Work order duplicate checking method and device, equipment, storage medium and product
By summarizing work orders and calculating feature vector similarity, duplicate work orders can be quickly identified, solving the problem of long processing time in existing technologies and improving work order processing efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA MOBILE ZHIJIE TECHNOLOGY (BEIJING) CO LTD
- Filing Date
- 2025-12-15
- Publication Date
- 2026-05-08
AI Technical Summary
Existing methods for checking duplicate work orders are time-consuming, causing work order processing personnel to handle duplicate work orders, resulting in a waste of manpower.
By acquiring historical work orders and work orders to be checked for duplicates, feature vectors are obtained through summarization, and similarity is calculated to quickly determine whether a work order to be checked for duplicates is a duplicate work order.
It enables the rapid and accurate identification of duplicate work orders, reducing the time spent processing duplicate work orders and lowering labor costs.
Smart Images
Figure CN121998290A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of data processing, specifically relating to a method, apparatus, equipment, storage medium, and product for work order deduplication. Background Technology
[0002] With the development of online services (such as the widespread use of online hotlines), the number of work orders is constantly accumulating. Users may call repeatedly within a short period of time to report the same problem, or different users may call to report similar problems. This can lead to work order processing personnel having to handle duplicate work orders, resulting in wasted manpower. In this situation, by querying duplicate work orders, agents can view the progress of users' historical duplicate work orders and provide timely feedback to the caller. On the other hand, work order processing personnel can consolidate duplicate work orders for processing, thereby reducing the manpower cost required for work order processing. However, existing work order deduplication methods are time-consuming.
[0003] Therefore, a method is needed to quickly check for duplicate work orders. Summary of the Invention
[0004] This application provides a method for checking duplicate work orders, which can quickly check for duplicate work orders.
[0005] In a first aspect, embodiments of this application provide a method for checking duplicate work orders. The method includes: acquiring multiple historical work orders and acquiring work orders to be checked for duplicates, wherein the historical work orders are historically generated work order data, and the work orders to be checked for duplicates are work order data that need to be checked for duplicates; summarizing each of the historical work orders to obtain multiple first summary work orders, and summarizing the work orders to be checked for duplicates to obtain second summary work orders; determining a first feature vector corresponding to each of the first summary work orders and a second feature vector corresponding to each of the second summary work orders; determining a first similarity between each of the first feature vectors and the second feature vectors, and determining whether the work order to be checked for duplicates is a duplicate work order based on the first similarity.
[0006] Secondly, embodiments of this application provide a work order deduplication device, which includes: a first acquisition module, used to acquire multiple historical work orders and acquire work orders to be deduplicated, wherein the historical work orders are historically generated work order data, and the work orders to be deduplicated are work order data that need to be deduplicated; a first summary module, used to summarize each of the historical work orders to obtain multiple first summary work orders, and to summarize the work orders to be deduplicated to obtain second summary work orders; a first determination module, used to determine a first feature vector corresponding to each of the first summary work orders and a second feature vector corresponding to the second summary work order; and a second determination module, used to determine a first similarity between each of the first feature vectors and the second feature vectors, and to determine whether the work order to be deduplicated is a duplicate work order based on the first similarity.
[0007] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.
[0008] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.
[0009] Fifthly, embodiments of this application provide a computer program product that, when executed by a processor, implements the steps of the method described in the first aspect.
[0010] In a sixth aspect, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.
[0011] In this embodiment, by acquiring multiple historical work orders and work orders to be checked for duplicates, where the historical work orders are historically generated work order data and the work orders to be checked for duplicates are work order data that need to be checked for duplicates; summarizing each of the historical work orders to obtain multiple first summary work orders, and summarizing the work orders to be checked for duplicates to obtain second summary work orders; determining the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders; determining the first similarity between each of the first feature vectors and the second feature vectors, and determining whether the work order to be checked for duplicates is a duplicate work order based on the first similarity, work order duplication can be quickly performed. Attached Figure Description
[0012] Figure 1 This is a flowchart illustrating a work order deduplication method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating the second work order deduplication method provided in this application embodiment; Figure 3 This is a flowchart illustrating the third work order deduplication method provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of a work order deduplication device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a work order deduplication device provided in an embodiment of this application. Detailed Implementation
[0013] 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 some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0014] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.
[0015] The work order deduplication method provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.
[0016] Figure 1 This illustration shows a work order deduplication method according to an embodiment of the present invention. The method can be executed by an electronic device, which may include a server and / or a terminal device, wherein the terminal device may be, for example, a vehicle-mounted terminal or a mobile terminal. In other words, the method can be executed by software or hardware installed in a work order deduplication device, and the method includes the following steps: Step 102: Obtain multiple historical work orders and obtain work orders to be checked for duplicates.
[0017] The historical work orders refer to historically generated work order data, and the work orders to be checked for duplicates refer to work order data that need to be checked for duplicates.
[0018] The execution entity of the work order deduplication method described in this application can be a work order deduplication system, work order deduplication software, or other execution entities. Here, we take a work order deduplication system as an example for explanation.
[0019] Before performing work order deduplication, the work order deduplication system first acquires multiple historical work orders, which are work order data generated in the past. During the work order deduplication process, the system needs to acquire the work orders to be checked, which are the work order data that require deduplication. The format of the historical work orders acquired by the system and the work orders to be checked can be the same or different. However, if their formats differ, the system needs to adjust their formats to make them identical.
[0020] Work order data includes various data types, such as work order title data, work order address data, and work order content data. Work order title data is related to the work order title; work order address data is the address where the problem described in the work order occurred; and work order content data is a description of the problem. Work order title data is text data and serves as the name of the work order. Work order address data is structured address data, including province / city / district / street / house number / specific location, such as "Kitchen, Unit 2, Building 5, Xingfu Community, No. 100 Xingfu Road, XX District, XX City, XX Province". Work order content data is long text data, containing a problem description, urgency level, etc. For example, a complete work order could be: "Work Order 1: My home is in Building 5, Xingfu Community. The kitchen faucet is leaking, and the floor is wet. Repair service is needed."
[0021] After acquiring multiple historical work orders and / or work orders to be checked for duplicates, the work order deduplication system can also preprocess the unstructured data (such as work order content) in the collected historical work orders and / or work orders to be checked for duplicates, such as removing special characters and redundant spaces, and unifying the text encoding (UTF-8) for subsequent processing.
[0022] Step 104: Summarize each of the historical work orders to obtain multiple first summary work orders, and summarize the work orders to be checked for duplicates to obtain second summary work orders.
[0023] After acquiring multiple historical work orders, the work order deduplication system summarizes each historical work order based on a preset historical work order summary strategy, resulting in a first summary work order for each historical work order. After acquiring the work order to be checked for deduplication, the system summarizes it based on a preset summary strategy for the work order to be checked, resulting in a second summary work order for the work order to be checked. Furthermore, the preset historical work order summary strategy and the work order to be checked summary strategy can be the same or different, but both strategies can summarize the key elements in the work order data.
[0024] Specifically, the strategies for summarizing historical work orders and / or work orders to be deduplicated can be strategies that use a pre-trained work order summary model, or strategies that use template filling, traditional summarization algorithms, or fine-tuned generative models. Template filling involves using a template plus rules to fill in the work order information, such as using regular expressions / NER to extract fields (phone number, fault symptoms, root causes, etc.), then applying a fixed template to generate a single sentence to summarize the work order. Traditional summarization algorithms can use TextRank, Lead-3, MMR, etc., to extract keywords or sentences from the "fault description + handling process," then assemble them into a 30-character summary. Fine-tuned generative models can use models such as BART, T5, GPT-2-Chinese to fine-tune the work order, inputting the full text and outputting a 30-character summary in plain language.
[0025] The historical work orders or work orders to be checked obtained by the work order plagiarism detection system may contain a lot of text content. This large amount of text content can cause interference during the plagiarism detection process. Therefore, during the plagiarism detection, it is necessary to summarize each historical work order or work order to be checked, and reduce the amount of text content in the historical work orders or work orders to be checked while retaining the key content. This reduces the impact of irrelevant content in the historical work orders or work orders to be checked on the plagiarism detection, thereby improving the accuracy of the plagiarism detection.
[0026] Step 106: Determine the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders.
[0027] After identifying the first summary work order corresponding to each historical work order and the second summary work order corresponding to the work order to be checked for duplicates, the work order deduplication system determines the first feature vector corresponding to each first summary work order and the second feature vector corresponding to each second summary work order. In other words, the work order summary system determines the first feature vector of each first summary work order and the second feature vector of each second summary work order.
[0028] When determining the first feature vector of the first summary work order or the second feature vector of the second summary work order, the work order plagiarism detection system can use models such as Word2Vec-SG / CBOW, GloVe, FastText, and BERT WordPiece to determine the feature vector of each first summary work order or second summary work order.
[0029] Therefore, after acquiring multiple historical work orders, the work order deduplication system summarizes each historical work order to obtain a first summary work order. Then, it determines the feature vector of each first summary work order, thereby determining multiple first feature vectors. Similarly, after acquiring the work order to be deduplicated, the system summarizes it to obtain a second summary work order. Then, it determines the feature vector of the second summary work order, thereby determining the second feature vector.
[0030] Specifically, the work order deduplication system can pre-determine and store the feature vectors of each historical work order (or first summary work order). This allows it to directly determine and process the feature vector corresponding to the work order to be deduplicated (or the second summary work order) after receiving it. In other words, the system first acquires multiple historical work orders, then determines the first summary work order corresponding to each historical work order, and then determines the first feature vector corresponding to each first summary work order. Finally, it inputs the determined first feature vectors into the database for storage. Thus, when a work order to be deduplicated is received from a user, the system determines the second summary work order, then determines the second feature vector corresponding to the second summary work order, and finally performs the work order deduplication operation directly based on the determined second feature vector and the multiple first feature vectors pre-stored in the database.
[0031] Step 108: Determine the first similarity between each of the first feature vectors and the second feature vectors, and determine whether the work order to be checked is a duplicate work order based on the first similarity.
[0032] After determining the first feature vector corresponding to each first summary work order and the second feature vector corresponding to each second summary work order, the work order deduplication system determines the similarity between each first feature vector and the second feature vector, and defines the determined similarity as the first similarity. This first similarity can be used to characterize the similarity between each first summary work order and the second summary work order, and can also be used to characterize the similarity between each historical work order and the work order to be deduplicated.
[0033] Specifically, when determining the first similarity between the first feature vector and the second feature vector, the work order plagiarism detection system can use the cosine similarity algorithm (formula: \(sim(v_1,v_2)=\frac{v_1\cdotv_2}{||v_1||\cdot||v_2||}\)) to calculate the first similarity between the second feature vector and each of the first feature vectors (range: 0-1, the closer the value is to 1, the more similar the semantics).
[0034] After determining the first similarity between each first feature vector and the second feature vector, the work order deduplication system determines whether the work order to be checked is a duplicate work order based on the first similarity corresponding to each first feature vector.
[0035] Specifically, the work order deduplication system determines whether there is a first feature vector with a first similarity greater than a first threshold. If there is, the work order to be deduplicated is determined to be a duplicate work order; otherwise, the work order to be deduplicated is determined to be a non-duplicate work order. The first threshold is a pre-set value.
[0036] After determining that the work order to be checked is a duplicate work order, the work order deduplication system can further identify target work orders and send them to the operators. Target work orders are those that duplicate the work order to be checked in the historical work order database. The number of target work orders can be one or more. The specific steps for the work order deduplication system to identify target work orders are as follows: First, a first feature vector with a first similarity greater than a first threshold is identified as the target feature vector; the target summary work order corresponding to the target feature vector is identified; and then the target work order corresponding to the target summary work order is identified. The target summary work order is the first summary work order corresponding to the target feature vector. Further, the number of first feature vectors with a first similarity greater than the first threshold may be one or more. When the number is one, this first feature vector is identified as the target feature vector. When the number is multiple, the work order deduplication system can identify the first feature vector with the highest first similarity as the target feature vector, or it can identify multiple first feature vectors with a first similarity greater than the first threshold as target feature vectors and identify the target work orders corresponding to each target feature vector.
[0037] In other words, if the work order deduplication system determines that there are at least one first feature vector in the search results with a similarity ≥ the first threshold to the second feature vector, then the work order to be checked is determined to be a "duplicate work order," and all historical work order information that meets the conditions is output (including work order ID, standardized summary, similarity value, and submission time). If the similarity of all first feature vectors in the search results is < the preset threshold, then the work order to be checked is determined to be a "non-duplicate work order."
[0038] Furthermore, after identifying the target summary work order, the work order deduplication system can directly send the target summary work order to the operator's terminal device. This allows the operator to quickly and accurately understand work orders in the historical work orders that are similar to the work order to be checked for deduplication. It can also send the target work order to the operator's terminal device, allowing the operator to have a complete understanding of the historical work orders that are similar to the work order to be checked for deduplication. In other words, the work order deduplication system can send historical work orders that overlap with the work order to the operator's terminal device. Since the work orders sent to the operator's terminal device are complete historical work orders, the operator can have a complete understanding of the work orders. The work order deduplication system can also send summary work orders of historical work orders that overlap with the work order to the operator's terminal device. Again, since the work orders sent to the operator's terminal device are summary historical work orders, the operator can quickly and accurately understand the work orders. Furthermore, the work order deduplication system can send both the target summary work order and the target work order to the operator's terminal device, so that the operator can not only quickly and accurately understand the work order, but also have a complete understanding of the work order content.
[0039] The work order deduplication method provided in this embodiment of the invention obtains multiple historical work orders and work orders to be deduplicated. The historical work orders are historically generated work order data, and the work orders to be deduplicated are work order data that need to be deduplicated. The method summarizes each historical work order to obtain multiple first summary work orders and summarizes the work orders to be deduplicated to obtain second summary work orders. It determines a first feature vector corresponding to each first summary work order and a second feature vector corresponding to each second summary work order. It determines a first similarity between each first feature vector and the second feature vector, and determines whether the work order to be deduplicated is a duplicate work order based on the first similarity. This method enables rapid work order deduplication.
[0040] In one implementation, determining the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders (step 106) can be performed via steps A1-A2: Step A1: Input the first summary work order into the pre-trained semantic vector model to obtain the first feature vector.
[0041] The first feature vector is a floating-point vector related to the first summary work order.
[0042] When determining the first feature vector, the work order deduplication system inputs each first summary work order into the pre-trained semantic vector model, thereby generating the first feature vector corresponding to each first summary work order. This first feature vector is a floating-point vector, meaning that the feature vector determined by the work order deduplication system is the floating-point vector of each first summary work order.
[0043] Specifically, the pre-trained semantic vector model must meet the following requirements: uniform output vector dimension (e.g., 768-dimensional, 1024-dimensional) and semantic mapping accuracy (cosine similarity of vectors with the same / similar semantics ≥ 0.8). Taking the first summary work order (string) as input, the vector model's API (or local inference interface) is called to output a fixed-dimensional floating-point vector (e.g., [0.231, 0.562, -0.123, ..., 0.891]), with vector element precision retained to three decimal places.
[0044] Step A2: Input the second summary work order into the pre-trained semantic vector model to obtain the second feature vector.
[0045] The second feature vector is a floating-point vector related to the second summary work order.
[0046] When determining the second feature vector, the work order deduplication system inputs the second summary work order into the pre-trained semantic vector model to obtain the second feature vector corresponding to the second summary work order. The second feature vector is a floating-point vector, that is, the feature vector determined by the work order deduplication system is the floating-point vector of the second summary work order.
[0047] Specifically, the method and model used to determine the second feature vector based on the second summary work order are the same as the method and model used to determine the first feature vector based on the first summary work order, and will not be repeated here.
[0048] In one implementation, summarizing each of the historical work orders (step 104) can be performed via steps B1-B2: Step B1: Obtain the work order summary elements.
[0049] The work order summary elements are necessary elements for constituting the first summary work order and the second summary work order.
[0050] When summarizing historical work orders, the work order deduplication system first obtains the work order summary elements. These elements are essential for constructing both the first and second summarized work orders. In other words, the work order summary elements are those elements extracted or summarized from the work orders to be summarized. For example, these elements may include the work order title, work order address, and work order content. The work orders to be summarized can be historical work orders or work orders subject to deduplication.
[0051] Specifically, when summarizing historical work orders using a preset model, the work order summary elements can be a Prompt template. This template instructs the preset model to extract the work order summary elements from the work orders to be summarized and to output them according to a preset format, clearly defining the output format and constraints of the preset model. For example, the constructed Prompt template can be: Based on the following work order elements, output a standardized event summary, strictly in the format of [Request Location + Event Description]: 1. Work Order Title: {Work Order Title}; 2. Work Order Address: {Work Order Address}; 3. Work Order Content: {Work Order Content}. Requirements: The request location must include the province / city / district / specific location (e.g., room, equipment location); the event description must extract the core issue (including issue type, scope of impact, urgency level), removing redundant information.
[0052] Step B2: Input the historical work orders and the work order summary elements into a preset large language model, so that the large language model can summarize the historical work orders based on the work order summary elements to obtain the first summary work order.
[0053] After determining the elements of the work order summary, the work order deduplication system inputs historical work orders and their summary elements into a pre-set large language model. This allows the large language model to summarize the historical work orders based on these elements, resulting in the first summarized work order. The pre-set large language model has preset temperature coefficients (to ensure output stability and avoid format deviations) and maximum output lengths (to meet the length requirements of standardized summaries). It is a large language model with semantic understanding capabilities, supporting long text input and providing structured output capabilities (generating results according to a specified format).
[0054] Specifically, after inputting historical work orders and work order summary elements (Prompt template) into the large language model, the large language model can generate a standard format work order event summary, namely the first summary work order, such as: kitchen + faucet leaking, floor damp, emergency repair needed for apartment 101, Unit 2, Building 5, Xingfu Community, XX District, XX City, XX Province.
[0055] Furthermore, the method and model used to obtain the second summary work order by summarizing the work orders to be checked for duplicates are the same as the method and model used to obtain the first summary work order by summarizing the historical work orders, and will not be elaborated here.
[0056] Figure 2 This is a flowchart illustrating the second work order deduplication method provided in one embodiment of this specification, as shown below. Figure 2 As shown, the schematic diagram includes: Step 202: Obtain multiple historical work orders and obtain work orders to be checked for duplicates.
[0057] The historical work orders refer to historically generated work order data, and the work orders to be checked for duplicates refer to work order data that need to be checked for duplicates.
[0058] Step 204: Obtain the work order summary elements.
[0059] The work order summary elements are necessary elements for constituting the first summary work order and the second summary work order.
[0060] Step 206: Input the historical work orders and the work order summary elements into a preset large language model, so that the large language model can summarize the historical work orders based on the work order summary elements to obtain the first summary work order.
[0061] Step 208: Input the work order to be checked for duplicates and the work order summary elements into the preset large language model, so that the large language model summarizes the work order to be checked for duplicates based on the work order summary elements, and obtains a second summary work order.
[0062] Step 210: Input the first summary work order into the pre-trained semantic vector model to obtain the first feature vector.
[0063] The first feature vector is a floating-point vector related to the first summary work order.
[0064] Step 212: Input the second summary work order into the pre-trained semantic vector model to obtain the second feature vector.
[0065] The second feature vector is a floating-point vector related to the second summary work order.
[0066] Step 214: Determine the first similarity between each of the first feature vectors and the second feature vectors, and determine whether the work order to be checked is a duplicate work order based on the first similarity.
[0067] In the embodiments described in the specification, by inputting historical work orders and work orders to be checked for duplicates into a large language model, a first summary work order and a second summary work order are obtained. This eliminates the need for pre-training or fine-tuning the large language model, allowing direct utilization of the native capabilities of the general-purpose large language model to summarize historical work orders and work orders to be checked for duplicates. This not only improves the accuracy of work order deduplication but also reduces costs. By determining the similarity between the floating-point vectors of the first summary work order and the floating-point vectors of the second summary work order, a duplicate work order query is performed on the work orders to be checked for duplicates, thereby improving the efficiency of duplicate work order queries.
[0068] In one implementation, after determining the first feature vector corresponding to each of the first summary work orders (step 106), the method may further perform steps C1-C3: Step C1: Input the first summary work order into a preset first database for storage, and input the first feature vector into a preset second database for storage.
[0069] After identifying the first summary work order for each historical work order, the work order deduplication system can input each first summary work order into a preset first database for storage. After identifying the first feature vector corresponding to the first summary work order, the work order deduplication system can input each first feature vector into a preset second database for storage. In other words, the work order deduplication system stores the first work order data and the first feature vector in different databases.
[0070] When storing the first summary work order in the first database, the work order deduplication system stores the work order content and the corresponding code ID of each first summary work order. When storing the first feature vector in the second database, the work order deduplication system first constructs a vector data table structure with fields including: vector ID (associated with the unique ID of the work order, in the format "WORKORDER_XXX"), semantic vector (array type, storing floating-point vectors), work order event summary (redundant storage for subsequent result backtracking), and storage timestamp (UTC time format).
[0071] Specifically, the second database is a vector database that supports Approximate Nearest Neighbor (ANN) retrieval. When building the vector index: the work order plagiarism detection system creates an index based on IVF_FLAT (Inverted File Flat Index) or HNSW (Hierarchical Navigable Small World) in the vector database, configuring index parameters (such as the number of cluster centers (nlist), and the number of probes during retrieval (balancing retrieval speed and accuracy)). During retrieval, the work order plagiarism detection system calls the ANN retrieval interface of the vector database, setting "Return Top-K Similar Results" (ensuring coverage of potentially duplicate work orders), and simultaneously setting a "Similarity Filtering Threshold".
[0072] Furthermore, before inputting each first feature vector into the second database, the work order deduplication system can also normalize each first feature vector. Specifically, it can use the L2 normalization algorithm (formula: \(v_{norm} = \frac{v}{\sqrt{\sum_{i=1}^{n}v_i^2}}\)) to scale the vector to a unit vector, eliminating the influence of vector length on subsequent similarity calculation.
[0073] Step C2: Determine the first index table based on the correspondence between the first summary work order and the first feature vector.
[0074] After storing each first summary work order in the first database and each first feature vector in the second database, the work order deduplication system determines a first index table based on the correspondence between the first summary work orders and the first feature vectors. The first index table includes the correspondence between the coded ID of the first summary work order and the coded ID of the first feature vector.
[0075] The work order deduplication system uses the batch write interface of the vector database to write the structured vector data into the index table, ensuring that the vector ID and work order ID are mapped one-to-one. After writing, the index is synchronized (ensuring that subsequent searches can obtain data in real time).
[0076] In other words, the first index table includes the mapping relationship between each first feature vector ID and the first summary work order. In this way, after the work order deduplication system determines a certain feature vector in the second database as the target feature vector, it can determine the work order ID of the target feature vector based on the vector ID of the target feature vector and the first index table, and then determine the target summary work order that is associated with the target feature vector.
[0077] Step C3: Determine the third feature vector in the second database, and determine the third summary work order in the first database based on the third feature vector and the first index table.
[0078] Wherein, the third feature vector is the first feature vector whose similarity to the second feature vector is greater than a first threshold, and the third summary work order is the first summary work order corresponding to the third feature vector.
[0079] When performing work order deduplication, the work order deduplication system first determines the third feature vector in the second database, and then determines the third summary work order in the first database based on the mapping relationship between the third feature vector and the ID in the first index table. The third feature vector is the first feature vector in the first feature vector whose similarity to the second feature vector is greater than the first threshold, and the third summary work order is the first summary work order corresponding to the third feature vector.
[0080] In other words, when performing work order deduplication, the work order deduplication system determines the first similarity of each first feature vector and second feature vector, then determines the first feature vector with a first similarity greater than a first threshold as the third feature vector, then determines the vector ID corresponding to the third feature vector, then determines the work order ID corresponding to the vector ID according to the first index table, and finally determines the corresponding first summary work order according to the work order ID in the first database and determines it as the third summary work order, that is, the work order is a duplicate of the work order to be deduplicated.
[0081] In one implementation, after determining whether the work order to be checked is a duplicate work order based on the first similarity (step 108), the method may further perform steps D1-D2: Step D1: Input the second summary work order into the first database for storage, and input the second feature vector into the second database for storage.
[0082] After determining whether the work order to be checked is a duplicate work order, regardless of whether the work order to be checked is a duplicate work order, the work order deduplication system will input the second summary work order into the first database for storage, and input the second feature vector into the second database for storage.
[0083] In other words, after performing a plagiarism check on a work order, the work order plagiarism check system will store the second summary work order and the second feature vector of the work order to be checked into the first database and the second database respectively, so as to update the first database and the second database, so that the work order plagiarism check can be performed based on the updated first database and the second database when performing subsequent work order plagiarism checks.
[0084] Step D2: Update the first index table based on the correspondence between the second summary work order and the second feature vector.
[0085] After storing the second summary work order and the second feature vector into the first database and the second database respectively, the work order deduplication system will also update the first index table based on the correspondence between the second summary work order and the second feature vector, so that the work order deduplication operation can be performed based on the updated first index table in subsequent work order deduplication operations.
[0086] Furthermore, after identifying the work order to be checked for duplicates and the target work order, the work order deduplication system can also store the correspondence between the work order to be checked for duplicates and the target work order. This correspondence includes whether the work order to be checked for duplicates and the target work order are duplicate work orders, and the first similarity value between them. The target work order is a historical work order that is confirmed to be duplicated by the work order to be checked for duplicates. Thus, when performing subsequent work order deduplication operations and determining that the work order to be checked for duplicates is a duplicate of the third work order, the target work order can also be determined as a duplicate of the third work order. The work order to be checked for duplicates is the work order data to be checked during the first deduplication operation, and the third work order is the work order data to be checked during the second deduplication operation.
[0087] In other words, the work order deduplication system can store the duplicate relationships existing in historical work orders. This means that when performing duplicate work order deduplication and identifying a target work order as a duplicate historical work order, it can also identify historical work orders with stored duplicate relationships as duplicate work orders, thus increasing the accuracy of identifying duplicate historical work orders. The duplicate relationships in historical work orders can be determined by determining the similarity between any two historical work orders. For example, two historical work orders can be randomly selected, and their similarity can be determined. If the similarity is greater than a second threshold, then a duplicate relationship exists between the two historical work orders; otherwise, no duplicate relationship exists.
[0088] Figure 3 This is a flowchart illustrating the third work order deduplication method provided in one embodiment of this specification, as shown below. Figure 3 As shown, the schematic diagram includes: Step 302: Obtain multiple historical work orders.
[0089] The historical work orders refer to historically generated work order data.
[0090] Step 304: Summarize each of the historical work orders to obtain multiple first summary work orders.
[0091] Step 306: Determine the first feature vector corresponding to each of the first summary work orders.
[0092] Step 308: Input the first summary work order into a preset first database for storage, and input the first feature vector into a preset second database for storage.
[0093] Step 310: Determine the first index table based on the correspondence between the first summary work order and the first feature vector.
[0094] Step 312: Obtain the duplicate work order to be checked.
[0095] The work order to be checked for duplicates refers to the work order data that needs to be checked for duplicates.
[0096] Step 314: Summarize the duplicate work orders to be checked to obtain the second summary work order.
[0097] Step 316: Determine the second feature vector corresponding to the second summary work order.
[0098] Step 318: Determine the third feature vector in the second database, and determine the third summary work order in the first database based on the third feature vector and the first index table.
[0099] Wherein, the third feature vector is the first feature vector whose similarity to the second feature vector is greater than a first threshold, and the third summary work order is the first summary work order corresponding to the third feature vector.
[0100] Step 320: Input the second summary work order into the first database for storage, and input the second feature vector into the second database for storage.
[0101] Step 322: Update the first index table based on the correspondence between the second summary work order and the second feature vector.
[0102] In the embodiments described in the specification, by pre-storing the feature vectors of each historical work order, the feature vector of the work order to be checked can be directly determined and the check operation can be performed when the work order to be checked is obtained, saving the check time and improving the check efficiency. By storing the second summary work order and the second feature vector into the database after the check, the data in the database can be improved through continuous check, thus improving the accuracy of the check.
[0103] In one implementation, obtaining the duplicate work order to be checked (step 102) can be performed via steps E1-E2: Step E1: Obtain the session data to be checked for plagiarism.
[0104] The deduplication session data to be checked is the data generated from conversations between users and customer service.
[0105] When obtaining work orders to be checked for duplicates, the work order deduplication system can obtain the work orders to be checked for duplicates in two ways. One is to directly obtain the work order and identify it as a work order to be checked for duplicates. The second is to obtain the dialogue content between the agent and the caller and identify the work order to be checked for duplicates based on the dialogue content.
[0106] Specifically, when determining the work order to be checked for plagiarism through the second method, the work order plagiarism checking system first obtains the conversation data to be checked. The conversation data to be checked is the dialogue data between the agent and the caller, that is, the data generated by the user and the customer service representative during the conversation. The conversation content in the conversation data to be checked can clearly distinguish the roles of the customer service representative and the user. For example, the conversation data to be checked includes voice data sent by the user (or the caller) and voice data sent by the agent (or customer service representative). The voice data sent by the user (or the caller) can be voice data such as "My home is in Building 5 of Xingfu Community, the kitchen faucet is leaking, and the floor is wet".
[0107] When acquiring the data of the session to be checked for plagiarism, the work order plagiarism checking system collects all the data in the session. For example, when the session is a real-time session, the work order plagiarism checking system collects all messages sent by the user through the session interface (sorted by timestamp, retaining the original dialogue order); when the session is a historical session, the work order plagiarism checking system reads the complete session record (including user ID, session ID, message content, and sending time) from the session database.
[0108] Step E2: Determine the deduplication work order based on the deduplication session data to be checked.
[0109] After determining the session data to be checked for plagiarism, the work order plagiarism checking system parses the session data to obtain the work order to be checked for plagiarism.
[0110] The work order plagiarism detection system can identify the work orders to be checked for plagiarism through a preset large language model. For example, the work order plagiarism detection system sends the session data to be checked to the large language model so that the large language model can parse the session data and identify the work orders to be checked for plagiarism.
[0111] Specifically, the work order plagiarism detection system can send preset prompt templates and the session data to be checked to the large language model, enabling the large language model to analyze the address information, request content information, etc. included in the session. The work orders to be checked output by the large language model strictly follow the format of "request location + event description", maintaining consistency with the format of historical work orders, and generate the work order to be checked. For example, when the session text of the session data to be checked is "My home is in Building 5 of Xingfu Community, the kitchen faucet is leaking, and the floor is wet", the summary (work order to be checked) output by the large language model can be "Building 5 of Xingfu Community, XX District, XX City, XX Province, kitchen + faucet leaking, floor damp, repair consultation".
[0112] Furthermore, the work order plagiarism detection system can input the session data to be checked into the large language model, so that the large language model can directly generate a summary work order, and then perform plagiarism detection based on the summary work order, saving operation steps and improving plagiarism detection efficiency.
[0113] It should be noted that the work order deduplication method provided in this application embodiment can be executed by a work order deduplication device or a control module within that work order deduplication device for executing the work order deduplication method. This application embodiment uses the execution of the work order deduplication method by a work order deduplication device as an example to illustrate the work order deduplication device provided in this application embodiment.
[0114] Figure 4 This is a schematic diagram of the work order deduplication device according to an embodiment of the present invention. Figure 4 As shown, the work order deduplication device includes: a first acquisition module 402, a first summary module 404, a first determination module 406, and a second determination module 408.
[0115] The first acquisition module 402 is used to acquire multiple historical work orders and acquire work orders to be checked for duplicates. The historical work orders are historically generated work order data, and the work orders to be checked for duplicates are work order data that need to be checked for duplicates. The first summary module 404 is used to summarize each of the historical work orders to obtain multiple first summary work orders, and to summarize the work orders to be checked for duplicates to obtain second summary work orders. The first determining module 406 is used to determine the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders; The second determining module 408 is used to determine the first similarity between each of the first feature vectors and the second feature vectors, and to determine whether the work order to be checked is a duplicate work order based on the first similarity.
[0116] The work order deduplication device in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network attached storage (NAS), personal computers (PCs), televisions (TVs), ATMs, or self-service machines, etc. This application embodiment does not impose specific limitations.
[0117] The work order deduplication device in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.
[0118] The virtual device provided in this application embodiment can achieve Figures 1 to 3 The various processes implemented in the method embodiments are not described in detail here to avoid repetition.
[0119] Based on the same technical concept, embodiments of this application also provide an electronic device for performing the above-described work order deduplication method. Figure 5This is a schematic diagram of the structure of an electronic device to implement various embodiments of this application. The electronic device can vary significantly due to differences in configuration or performance, and may include a processor 502, a communications interface 504, a memory 506, and a communication bus 508. The processor 502, communications interface 504, and memory 506 communicate with each other via the communication bus 508. The processor 502 can call a computer program stored in the memory 506 and executable on the processor 502 to perform the following steps: The system retrieves multiple historical work orders and retrieves work orders to be checked for duplicates. The historical work orders are historically generated work order data, and the work orders to be checked for duplicates are work order data that need to be checked for duplicates. Summarize each of the historical work orders to obtain multiple first summary work orders, and summarize the work orders to be checked for duplicates to obtain second summary work orders; Determine the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders; Determine the first similarity between each of the first feature vectors and the second feature vectors, and determine whether the work order to be checked is a duplicate work order based on the first similarity.
[0120] In one implementation, determining the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders includes: The first summary work order is input into a pre-trained semantic vector model to obtain the first feature vector, which is a floating-point vector related to the first summary work order. The second summary work order is input into the pre-trained semantic vector model to obtain the second feature vector, which is a floating-point vector related to the second summary work order.
[0121] In one implementation, summarizing each of the historical work orders includes: Obtain work order summary elements, which are necessary elements for constituting the first summary work order and the second summary work order; The historical work orders and the work order summary elements are input into a preset large language model, so that the large language model summarizes the historical work orders based on the work order summary elements to obtain the first summary work order.
[0122] In one implementation, after determining the first feature vector corresponding to each of the first summary work orders, the method further includes: The first summary work order is input into a preset first database for storage, and the first feature vector is input into a preset second database for storage; The first index table is determined based on the correspondence between the first summary work order and the first feature vector; A third feature vector is determined in the second database, and a third summary work order is determined in the first database based on the third feature vector and the first index table. The third feature vector is the first feature vector whose similarity to the second feature vector is greater than a first threshold, and the third summary work order is the first summary work order corresponding to the third feature vector.
[0123] In one implementation, after determining whether the work order to be checked is a duplicate work order based on the first similarity, the method further includes: The second summary work order is input into the first database for storage, and the second feature vector is input into the second database for storage; The first index table is updated based on the correspondence between the second summary work order and the second feature vector.
[0124] In one implementation, obtaining the duplicate work order to be checked includes: Obtain the deduplication session data to be checked, which is the data generated from the conversation between the user and customer service; The deduplication work order is determined based on the deduplication session data.
[0125] The specific execution steps can be found in the various steps of the above-mentioned work order deduplication method embodiment, and can achieve the same technical effect. To avoid repetition, they will not be repeated here.
[0126] It should be noted that the electronic devices in the embodiments of this application include: servers, terminals, or other devices besides terminals.
[0127] The above electronic device structure does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or arrange them differently. For example, an input unit may include a Graphics Processing Unit (GPU) and a microphone, and a display unit may use a liquid crystal display (LCD), organic light-emitting diode (OLED), or other similar display panels. User input units include at least one of a touch panel and other input devices. A touch panel is also called a touchscreen. Other input devices may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be elaborated further here.
[0128] Memory can be used to store software programs and various data. Memory can primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area can store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, memory can include volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).
[0129] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly handles operations related to the operating system, user interface, and applications, while the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into the processor.
[0130] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described work order deduplication method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.
[0131] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.
[0132] This application also provides a computer program product. When the computer program product is executed by a processor, it implements the various processes of the above-described work order deduplication method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0133] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described work order deduplication method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.
[0134] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
[0135] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0137] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A method for checking duplicate work orders, characterized in that, include: The system retrieves multiple historical work orders and retrieves work orders to be checked for duplicates. The historical work orders are historically generated work order data, and the work orders to be checked for duplicates are work order data that need to be checked for duplicates. Summarize each of the historical work orders to obtain multiple first summary work orders, and summarize the work orders to be checked for duplicates to obtain second summary work orders; Determine the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders; Determine the first similarity between each of the first feature vectors and the second feature vectors, and determine whether the work order to be checked is a duplicate work order based on the first similarity.
2. The method according to claim 1, characterized in that, Determining the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders includes: The first summary work order is input into a pre-trained semantic vector model to obtain the first feature vector, which is a floating-point vector related to the first summary work order. The second summary work order is input into the pre-trained semantic vector model to obtain the second feature vector, which is a floating-point vector related to the second summary work order.
3. The method according to claim 1, characterized in that, The summary of each of the aforementioned historical work orders includes: Obtain work order summary elements, which are necessary elements for constituting the first summary work order and the second summary work order; The historical work orders and the work order summary elements are input into a preset large language model, so that the large language model summarizes the historical work orders based on the work order summary elements to obtain the first summary work order.
4. The method according to claim 1, characterized in that, After determining the first feature vector corresponding to each of the first summary work orders, the method further includes: The first summary work order is input into a preset first database for storage, and the first feature vector is input into a preset second database for storage; The first index table is determined based on the correspondence between the first summary work order and the first feature vector; A third feature vector is determined in the second database, and a third summary work order is determined in the first database based on the third feature vector and the first index table. The third feature vector is the first feature vector whose similarity to the second feature vector is greater than a first threshold, and the third summary work order is the first summary work order corresponding to the third feature vector.
5. The method according to claim 4, characterized in that, After determining whether the work order to be checked is a duplicate work order based on the first similarity, the method further includes: The second summary work order is input into the first database for storage, and the second feature vector is input into the second database for storage; The first index table is updated based on the correspondence between the second summary work order and the second feature vector.
6. The method according to claim 1, characterized in that, The process of obtaining the duplicate work order to be checked includes: Obtain the deduplication session data to be checked, which is the data generated from the conversation between the user and customer service; The deduplication work order is determined based on the deduplication session data.
7. A work order duplicate checking device, characterized in that, include: The first acquisition module is used to acquire multiple historical work orders and acquire work orders to be checked for duplicates. The historical work orders are historically generated work order data, and the work orders to be checked for duplicates are work order data that need to be checked for duplicates. The first summary module is used to summarize each of the historical work orders to obtain multiple first summary work orders, and to summarize the work orders to be checked for duplicates to obtain second summary work orders. The first determining module is used to determine the first feature vector corresponding to each of the first summary work orders and the second feature vector corresponding to each of the second summary work orders; The second determining module is used to determine the first similarity between each of the first feature vectors and the second feature vectors, and to determine whether the work order to be checked is a duplicate work order based on the first similarity.
8. An electronic device, characterized in that, The device includes: Processor; and A memory configured to store computer-executable instructions configured to be executed by the processor, the executable instructions including steps for performing the work order deduplication method as described in any one of claims 1 to 6.
9. A storage medium, characterized in that, The storage medium is used to store computer-executable instructions that cause the computer to perform the work order deduplication method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the work order deduplication method according to any one of claims 1 to 6.