it is determined that the current noun belongs to the low-active identification noun candidate set, the dynamic keyword matching rule is started, and a dynamic keyword library which is the core component of the dynamic keyword matching rule is called. The dynamic keyword library integrates the basic standard expressions of the static keyword library and the synonymous extended keywords in the preset time window, including the effective synonyms appearing recently with high frequency. All keyword entries corresponding to the low-active identification noun candidate set are read from the dynamic keyword library through a data interface, including the static basic words and the dynamic extended words, to form a complete matching word pool. Then, a multi-mode matching algorithm is used to compare the current noun to be inspected with all entries in the word pool one by one. If the current noun matches any keyword in the word pool successfully, whether it is a static standard expression or a recent high-frequency synonymous expression, it is determined that the noun passes the inspection. If no entry is matched, it is marked as unqualified, and the specific information of “no static basic word or recent extended word in the dynamic keyword library” is recorded, which ensures the expression standardization of the low-active noun and flexibly adapts to its limited expression updating characteristics.

[0068] In a possible implementation manner, the step S420 further includes:

[0069] The step S424: the semantic model matching rule includes a semantic model pre-trained based on a preset semantic sample.

[0070] The step S425: similarity matching inspection is performed on the high-active identification noun candidate set and the preset semantic sample according to the pre-trained semantic model.

[0071] Specifically, the semantic model matching rule is that a semantic model pre-trained based on a preset semantic sample is adopted by using a Transformer architecture. The model takes the preset semantic sample of the high-active identification noun as the training basis, and the sample covers the standard expression and various synonymous variants generated at different times, which are all verified by business to have semantic correlation. The model can simultaneously capture the bidirectional dependency relationship between words in the expression and the context semantic correlation by using the multi-head self-attention mechanism of the Transformer architecture, and converts the literal expression into a high-dimensional vector containing deep semantic information through the pre-training process. During training, the basic language rules are learned on the general text corpus first, and then the synonymous expression samples in the inspection work order field are used for fine-tuning, so that the model is accurately adapted to the expression characteristics of the high-active noun in the business scenario, and finally has the ability to judge whether the expressions are synonymous by calculating the semantic vector similarity, thereby providing flexible and accurate algorithm support for dynamic inspection of the high-active identification noun.

[0072] When it is determined that the current noun belongs to the high active identification noun candidate set, the pre-trained semantic model is called. The current high active identification noun to be inspected and the preset semantic sample, including the standard expression of the noun, the historical synonymous variant and the recent newly added effective expression, are input into the model. The model converts the two into high-dimensional vectors containing deep semantic information through the internal multi-head self-attention mechanism and the full connection layer, and then calculates the similarity value between the two vectors by using the cosine similarity algorithm. Then, the calculated similarity value is compared with the preset threshold value, i.e. the threshold value determined through the cross validation of the business scene, such as 0.85. If the similarity is greater than or equal to the threshold value, it is determined that the current high active identification noun is consistent with the preset semantic sample in semantics, and the inspection is qualified. If the similarity is less than the threshold value, it is marked as unqualified, and the specific similarity value and the corresponding preset semantic sample are recorded synchronously, which not only adapts to the frequent expression changes of high active nouns, but also ensures the accuracy and traceability of the inspection results.

[0073] In the second embodiment, based on the same inventive concept as the semantic feature-based intelligent inspection work order quality inspection method in the foregoing embodiments, as shown in the following table, the present application provides a semantic feature-based intelligent inspection work order quality inspection system, and the system and method embodiments in the present application are based on the same inventive concept. The system includes: Figure 2

[0074] The candidate set identification module 10 is configured to read historical inspection work order data, and identify an identification noun candidate set of the historical inspection work order data. The identification noun is a non-professional term noun of the work order text.

[0075] The update frequency set calculation module 20 is configured to calculate a synonymous expression update frequency set of the identification noun candidate set according to semantic features.

[0076] The candidate set division module 30 is configured to divide the identification noun candidate set into multiple identification noun candidate sets according to the size of the synonymous expression update frequency set, wherein the multiple identification noun candidate sets at least include a stable identification noun candidate set, a low active identification noun candidate set and a high active identification noun candidate set.

[0077] The rule inspection module 40 is configured to be connected to an inspection work order inspection system. The inspection work order inspection system performs priority rule inspection on a set of identification nouns of a current inspection work order according to the multiple identification noun candidate sets, and outputs an identification noun inspection return result. The rule inspection includes a static keyword matching rule, a dynamic keyword matching rule and a semantic model matching rule.

[0078] Further, the system is further configured to implement the following functions:

[0079] ​The set of identified noun candidates is sliced by time data to obtain a plurality of identified noun candidate subsets corresponding to a plurality of time slices; the plurality of sets of identified noun candidates is subjected to semantic vectorization processing to obtain a set of semantic feature vectors; similarity calculation is performed on the identified noun candidate subset corresponding to each time slice according to the set of semantic feature vectors, and a synonymous expression set of each time slice is obtained according to the similarity calculation result; and synonymous expression update frequency calculation is performed according to the synonymous expression set of each time slice, and a set of synonymous expression update frequencies is output.

[0080] Further, the system is also used to implement the following functions:

[0081] The frequency proportion of each synonymous expression is calculated according to the synonymous expression set of each time slice, the dominant synonymous expression is determined based on the frequency proportion of each synonymous expression, the dominant synonymous expressions of adjacent time slices in the plurality of time slices are compared, and an expression update event is recorded; the synonymous expression update frequency of each synonymous expression is calculated according to the number of expression update events, and a set of synonymous expression update frequencies is output.

[0082] Further, the system is also used to implement the following functions:

[0083] The dominant synonymous expressions of adjacent time slices in the plurality of time slices are compared, and an expression update event is recorded; wherein the expression update event is an event that the dominant synonymous expression of adjacent time slices changes and the frequency proportion of the updated dominant synonymous expression exceeds a preset threshold.

[0084] Further, the system is also used to implement the following functions:

[0085] A first threshold and a second threshold are set; according to the size of the set of synonymous expression update frequencies, the set of identified noun candidates less than the first threshold is divided into a stable set of identified noun candidates, the set of identified noun candidates greater than or equal to the first threshold and less than the second threshold is divided into a low-active set of identified noun candidates, and the set of identified noun candidates greater than or equal to the second threshold is divided into a high-active set of identified noun candidates.

[0086] Further, the system is also used to implement the following functions:

[0087] The inspection work order quality inspection system identifies a set of identified nouns of a current inspection work order, iterates through the set of identified nouns of the current inspection work order, and judges the type to which the current noun belongs; if the type to which the current noun belongs is a stable set of identified noun candidates, a static keyword matching rule is triggered for quality inspection, if the type to which the current noun belongs is a low-active set of identified noun candidates, a dynamic keyword matching rule is triggered for quality inspection, and if the type to which the current noun belongs is a high-active set of identified noun candidates, a semantic model matching rule is triggered for quality inspection; until the iteration is completed, a return result of identified noun quality inspection is output.

[0088] Further, the system is further used to realize the following functions:

[0089] If the type of the current noun belongs to the stable type identification noun candidate set, a static keyword matching rule is triggered, and the stable type identification noun candidate set is read by using a static keyword library of the static keyword matching rule for matching quality inspection.

[0090] Further, the system is further used to realize the following functions:

[0091] The dynamic keyword matching rule includes a dynamic keyword library, and the dynamic keyword library includes a static keyword library and a synonym expansion keyword library in a preset time window; the low active identification noun candidate set is read by using a dynamic keyword library of the dynamic keyword matching rule for matching quality inspection.

[0092] Further, the system is further used to realize the following functions:

[0093] The semantic model matching rule includes a semantic model pre-trained based on a preset semantic sample; the high active identification noun candidate set is read according to the pre-trained semantic model to perform similarity matching quality inspection on the preset semantic sample.

[0094] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0095] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0096] The specification and drawings are only exemplary of the present application, and any and all modifications, changes, combinations or equivalents within the scope of the present application are considered to be covered by the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A method for intelligent quality inspection of an inspection order based on semantic features, characterized in that, The method comprises: reading historical inspection work order data, identifying a set of identified noun candidates of the historical inspection work order data, the identified noun being a non-professional term noun of the work order text; calculating a set of synonym expression update frequencies of the set of identified noun candidates according to semantic features; dividing the set of identified noun candidates into a plurality of sets of identified noun candidates according to the sizes of the set of synonym expression update frequencies, the plurality of sets of identified noun candidates including at least a stable set of identified noun candidates, a low-active set of identified noun candidates, and a high-active set of identified noun candidates; connecting an inspection work order quality inspection system, the inspection work order quality inspection system performing priority rule quality inspection on a set of identified nouns of a current inspection work order according to the plurality of sets of identified noun candidates, outputting an identified noun quality inspection return result, and the rule quality inspection including a static keyword matching rule, a dynamic keyword matching rule, and a semantic model matching rule; calculating a set of synonym expression update frequencies of the set of identified noun candidates according to semantic features, the method comprising: performing time series data slicing on the set of identified noun candidates, obtaining a plurality of sets of identified noun candidate subsets corresponding to a plurality of time series slices; performing semantic vectorization processing on the plurality of sets of identified noun candidates, obtaining a set of semantic feature vectors; performing similarity calculation on the set of identified noun candidate subsets corresponding to each time series slice according to the set of semantic feature vectors, and obtaining a set of synonym expressions of each time series slice according to the similarity calculation result; performing synonym expression update frequency calculation on the set of synonym expressions of each time series slice, and outputting a set of synonym expression update frequencies, the method comprising: calculating a frequency proportion of each synonym expression according to the set of synonym expressions of each time series slice, and determining a dominant synonym expression based on the frequency proportion of each synonym expression; comparing the dominant synonym expressions of adjacent time series slices in the plurality of time series slices, and recording expression update events; calculating a synonym expression update frequency of each synonym expression according to the number of expression update events, and outputting a set of synonym expression update frequencies. 2.The method of claim 1, wherein, comparing the dominant synonym expressions of adjacent time series slices in the plurality of time series slices, and recording expression update events; wherein the expression update event is an event in which the dominant synonym expression of adjacent time series slices changes, and the frequency proportion of the updated dominant synonym expression exceeds a preset threshold. 3.The method of claim 1, wherein, dividing the set of identified noun candidates into a plurality of sets of identified noun candidates according to the sizes of the set of synonym expression update frequencies, the method comprising: setting a first threshold and a second threshold; dividing the set of identified noun candidates into a stable set of identified noun candidates, a low-active set of identified noun candidates, and a high-active set of identified noun candidates according to the sizes of the set of synonym expression update frequencies, wherein the set of identified noun candidates less than the first threshold is divided into the stable set of identified noun candidates, the set of identified noun candidates greater than or equal to the first threshold and less than the second threshold is divided into the low-active set of identified noun candidates, and the set of identified noun candidates greater than or equal to the second threshold is divided into the high-active set of identified noun candidates. 4.The method of claim 1, wherein, the inspection work order quality inspection system performing priority rule quality inspection on a set of identified nouns of a current inspection work order according to the plurality of sets of identified noun candidates, the method comprising: the inspection work order quality inspection system identifying a set of identified nouns of a current inspection work order, traversing the set of identified nouns of the current inspection work order, and judging the type of a current noun; If the current noun belongs to the stable type of identification noun candidate set, trigger the static keyword matching rule for quality inspection, if the current noun belongs to the low active identification noun candidate set, trigger the dynamic keyword matching rule for quality inspection, if the current noun belongs to the high active identification noun candidate set, trigger the semantic model matching rule for quality inspection; Until the traversal is completed, the identification noun quality inspection return result is output.

5. The method of claim 4, wherein the semantic feature-based intelligent quality inspection of the inspection ticket is performed by a computer system. If the current noun belongs to the stable type of identification noun candidate set, trigger the static keyword matching rule, and read the stable type of identification noun candidate set using the static keyword library of the static keyword matching rule for matching quality inspection. 6.The method of claim 4, wherein, If the current noun belongs to the low active identification noun candidate set, trigger the dynamic keyword matching rule for quality inspection, method Comprising: Wherein, the dynamic keyword matching rule includes a dynamic keyword library, and the dynamic keyword library includes a static keyword library and a synonym expansion keyword library within a preset time window; Read the low active identification noun candidate set using the dynamic keyword library of the dynamic keyword matching rule for matching quality inspection.

7. The method of claim 4, wherein the semantic feature-based intelligent quality inspection of the inspection ticket is performed by a computer system. If the current noun belongs to the high active identification noun candidate set, trigger the semantic model matching rule for quality inspection, method comprising: The semantic model matching rule includes a pre-trained semantic model based on a preset semantic sample; According to the pre-trained semantic model, read the high active identification noun candidate set and the preset semantic sample for similarity matching quality inspection.

8. The intelligent quality inspection system for inspection work order based on semantic features, characterized in that, The system is used to implement the semantic feature-based inspection work order intelligent quality inspection method of any one of claims 1-7, and the system comprises: A candidate set identification module is configured to read historical inspection work order data, identify an identification noun candidate set of the historical inspection work order data, and identify a noun as a non-professional term noun of a work order text; An update frequency set calculation module is configured to calculate a synonym expression update frequency set of the identification noun candidate set according to semantic features; A candidate set division module is configured to divide the identification noun candidate set into multiple types of identification noun candidate sets according to the size of the synonym expression update frequency set, wherein the multiple types of identification noun candidate sets at least include a stable type of identification noun candidate set, a low active identification noun candidate set, and a high active identification noun candidate set; A rule quality inspection module is configured to connect an inspection work order quality inspection system, wherein the inspection work order quality inspection system performs priority rule quality inspection on an identification noun set of a current inspection work order according to the multiple types of identification noun candidate sets, outputs an identification noun quality inspection return result, and rule quality inspection includes a static keyword matching rule, a dynamic keyword matching rule, and a semantic model matching rule.

Images