A work order duplicate checking method, system, device and storage medium

Through the combination of Word2vector model and BERT-based sentiment analysis, the problem of low accuracy of existing work order plagiarism check technology is solved. Multi-dimensional similarity calculations are used to improve the accuracy of work order plagiarism check and reduce misjudgment.

CN118569612BActive Publication Date: 2024-12-06国泰新点软件股份有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411053509.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-12-06
Estimated Expiration
2044-08-02

AI Technical Summary

Technical Problem

The existing work order plagiarism checking technology has the problem of low accuracy. The calculation methods based on surface text and semantic similarity have limitations, making it difficult to accurately judge repeated work orders.

Method used

Word2vector model is used to calculate word vector transformation and Jaccard similarity, combined with BERT-based sentiment analysis, the weighted sum of three similarities is determined whether the work order is repeated, and the accuracy of plagiarism check is improved.

Benefits of technology

Through multi-dimensional similarity calculation, comprehensive inspection of work order repetition is improved, and the accuracy and accuracy of work order plagiarism checks are reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118569612B_ABST
    Figure CN118569612B_ABST
Patent Text Reader

Abstract

The present invention provides a method, system, device and storage medium for checking duplicate work orders. The method for checking duplicate work orders includes: obtaining query work order data and historical work order data and performing preprocessing to obtain a query work order word list and a historical work order word list; performing word vector conversion and summing on the query work order word list and the historical work order word list respectively through a pre-trained Word2vector model, obtaining a query work order vector and a historical work order vector and calculating a first similarity; calculating a second similarity through Jaccard based on the query work order word list and the historical work order word list; extracting the sentiment feature vectors of the query work order data and the historical work order data through BERT‑based and calculating a third similarity; determining whether the query work order data is repeated based on the first similarity, the second similarity and the third similarity. The present invention improves the accuracy of checking duplicate work orders by calculating the similarity of the query work order data and the historical work order data from three aspects: semantics, word co-occurrence times and sentiment, and checking duplicate work orders through the similarity.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a work order duplicate checking method, system, device and storage medium. Background Art

[0002] In the hotline business flow, the process from the operator receiving citizens' demands to form a work order to dispatching the work order to the corresponding department for processing is the dispatching link. In the dispatching link, the operator can check the work order for duplicates after forming the work order, and the department can check the work order for duplicates before processing the work order. The behavior of finding duplicate work orders and merging them for processing is called merging. Work order duplicate checking is a pre-merging technology, which is essentially a method of finding duplicate work orders based on the similarity of work order texts. At present, the similarity calculation methods on the market can be summarized into two categories. One is the similarity calculation based on surface text. First, the work order is segmented, and the segmentation results of different work orders are compared. The similarity between the current work order and the historical work order is calculated by cosine similarity and a similarity threshold is set to confirm the duplicate work order. When the work order is judged to be duplicated by this method, the similarity will increase when the words in the text are consistent, but the same word will express different meanings when used in different positions, which will be mistaken for duplicate work orders; the other is the similarity calculation based on semantics. First, a feature mining model is built to mine the elements in the duplicate work order to be checked and the target work order, and then the semantic analysis technology LSA is used to calculate the semantic similarity between the elements of the duplicate work order to be checked and the elements of the target work order to form a duplicate work order. When the work order is judged to be duplicated by this method, the similarity will increase when the semantics are consistent, but the same semantics will have multiple expressions of emotions. When the emotions are different, it will be mistaken for a duplicate work order. Both methods have limitations and the accuracy of checking duplicate work orders is low. Summary of the invention

[0003] In view of this, the embodiments of the present application are committed to providing a work order duplication checking method, system, device and storage medium, which are used to solve the limitations of current work order duplication checking technology and the problem of low work order duplication checking accuracy.

[0004] According to a first aspect of the present invention, the present invention provides a work order duplicate checking method, comprising:

[0005] Obtain query work order data and historical work order data and perform preprocessing to obtain query work word list and historical work word list;

[0006] The query work order word list and the historical work order word list are respectively converted into word vectors and summed up using a pre-trained Word2vector model to obtain a query work order vector and a historical work order vector and calculate a first similarity;

[0007] Calculate a second similarity by Jaccard based on the query word list and the historical word list;

[0008] Extracting sentiment feature vectors of the query work order data and the historical work order data through BERT-based and calculating a third similarity;

[0009] Whether the query work order data is repeated is determined based on the first similarity, the second similarity, and the third similarity.

[0010] Preferably, the obtaining of query work order data and historical work order data and preprocessing to obtain a query work word list and a historical work word list includes:

[0011] Utilize the historical work order data to perform word segmentation through Jieba and construct a related stop word list;

[0012] Performing word segmentation and stop word processing on the query work order data and the historical work order data through the jieba and the related stop word list;

[0013] The processed query work order data and the historical work order data are concatenated in terms of title, content and title to obtain a query work order word list and a historical work order word list.

[0014] Preferably, the training of the Word2vector model includes:

[0015] The training sample is used as the input of the Word2vector model, and the word vector is the output of the Word2vector model;

[0016] The training samples are constructed through training corpus, and each training sample includes a central word and context words.

[0017] Preferably, the pre-trained Word2vector model is used to convert the query work order word list and the historical work order word list into word vectors respectively and sum them up to obtain the query work order vector and the historical work order vector and calculate the first similarity, including:

[0018] Initialize the vector of the historical word list by using the Word2vector model;

[0019] Input the historical work order word list into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the historical work order vector;

[0020] Input the query work order word list into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the query work order vector;

[0021] The first similarity is calculated by using the historical work order vector and the query work order vector through Cosine.

[0022] Preferably, the calculating the second similarity based on the query word list and the historical word list by using Jaccard includes:

[0023] Converting the query word list and the history word list into an N-Gram sequence;

[0024] Counting the co-occurrence times of N-Gram sequences of the query word list and the historical word list;

[0025] The second similarity is calculated by the Jaccard method based on the co-occurrence count.

[0026] Preferably, the extracting the sentiment feature vectors of the query work order data and the historical work order data through BERT-based and calculating the third similarity includes:

[0027] Performing sentiment analysis on the query work order data and the historical work order data through the BERT-based method to obtain query sentiment labels and historical sentiment labels;

[0028] Converting the query sentiment label and the historical sentiment label into a query sentiment vector and a historical sentiment vector respectively;

[0029] Calculating a third similarity by using the query sentiment vector and the historical sentiment vector;

[0030] The emotion vector is shown in the following formula:

[0031] ;

[0032] Where: is the sentiment vector; is the positive sentiment vector; is the negative sentiment vector; is the neutral sentiment vector;

[0033] The third similarity is shown in the following formula:

[0034] ;

[0035] Where: To query the sentiment vector; is the historical sentiment vector.

[0036] Preferably, the determining whether the query work order data is repeated based on the first similarity, the second similarity and the third similarity further includes:

[0037] Setting weights for the first similarity, the second similarity, and the third similarity;

[0038] Summing the weighted first similarity, the second similarity and the third similarity to obtain a query work order similarity;

[0039] Sorting the query work order data according to the query work order similarity;

[0040] According to the sorting, the repeated query work order data is determined.

[0041] According to a third aspect of the present invention, a work order duplicate checking system is also provided, comprising:

[0042] The vocabulary acquisition module is used to acquire the query work order data and the historical work order data and perform preprocessing to obtain the query work order vocabulary list and the historical work order vocabulary list;

[0043] A first similarity module is used to convert the query work order word list and the historical work order word list into word vectors and sum them up respectively through a pre-trained Word2vector model, obtain the query work order vector and the historical work order vector and calculate the first similarity;

[0044] A second similarity module, used for calculating a second similarity between the query word list and the historical word list through Jaccard;

[0045] A third similarity module, configured to extract sentiment feature vectors of the query work order data and the historical work order data through BERT-based extraction and calculate a third similarity;

[0046] The query module is used to determine whether the query work order data is repeated according to the first similarity, the second similarity and the third similarity.

[0047] According to a fourth aspect of the present invention, there is also provided an electronic device comprising a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the work order duplication checking method.

[0048] According to a fifth aspect of the present invention, a computer-readable storage medium is also provided, wherein the computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the work order duplication checking method.

[0049] According to the first aspect of the present invention, query work order data and historical work order data are obtained, and the first similarity is calculated by converting them into vectors and summing them up to query whether the semantics between the work orders are similar; the second similarity is calculated by converting them into N-Gram sequences, and the number of co-occurrences of phrases between the work orders are queried; the sentiment feature vector is obtained through BERT-based to calculate the third similarity, and whether the sentiment expressions between the work orders are the same. Even if the words in the query work word list and the historical work word list are consistent, the similarity calculated by the sentiment vector can also check that there are no duplicate work orders. Conversely, even if there are no identical words in the two work word lists, but the semantics are the same and the emotions expressed are the same, it can also be confirmed as a duplicate work order. Based on the calculation of the three similarities, it is possible to comprehensively check whether the query work order data is repeated, and improve the accuracy of work order duplicate checking.

[0050] Furthermore, since the first similarity, the second similarity and the third similarity represent different meanings respectively, the weight of each similarity is different. Each similarity is weighted according to the importance of the three similarities, and the three weighted similarities are summed up, thereby improving the accuracy of duplicate checking for work orders. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic flow chart of a work order duplicate checking method according to an embodiment of the present invention is shown;

[0052] Figure 2 Shows Figure 1 A schematic flow chart of a method for obtaining a query word list and a historical word list in step S100 is shown;

[0053] Figure 3 Shows Figure 1 A schematic flow chart of a method for calculating the first similarity in step S200 is shown;

[0054] Figure 4 Shows Figure 1 A schematic flow chart of a method for calculating the second similarity in step S300 is shown;

[0055] Figure 5 Shows Figure 1 A schematic flow chart of a method for calculating the third similarity in step S400 is shown;

[0056] Figure 6 Shows Figure 1 A schematic flow chart of the method for querying work order sorting in step S500 is shown;

[0057] Figure 7 A schematic diagram of the application of a work order duplicate checking method according to an embodiment of the present invention is shown;

[0058] Figure 8 A schematic diagram of a work order duplicate checking system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some structures related to the present application are shown in the accompanying drawings, rather than all structures. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0060] The terms "including" and "having" and any variations thereof in this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0061] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0062] Figure 1 A schematic flow chart of a work order duplicate checking method according to an embodiment of the present invention is shown. Figure 1 As shown, a work order duplicate checking method includes:

[0063] Step S100: Acquire query work order data and historical work order data and perform preprocessing to obtain a query work word list and a historical work word list;

[0064] Step S200: using a pre-trained Word2vector model, respectively convert the query work word list and the history work word list into word vectors and sum them up, obtain the query work order vector and the history work order vector, and calculate the first similarity;

[0065] Step S300: Calculating a second similarity by Jaccard distance based on the query word list and the historical word list;

[0066] Step S400: extracting sentiment feature vectors of query work order data and historical work order data through BERT-based (sentiment analysis model) and calculating the third similarity;

[0067] Step S500: Determine whether the query work order data is repeated through the first similarity, the second similarity and the third similarity.

[0068] Obtain query work order data and historical work order data, convert them into vectors and calculate the first similarity by summing them up, to query whether the semantics between work orders are similar; calculate the second similarity by converting them into N-Gram sequences (N-gram models), to query the number of co-occurrences of phrases between work orders; obtain the sentiment feature vector through BERT-based to calculate the third similarity, to query whether the sentiment expressions between work orders are the same, and then obtain whether the query work order data is a duplicate work order based on the relationship between the first similarity, the second similarity and the third similarity. Based on the calculation of the three similarities, a comprehensive check can be made to check whether the query work order data is duplicated, and the accuracy of duplicate work order checking can be improved.

[0069] Figure 2 Shows Figure 1 The schematic flow chart of the method for obtaining the query word list and the historical word list in step S100 is shown as follows: Figure 2 As shown, in step S100, the query work order data and the historical work order data are obtained and preprocessed to obtain the query work word list and the historical work word list, including:

[0070] Step S110: Use the historical work order data to perform word segmentation through jieba (jieba) and construct a related stop word list;

[0071] Step S120: perform word segmentation and stop word processing on the concatenated query work order data and historical work order data through Jieba and related stop word lists;

[0072] Step S130: concatenate the title, content and title of the processed query work order data and historical work order data to obtain a query work word list and a historical work word list.

[0073] Through Jieba, historical work order data is segmented to select words that have little impact on expressing the meaning of the work order, and a related stop word list is constructed based on these words. The related stop word list is then used to filter the words in subsequent query work order data and historical work order data, and the words in the related stop word list are removed. Then, the filtered query work order data and historical work order data are spliced ​​through the title, content, and title format. Since the title is a summary of the content, the weight of the title is set higher.

[0074] In step S200, the training of the Word2vector model includes:

[0075] The training sample is used as the input of the Word2vector model, and the word vector is the output of the Word2vector model;

[0076] The training samples are constructed through training corpus, and each training sample includes a central word and context words.

[0077] Set up training samples, including central words and context words, to prepare for the subsequent calculation of semantic similarity. Therefore, it is necessary to train the model in combination with central words and context words to obtain more accurate semantics.

[0078] Figure 3 Shows Figure 1 The schematic flow chart of the method for calculating the first similarity in step S200 is shown as follows. Figure 3 As shown, in step S200, the query work word list and the history work word list are respectively transformed into word vectors and summed up through the pre-trained Word2vector model to obtain the query work order vector and the history work order vector and calculate the first similarity, including;

[0079] Step S210: Initialize the vector of the historical word list through the Word2vector model;

[0080] Step S220: Input the historical work order word list into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the historical work order vector;

[0081] Step S230: Input the query work word list into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the query work order vector;

[0082] Step S240: Calculate the first similarity by using Cosine using the historical work order vector and the query work order vector.

[0083] The model initializes the vector of the historical word list and randomly initializes a word vector for each word. The historical word list and the query word list are then input into the model to obtain the vector of each word in each word list, and the vector of each word list is calculated by adding the vectors of each word in each word list to calculate the semantic similarity. When a word in the query word list does not obtain a vector, it is initialized to a vector of 1.

[0084] Figure 4 Shows Figure 1 The schematic flow chart of the method for calculating the second similarity in step S300 is shown as follows. Figure 4 As shown, in step S300, the second similarity is calculated by Jaccard based on the query word list and the historical word list, including:

[0085] Step S310: converting the query word list and the history word list into an N-Gram sequence;

[0086] Step S320: Count the co-occurrence times of N-Gram sequences in the query word list and the historical word list;

[0087] Step S330: Calculate the second similarity by Jaccard based on the number of co-occurrences.

[0088] The query word list and the historical word list are restored into strings, the two words are combined into a sequence through the N-Gram sequence, and the co-occurrence times of the N-Gram sequences in the two word lists are counted to obtain the similar contents in the two word lists. Finally, the second similarity is calculated by Jaccard.

[0089] Figure 5 Shows Figure 1 The schematic flow chart of the method for calculating the third similarity in step S400 is shown as follows. Figure 5 As shown, in step S400, the sentiment feature vectors of the query work order data and the historical work order data are extracted through BERT-based and the third similarity is calculated, including:

[0090] Step S410: Perform sentiment analysis on the query work order data and the historical work order data through BERT-based to obtain query sentiment labels and historical sentiment labels;

[0091] Step S420: converting the query emotion label and the historical emotion label into a query emotion vector and a historical emotion vector respectively;

[0092] Step S430: calculating a third similarity by querying the sentiment vector and the historical sentiment vector;

[0093] The sentiment vector is as follows:

[0094] ;

[0095] Where: is the sentiment vector; is the positive sentiment vector; is the negative sentiment vector; is the neutral sentiment vector;

[0096] The third similarity is shown in the following formula:

[0097] ;

[0098] Where: To query the sentiment vector; is the historical sentiment vector.

[0099] In the expression of work orders, sometimes the words are the same but the meanings to be expressed are different. Therefore, it is necessary to calculate the sentiment similarity of the query work order data and the historical work order data, divide the sentiment into positive, negative and neutral, perform sentiment analysis on the words in the query work word list and the historical work word list, obtain the sentiment label of each word in the query work word list and the historical work word list respectively, convert the label into a vector, and calculate the third similarity. Even if the words in the query work word list and the historical work word list are the same, the similarity calculated by the sentiment vector can also check that there is no duplication of the work order. On the contrary, even if there are no identical words in the two work word lists, but the semantics are the same and the sentiments expressed are the same, it can also be confirmed as a duplicate work order.

[0100] Figure 6 Shows Figure 1 The schematic flow chart of the method for querying the work order sorting in step S500 is shown as follows. Figure 6 As shown, in step S500, based on the first similarity, the second similarity and the third similarity, determining whether the query work order data is repeated also includes:

[0101] Step S510: setting weights for the first similarity, the second similarity and the third similarity;

[0102] Step S520: summing the weighted first similarity, second similarity and third similarity to obtain the query work order similarity;

[0103] Step S530: sorting the query work order data according to the query work order similarity;

[0104] Step S540: Determine the duplicate query work order data according to the sorting.

[0105] Three similarities are calculated through three aspects. However, since the importance of similarities is not constant, a matching weight is added to each similarity. Each similarity is weighted and then summed up. The query work orders are sorted according to their similarities, thereby selecting duplicate query work order data.

[0106] Figure 7 FIG. 4 shows a schematic diagram of an application of a work order duplicate checking method according to an embodiment of the present invention. Figure 7 As shown, work order duplication checking can be applied to hotline seats and processing units, and the processing procedures are roughly similar.

[0107] Hotline agents receive complaints. After receiving the complaints, they check for duplicates. If there are no duplicate complaints, they will be dispatched regularly. If the complaints are duplicate work orders, they will be combined. Duplicate work orders will be suspended and then handled, and finally returned for handling. After the work order is suspended, it will be combined with similar historical work orders for regular dispatch or cancellation of duplicate work orders.

[0108] The handling unit includes signing for the work orders sent by the hotline agents, checking for duplicates, and performing regular processing when there are no duplicate tasks. When it is found to be a duplicate task, the task is merged and suspended, and then task feedback is attached, and finally task feedback is given. After the task is suspended, the merge can also be canceled.

[0109] In one embodiment of the present invention, the words that have little influence on the meaning of the work order are sorted out manually based on the historical work order data and then the stop word list is formed to reduce the interference caused by subsequent similar calculations. Example words include related, department, hope, appeal, help, platform, call, artificial, request, netizen, requirement, information, confidentiality, solution, thank you, trouble, and problem.

[0110] Input historical work order data, segment the historical work order data through jieba, load the stop word library to remove stop words, and then train the Word2vector model. The training process of the Word2vector model includes: first build a vocabulary, that is, the vocabulary obtained after processing the historical work order data. Then initialize the word vector, and randomly initialize a word vector for each word in the vocabulary. Then build training samples, build training samples from the training corpus, and each sample contains a central word and its context words. Then train the neural network, use the training samples to train the neural network model, and optimize the word vector. Finally, extract the word vector. After the training is completed, output the word vector corresponding to each word as the final result. The Word2vector core is selected because it can learn the semantic relationship between words, generate high-quality word vectors, and is suitable for large-scale corpora.

[0111] Input two work order data to be compared, namely historical work order data and query work order data, and first perform the splicing operation of title + content + title. The title is spliced ​​twice because the title has a high degree of summary and generalization in the work order data and can be given a higher weight. Then perform Jieba word segmentation and stop word processing respectively, and finally output two processed work word lists.

[0112] Load the trained Word2vector model, convert the two processed work word lists into word vectors, and obtain the vector of each word. If the vector of the word cannot be obtained, initialize it to a vector of the same dimension of 1, and sum the vectors of each word in each work word list to obtain the work order vector. Finally, the semantic similarity of the two work orders is calculated by Cosine (cosine similarity) between the two work order vectors.

[0113] Restore the word list to a string, convert the word list into an N-Gram sequence, where n is 2, that is, two consecutive words form a sequence. Count the co-occurrence times of the N-Gram sequences in the two texts, use the Jaccard distance to calculate the similarity of the co-occurrence data, and finally output the similarity of the word co-occurrence times of the two work orders.

[0114] Through BERT-based sentiment analysis of query work order data and historical work order data, query sentiment labels and historical sentiment labels are obtained. In the expression of work orders, sometimes the same words appear but the meanings to be expressed are different. Therefore, it is necessary to calculate the sentiment similarity of query work order data and historical work order data, and divide the sentiment into positive, negative and neutral. The query sentiment label and historical sentiment label are converted into query sentiment vector and historical sentiment vector respectively, and the sentiment similarity is calculated through the query sentiment vector and historical sentiment vector;

[0115] The sentiment vector is as follows:

[0116] ;

[0117] Where: is the sentiment vector; is the positive sentiment vector; is the negative sentiment vector; is the neutral sentiment vector;

[0118] For example, when the word has a positive sentiment vector, the sentiment vector ;

[0119] The third similarity is shown in the following formula:

[0120] ;

[0121] Where: To query the sentiment vector; is the historical sentiment vector.

[0122] The final result is sorted in descending order based on the sum of the Cosine calculation results and the Jaccard calculation results and returns the top N (first N) ones. The output fields are the work order title, work order content, and similarity.

[0123] Figure 8 A schematic diagram of a work order duplicate checking system according to an embodiment of the present invention is shown. Figure 8 As shown, a work order duplicate checking system is also provided, including:

[0124] The word list acquisition module 810 is used to acquire the query work order data and the historical work order data and perform preprocessing to obtain the query work word list and the historical work word list;

[0125] The first similarity module 820 is used to convert the query work order word list and the history work order word list into word vectors and sum them up through the pre-trained Word2vector model, obtain the query work order vector and the history work order vector and calculate the first similarity;

[0126] A second similarity module 830, for querying the industrial word list and the historical industrial word list to calculate a second similarity through Jaccard;

[0127] A third similarity module 840, for extracting sentiment feature vectors of query work order data and historical work order data through BERT-based and calculating a third similarity;

[0128] The query module 850 is used to determine whether the query work order data is repeated through the first similarity, the second similarity and the third similarity.

[0129] In particular, an electronic device is also provided, the electronic device includes a processor and a memory, the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the work order duplicate checking method, for example, to execute the above-described Figure 1 Steps S100 to S500 of the method described above are also performed. Figure 6 The method comprises steps S510 to S540.

[0130] In particular, a computer-readable storage medium is also provided, the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the work order duplicate checking method, for example, to execute the above-described Figure 1 Steps S100 to S500 of the method described above are also performed. Figure 6 The method comprises steps S510 to S540.

[0131] The processor is used to complete various control logics of the device, which can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a single-chip microcomputer, an ARM (Acorn RISCMachine) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of these components. In addition, the processor can also be any traditional processor, microprocessor or state machine. The processor can also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a DSP and / or any other such configuration.

[0132] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0133] As an example, non-volatile storage media can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) as external cache memory. By way of illustration and not limitation, RAM can be obtained in many forms such as synchronous RAM (SRAM), dynamic RAM, (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The disclosed memory components or memories of the operating environment described herein are intended to include one or more of these and / or any other suitable types of memory.

[0134] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can include the processes of the above-mentioned method embodiments when executed. The storage medium can be a memory, a disk, a floppy disk, a flash memory, an optical storage device, etc.

[0135] The above are only some specific implementation methods of the present application, and any other improvements made based on the concept of the present application are deemed to be within the protection scope of the present application.

Claims

1. A work order duplicate checking method, characterized in that: include: Obtain query work order data and historical work order data and perform preprocessing to obtain query work word list and historical work word list; The query work order word list and the historical work order word list are respectively converted into word vectors and summed up using a pre-trained Word2vector model to obtain a query work order vector and a historical work order vector and calculate a first similarity; Calculating a second similarity based on the query word list and the historical word list by Jaccard; comprising: Converting the query word list and the history word list into an N-Gram sequence; Counting the co-occurrence times of N-Gram sequences of the query word list and the historical word list; Calculating the second similarity by using the Jaccard based on the number of co-occurrences; Extracting sentiment feature vectors of the query work order data and the historical work order data through BERT-based and calculating a third similarity; Determining whether the query work order data is repeated based on the first similarity, the second similarity, and the third similarity includes: Setting weights for the first similarity, the second similarity, and the third similarity; Summing the weighted first similarity, the second similarity and the third similarity to obtain a query work order similarity; Sorting the query work order data according to the query work order similarity; According to the sorting, the repeated query work order data is determined.

2. The method according to claim 1, characterized in that The step of obtaining the query work order data and the historical work order data and performing preprocessing to obtain the query work order word list and the historical work order word list includes: The historical work order data is segmented by Jieba and a related stop word list is constructed; Performing word segmentation and stop word processing on the query work order data and the historical work order data through the jieba and the related stop word list; The processed query work order data and the historical work order data are concatenated in terms of title, content and title to obtain a query work order word list and a historical work order word list.

3. The method according to claim 1, characterized in that The training of the Word2vector model includes: The training sample is used as the input of the Word2vector model, and the word vector is the output of the Word2vector model; The training samples are constructed through training corpus, and each training sample includes a central word and context words.

4. The method according to claim 3, characterized in that The method of converting the query work word list and the history work word list into word vectors and summing them up respectively by using the pre-trained Word2vector model to obtain the query work order vector and the history work order vector and calculate the first similarity includes: The Word2vector model performs vector initialization on the historical word list; The historical work order word list is input into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the historical work order vector; The query work order word list is input into the Word2vector model to obtain the word vector of each word in the word list and sum them up to obtain the query work order vector; The first similarity is calculated by using Cosine between the historical work order vector and the query work order vector.

5. The method according to claim 1, characterized in that The extracting the sentiment feature vectors of the query work order data and the historical work order data through BERT-based and calculating the third similarity includes: Performing sentiment analysis on the query work order data and the historical work order data through the BERT-based method to obtain query sentiment labels and historical sentiment labels; Converting the query sentiment label and the historical sentiment label into a query sentiment vector and a historical sentiment vector respectively; Calculating a third similarity by using the query sentiment vector and the historical sentiment vector; The emotion vector is shown in the following formula: Where: is the sentiment vector; is the positive sentiment vector; is the negative sentiment vector; is the neutral sentiment vector; The third similarity is shown in the following formula: Where: To query the sentiment vector; is the historical sentiment vector.

6. A work order duplicate checking system, characterized in that: include: Acquisition vocabulary module: used to acquire query work order data and historical work order data and perform preprocessing to obtain query work word lists and historical work word lists; The first similarity module is used to convert the query work order word list and the historical work order word list into word vectors and sum them up through the pre-trained Word2vector model, obtain the query work order vector and the historical work order vector and calculate the first similarity; The second similarity module is used to calculate the second similarity between the query word list and the historical word list through Jaccard; it includes: Converting the query word list and the history word list into an N-Gram sequence; Counting the co-occurrence times of N-Gram sequences of the query word list and the historical word list; Calculating the second similarity by using the Jaccard based on the number of co-occurrences; The third similarity module is used to extract the sentiment feature vectors of the query work order data and the historical work order data through BERT-based and calculate the third similarity; A query module: used for determining whether the query work order data is repeated based on the first similarity, the second similarity and the third similarity; including: Setting weights for the first similarity, the second similarity, and the third similarity; Summing the weighted first similarity, the second similarity and the third similarity to obtain a query work order similarity; Sorting the query work order data according to the query work order similarity; According to the sorting, the repeated query work order data is determined.

7. An electronic device, characterized in that: The electronic device includes a processor and a memory, the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the work order duplication checking method described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are called and executed by the processor, the computer-executable instructions prompt the processor to implement the work order duplication checking method described in any one of claims 1 to 5.