Work order processing method, apparatus, and electronic device

Abstract

The application discloses a work order processing method, belongs to the communication technical field, and can improve the efficiency of determining the target responsibility post of a work order. The method comprises the following steps: acquiring post description information and feedback content texts of each processing post of the work order; acquiring a corresponding word list of each processing post according to words in the feedback content text and words in the post description information; performing word vectorization on each word in each word list, and obtaining a corresponding text vector of each processing post according to the result of the word vectorization; taking the text vector corresponding to each processing post as the input of a preset work order responsibility determination model, obtaining the probability that the corresponding text vector output by the model matches the target responsibility post, and taking the probability as the probability that the corresponding processing post matches the target responsibility post; and determining the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post. The method realizes automatic determination of the target responsibility post of the work order according to work order data, and improves the work order processing efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, in particular to a work order processing method and device, electronic equipment and computer readable storage medium. BACKGROUND

[0002] At present, in the communication industry, fault processing is carried out by electronic work order (hereinafter referred to as "work order") and processed by manual. In the process of work order processing, multiple posts often need to be coordinated, or due to inaccurate work order assignment, the work order flows between multiple posts. When the work order is completed, in addition to the actual processing post of the fault, the work order contains the feedback content of other posts, which brings great difficulty to the archiving of the work order. Due to the uncertainty of the number of posts involved in the work order process, the uncertainty of the post responsibilities, and the uncertainty of the feedback content of each post, it is impossible to determine the main responsibility post of the work order through artificial intelligence in the process of work order archiving, so as to archive the work order.

[0003] In the prior art, the work order responsibility assignment and archiving process still cannot be completely automatically processed by information processing equipment, and the work order archiving has the problems of low processing efficiency and poor processing result standardization. SUMMARY

[0004] The embodiments of the present application provide a work order processing method and device, aiming to automatically identify the main responsibility post of the work order, so as to realize automatic archiving of the work order and improve the work order processing efficiency.

[0005] In a first aspect, the embodiments of the present application provide a work order processing method, comprising:

[0006] obtaining post description information and feedback content text of each processing post of a work order;

[0007] According to the words in the feedback content text of each processing post and the words in the post description information, a corresponding word list of each processing post is obtained;

[0008] Each word in each word list is subjected to word vectorization, and a corresponding text vector of each processing post is obtained according to the result of the word vectorization of each word;

[0009] Each text vector corresponding to each processing post is taken as the input of a preset work order responsibility assignment model, and the probability of matching a target responsibility post output by the work order responsibility assignment model is obtained as the probability of matching a target responsibility post of each processing post;

[0010] According to the probability of matching a target responsibility post of each processing post, a target responsibility post of the work order is determined.

[0011] In a second aspect, the embodiments of the present application provide a work order processing apparatus, comprising:

[0012] a processing post text acquisition module configured to acquire post description information and feedback content text of each processing post of a work order;

[0013] a word list acquisition module configured to acquire a word list corresponding to each processing post according to words in the feedback content text and words in the post description information of the processing post;

[0014] a vectorization module configured to perform word vectorization on each word in each word list, and obtain a text vector corresponding to each processing post according to a result of the word vectorization on the words;

[0015] a target responsibility post prediction module configured to acquire a probability that a target responsibility post matches each processing post as a probability that a target responsibility post matches each processing post by taking the text vector corresponding to each processing post as an input of a preset work order responsibility model, and taking a corresponding text vector output by the work order responsibility model as an output of the work order responsibility model;

[0016] a target responsibility post determination module configured to determine a target responsibility post of the work order according to the probability that a target responsibility post matches each processing post.

[0017] In a third aspect, the embodiments of the present application further disclose an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the work order processing method of the embodiments of the present application when executing the computer program.

[0018] In a fourth aspect, the embodiments of the present application provide a computer readable storage medium, which stores a computer program, and the program implements the steps of the work order processing method disclosed by the embodiments of the present application when executed by a processor.

[0019] The work order processing method disclosed by the embodiments of the present application comprises the following steps: obtaining post description information and feedback content texts of each processing post of a work order; obtaining a word list corresponding to each processing post according to words in the feedback content texts and words in the post description information; performing word vectorization on each word in each word list, and obtaining a text vector corresponding to each processing post according to the result of the word vectorization; taking the text vector corresponding to each processing post as an input of a preset work order responsibility assignment model, obtaining a probability that the work order responsibility assignment model outputs a corresponding text vector matching a target responsibility post as a probability that each processing post matches the target responsibility post; and determining a target responsibility post of the work order according to the probability that each processing post matches the target responsibility post, so as to automatically determine the target responsibility post of the work order according to work order data, thereby realizing automatic archiving of the work order and improving work order processing efficiency.

[0020] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to make the purposes, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0022] Figure 1 is one of the work order processing method flowcharts in the embodiments of the present application;

[0023] Figure 2 is a processing flowchart of the feedback content texts in the embodiments of the present application;

[0024] Figure 3 is the second work order processing method flowchart in the embodiments of the present application;

[0025] Figure 4 is a work order processing device structure diagram in the embodiments of the present application;

[0026] Figure 5 schematically shows a block diagram of an electronic device for executing the method according to the present application; and

[0027] Figure 6A storage unit for holding or carrying program code implementing the method according to the application is schematically shown. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] Embodiment one

[0030] As shown in the drawings, Figure 1 The work order processing method disclosed in the embodiments of the present application includes steps 100 to 140.

[0031] In step 100, the post description information and the feedback content text of each processing post of a work order are obtained.

[0032] The work order in the embodiments of the present application refers to work order data generated after flowing through one or more processing posts. For a work order, the work order may flow through and be processed by multiple different processing posts in sequence during the work order processing process, until the work order is processed and completed, and finally, the processing data of the work order is archived. During the work order processing process, each processing post may input the feedback content text for the work order, and flow the work order to the next processing post, or after performing the corresponding processing operation on the work order, end the flow of the work order, and submit the work order for archival management.

[0033] For each work order to be archived and managed, the work order data thereof usually includes the work order data of at least one processing post. The work order data includes but is not limited to one or more data: post description information of the processing post, feedback time, and feedback content text.

[0034] When performing the work order archiving processing, the main responsible post of the work order needs to be determined, and the work order is archived based on the main responsible post. The target responsible post in the embodiments of the present application is the main responsible post.

[0035] When the target responsible post of the work order is determined based on the work order data by using the artificial intelligence technology, the work order data corresponding to each processing post in the work order data of the work order needs to be obtained based on the dimension of the processing post. The work order data corresponding to each processing post at least includes the feedback content text of the processing post, the post description information of the processing post, and the feedback time.

[0036] In step 110, a word list corresponding to each processing post is obtained according to the words in the feedback content text of each processing post and the words in the post description information.

[0037] After obtaining the work order data corresponding to each processing post of the work order, for the work order data corresponding to each processing post, further text processing is performed to obtain the words in the feedback content text of each processing post for the work order and the key words in the post description information, such as the post name. Then, according to the words in the feedback content text and the words in the post description information, a word list corresponding to the corresponding processing post is obtained. Each word list is obtained based on the feedback content text of the corresponding processing post for the work order and carries the processing characteristics of the corresponding processing post, and thus can be used for subsequent target responsibility post identification.

[0038] As shown in some embodiments of the present application, Figure 2 According to the words in the feedback content text of each processing post and the words in the post description information, the word list corresponding to each processing post is obtained, which further includes: for each processing post, the following text processing operations in sub-step 1101 to sub-step 1104 are performed to obtain the word list corresponding to each processing post.

[0039] In sub-step 1101, the feedback content text of the processing post is subjected to word segmentation processing to obtain a word segmentation result.

[0040] In some embodiments of the present application, the feedback content text of the processing post is subjected to word segmentation processing to obtain a word segmentation result, which includes: the feedback content text of the processing post is subjected to word segmentation processing based on a preset word segmentation library. For example, based on a preset word segmentation library, regular expression is used for regular matching to perform word segmentation processing on the feedback content text to obtain each word constituting the feedback content text. For specific implementations of performing word segmentation processing on the feedback content text of a specified processing post based on a preset word segmentation library to obtain a word segmentation result, refer to the prior art, which will not be described herein.

[0041] The preset word segmentation library can use a general library in the prior art. In order to improve the accuracy of word segmentation, the preset word segmentation library is obtained by the following method: a communication field professional library is obtained by knowledge analysis on a preset corpus; login words are obtained based on historical work order data; and the preset word segmentation library is obtained by combining the login words and the communication field professional library.

[0042] For example, a corpus in the field of communication is selected, knowledge is sorted based on a knowledge graph analysis method to obtain a professional vocabulary in the field of communication. Then, knowledge is sorted from historical work order data to obtain words that do not appear in the preset expectation but appear in the work order data as login words, such as "tail fiber", "board card", "fusion", and the like. Finally, the login words and the words in the professional vocabulary in the field of communication jointly constitute the preset segmentation vocabulary. In some embodiments of the present application, the login words can be extracted from the work order data based on expert experience.

[0043] In sub-step 1102, words in the job description information of the processing post are obtained.

[0044] In some embodiments of the present application, the job description information can only include the name of the post, and the words in the job description information of the processing post of the work order are the job description information. In other embodiments of the present application, the job description information can also include information such as a province, and accordingly, the post name keyword in the job description information can also be obtained by performing entity recognition on the job description information, as the words in the job description information of the processing post that need to be obtained.

[0045] In step 1103, the words in the job description information are merged with the segmentation result to obtain a word list corresponding to the processing post.

[0046] Then, for a certain processing post, the job description information of the processing post is merged with the segmentation result of the feedback content text to obtain a word list corresponding to the processing post. For example, the feedback content of a certain processing post in a work order is "city electricity incoming call, alarm elimination, has been restored", and the job description information of the processing post is the post name "XX maintenance team". After the feedback content of the processing post is segmented, the segmentation result is obtained as: ["city electricity", "incoming call", "alarm", "elimination", "restore", "done"], and the keywords in the job description information are: ["XX", "maintenance", "team"]. Accordingly, after the job description information of the processing post is merged with the segmentation result of the feedback content text, the word list corresponding to the processing post is obtained as: ["XX", "maintenance", "team", "city electricity", "incoming call", "alarm", "elimination", "restore", "done"].

[0047] In the work order processing process, the post of post-processing work order often reuses the feedback content text of the processing post of the previously processed work order. Thus, the same feedback content appears in different posts, which reduces the recognition of feedback content to the processing post. By merging the segmentation results of the feedback content text and the post description information of the corresponding processing post, more rich text information for the processing post can be obtained, which helps to improve the accuracy of determining the matching target responsibility post probability based on the text content.

[0048] In step 1104, for each of the word lists, word denoising processing is performed respectively.

[0049] The denoising processing includes one or more of the following operations: deleting high-frequency common words, deleting stop words, deleting meaningless words, and deleting symbols.

[0050] To further improve the accuracy of determining the processing post based on the feedback content text, in some embodiments of the present application, the word list obtained in the foregoing step is further subjected to word denoising processing to delete noise introduced by high-frequency common words, stop words, meaningless words, and symbols.

[0051] In some embodiments of the present application, the word denoising processing on the word list includes deleting words with a predetermined meaning in the word list based on a regular expression. For example, the date, work order number, mobile phone number, and other contents with low relevance to work order processing included in the work order data are removed by the regular expression.

[0052] In some other embodiments of the present application, the word denoising processing on the word list includes performing word denoising processing on the word list based on a pre-established denoising word library, wherein the denoising word library includes any one or more of the following noise words: high-frequency common words, stop words, meaningless words, and symbols.

[0053] In some embodiments of the present application, a denoising word library can be constructed in advance, and words or symbols that are noise are added to the denoising word library. Then, when performing denoising processing on the word list, the word denoising processing on the word list can be performed based on the denoising word library by using a regular expression. The denoising word library is determined according to expert experience and word frequency statistics of historical work orders. For example, high-frequency common words and stop words in historical work order data can be analyzed and added to the denoising word library. For another example, meaningless words such as "de", "le", and "may" can also be added to the denoising word library according to the meaning of the text.

[0054] After the merging of the job description information of the processing post and the segmented results of the feedback content text in the preceding step, the obtained word list corresponding to the processing post is: ["XX", "maintenance", "team", "city power", "incoming call", "alarm", "eliminate", "restore", "finished"]. For example, after further denoising processing, the obtained word list is: ["XX", "maintenance", "city power", "incoming call", "alarm", "eliminate", "restore"].

[0055] After denoising processing, some meaningless words, stop words, high-frequency common words, symbols, and the like in the word list are deleted. In this way, when calculating the text vector of the word list subsequently, the obtained text vector has accurate expression ability for the characteristics of the processing post, thereby improving the accuracy of determining the matching target responsibility post probability of the processing post of the work order.

[0056] In step 120, word vectorization is performed on each word in each word list, and a text vector corresponding to the corresponding processing post is obtained according to the result of the word vectorization on the words.

[0057] Next, word vectorization is performed on each word in the word list to obtain a word vector of each word in the word list. Then, the word vectors of all words in the word list are concatenated to form a sequence as the text vector of the processing post.

[0058] In some embodiments of the present application, the word vectorization is performed on each word in each word list, and a text vector corresponding to the corresponding processing post is obtained according to the result of the word vectorization on the words, including: obtaining the term frequency-inverse document frequency of each word in the word list based on a preset work order sample set as the word vector corresponding to each word; and concatenating the word vectors corresponding to each word into a text vector corresponding to the corresponding processing post.

[0059] In the embodiments of the present application, the TF-IDF algorithm is used to realize word vectorization to obtain the word vector of each word in the word list. Taking a word list c including n words as an example, the word list can be represented as: [x1, x2, …, xn], where x1, x2, …, xn ∈ c. The term frequency TF(x) and the inverse document frequency IDF(x) of the word x can be calculated by the following formulas, respectively: i n i n i i i

[0060]

[0061] ​​​​​​​wherein, the term frequency of the term x i TF(x i ) represents the number of times that the term x i appears in all the work order samples included in the preset work order sample set (i.e. ) accounts for the proportion of the total number of terms included in the work order samples (i.e. x ).

[0062]

[0063] wherein, |D| represents the total number of all the work order samples included in the preset work order sample set, |{k:x i ∈d k}|+1 represents the total number of work order samples containing the term x i , and K is an integer.

[0064] Then, the term frequency-inverse document frequency (i.e., TD-IDF term frequency) of the term x i may be represented as:

[0065] TF-IDF(x i ) = TF(x i ) * IDF(x i ).

[0066] According to the above method, the term frequency-inverse document frequency of the terms x1, x2, …, x i , …, x n in the term list can be obtained respectively. In the embodiments of the present application, the term frequency-inverse document frequency of a term is taken as the word vector of the term, and then the sequence of the term frequency-inverse document frequency of each term is taken as the text vector of the processing post. For example, after the word vectorization of the terms in the term list [“XX”, “maintenance”, “city power”, “incoming call”, “alarm”, “eliminate”, “restore”] obtained by the foregoing steps, the text vector of the processing post can be represented as: [0.0011 0.024 0.0212 0.0211 0.0103 0.0082 0.0221].

[0067] In some embodiments of the present application, the preset work order sample set includes several work order samples. The work order samples can be the term list of each processing post in the corresponding work order obtained after the cleaning, word segmentation processing, and denoising processing of the original work order data of each processing post in each work order.

[0068] The several work order samples included in the preset work order sample set can be obtained by the following method: based on the processing post dimension, collecting the work order data of each processing post in each historical work order respectively, wherein the work order data at least includes the feedback content text of the corresponding processing post, and the post description information of the corresponding processing post; performing text processing on the work order data of each processing post to obtain the word list corresponding to the corresponding work order data as the work order sample corresponding to the work order data, wherein the word list includes the words in the feedback content text of the processing post in the corresponding work order data, and the words in the post description information of the processing post in the corresponding work order data.

[0069] The specific implementation of performing text processing (such as cleaning, word segmentation processing, and denoising processing) on the work order data of each processing post to obtain the word list corresponding to the corresponding work order data can be referred to the specific implementation of the sample data of the training sample of the work order responsibility assignment model in the following, which will not be described here.

[0070] Step 130, respectively taking the text vector corresponding to each processing post as the input of the preset work order responsibility assignment model to obtain the probability that the corresponding text vector output by the work order responsibility assignment model matches the target responsibility post as the probability that the corresponding processing post matches the target responsibility post.

[0071] Then, taking the text vector of each processing post as the input feature of the pre-trained work order responsibility assignment model, the work order responsibility assignment model classifies and maps the text vector of the processing post according to the pre-learned classification decision, so as to obtain the probability that each text vector matches the target responsibility post. Further, the probability that each text vector matches the target responsibility post is taken as the probability that the processing post corresponding to the text vector matches the target responsibility post.

[0072] For example, if the work order data of a work order records the feedback content text of P processing posts, the foregoing steps will obtain P word lists corresponding to the P processing posts one by one, each word list including the words in the feedback content text and the post description information of the corresponding processing post. After inputting the text vector generated according to each word list into the work order responsibility assignment model, the work order responsibility assignment model will obtain a corresponding probability value after processing the input text vector. After inputting the text vectors corresponding to the P processing posts into the work order responsibility assignment model, P probability values will be obtained, each probability value corresponding to the probability that a processing post matches the target responsibility post.

[0073] The specific training method of the work order responsibility assignment model is described in the following, which will not be described here.

[0074] Step 140, determining the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post.

[0075] In some embodiments of the present application, the greater the probability, the greater the likelihood that the corresponding processing post is the target responsibility post. The target responsibility post refers to the processing post that plays a major role in the process of processing the work order.

[0076] Through research on work order data, it is found that, in general, the feedback content of the target responsibility post is more detailed and complete than that of the secondary post, i.e., the text content is longer, and there is more effective content such as professional operation, which is significant. That is, when the processing post s i is the target responsibility post, the feedback content of other processing posts s j has c i ≥ c j i,j∈(0,N], N is the total number of processing posts processing the work order. After obtaining the probability of each processing post matching the target responsibility post for the work order, a processing post can be selected as the target responsibility post for the work order according to the probability.

[0077] In some embodiments of the present application, determining the target responsibility post for the work order according to the probability of each processing post matching the target responsibility post includes: obtaining the processing post that does not belong to a preset management post and satisfies a preset probability condition; and determining, among the obtained processing posts, the processing post with the earliest feedback time for the work order as the target responsibility post for the work order.

[0078] For example, a probability threshold can be set in advance, and the probability condition is set as: greater than or equal to the probability threshold. Only when the probability of the processing post matching the target responsibility post is greater than or equal to the preset probability threshold, the processing post can be the target responsibility post for the work order. The probability threshold can be determined according to experience or data verification results.

[0079] In some embodiments of the present application, the processing post that satisfies the preset probability condition can be obtained first. When there is only one processing post that satisfies the preset probability condition, the processing post can be directly determined as the target responsibility post. When there is no processing post that satisfies the preset probability condition, the work order can be submitted to a manual processing process for manual archiving.

[0080] In some embodiments of the present application, when the processing post that satisfies the preset probability condition includes multiple processing posts, the management post needs to be excluded.

[0081] In specific applications, the feedback content text of the target responsibility post of a work order is highly correlated with the post and profession to which the work order belongs, that is, the target responsibility post of a work order is usually a maintenance professional post, rather than a work order process control post. Therefore, when selecting the target responsibility post based on the probability of each processing post matching the target responsibility post output by the work order responsibility model, the control post needs to be excluded. In some embodiments of the present application, a control post set can be established in advance, and then, by matching each processing post with the control post in the control post set one by one, the control post in each processing post can be found and excluded.

[0082] In some embodiments of the present application, when only one processing post remains after excluding the control post, the remaining processing post can be directly used as the target responsibility post of the work order.

[0083] In some embodiments of the present application, after excluding the control post, if multiple processing posts remain and the probability of the remaining processing posts matching the target responsible post all meets the preset probability threshold, it is necessary to further determine the target responsible post based on the order in which the processing posts processed the work order. For example, from these multiple processing posts, the processing post that provided feedback on the work order the earliest may be selected as the target responsible post for the work order.

[0084] In a work order processing system, for each work order, the work order data records the feedback time and position description information of each processing position that performs the processing operation on the work order. In an embodiment of the present application, the processing position with the earliest feedback time for the work order can be determined based on the editing time of the feedback content text of the work order recorded in the work order data by each processing position.

[0085] In actual applications, for a work order, when the same feedback content appears, it is usually the work order control post or other processing post that copies the feedback content of the target responsibility post. Therefore, for feedback with the same content in the work order data, the target responsibility post has the characteristic of responding the earliest. That is, when the processing post i For target responsibility posts, for other processing posts j There is t i ≤t j i, j∈(0, N], where N is the total number of processing posts that process the work order. In the embodiment of the present application, by selecting the processing post with the earliest feedback time for the work order from the processing posts whose probabilities meet the preset probability conditions as the target responsible post for the work order, it is more in line with the characteristics of the actual work order data and ensures the accuracy of determining the target responsible post for the work order.

[0086] In order to facilitate readers' understanding of the work order processing method disclosed in the embodiments of this application, the training process of the work order responsibility model is illustrated below with examples.

[0087] like Figure 3As shown, the training method of the work order responsibility model includes: steps 310 to 330.

[0088] Step 310: Obtain several training samples.

[0089] Among them, each training sample corresponds to the work order data of a processing position in a historical work order, and the sample data of the training sample includes: words in the feedback content text of the corresponding processing position in the corresponding historical work order, and words in the job description information of the corresponding processing position; the sample label of the training sample is: whether the corresponding sample data matches the true value of the target responsibility position.

[0090] In an embodiment of the present application, sample data of several training samples can be obtained by collecting historical work order data and performing text processing on the historical work order data based on a natural language processing method. For example, historical work order data is collected, including but not limited to: feedback content (hereinafter represented by the symbol "c"), the name of the processing position (hereinafter represented by the symbol "s"), and the feedback time (hereinafter represented by the symbol "t"). For a work order D processed by multiple positions, it is assumed that the feedback contents of the N processing positions in the work order D are c1, c2, ..., c i ,…c N , the job titles are s1, s2, …, s i ,…s N , the feedback time is t1, t2,…, t i ,…t N , then according to the feedback content c1, c2, ..., c i ,…c N The words included in the work order D, and the job titles s1, s2, ..., s i ,…s n , generate sample data of N training samples corresponding to work order D.

[0091] The text of the work order data contains many formatting phrases or symbols, as well as dates, work order numbers, mobile phone numbers, and other content that has little relevance to work order processing. In order to improve the generalization ability of the processing results of the trained work order responsibility model, in some embodiments of the present application, it is also necessary to clean the feedback content text contained in the work order data to remove formatting phrases or symbols, dates, work order numbers, mobile phone numbers, and other content that has little relevance to work order processing. In some embodiments of the present application, the work order data can be cleaned using regular expressions.

[0092] For each work order, according to the feedback content text of each processing post in the work order cleaned work order data, and the name of each processing post, the sample data of the training sample corresponding to each processing post in the work order is generated. Specifically, the words in the feedback content text of each processing post in the work order cleaned work order data are obtained, and the name of each processing post is obtained, the work order data corresponding to each processing post in the work order is generated, and the work order data acquisition based on the processing post dimension is completed. In this way, each work order data corresponding to a specified processing post obtained at least includes: feedback content text and the name of the processing post (i.e. post description information). In the embodiment of the application, the work order data of the historical work order usually records which processing post of the historical work order is the target responsibility post, so when the work order data is collected based on the processing post dimension, the information of whether the processing post is the target responsibility post for the specified historical work order can also be collected, that is, the information of whether a piece of work order data matches the target responsibility post can be collected (for example, when a processing post is the target responsibility post for a specified historical work order, it can be considered that the work order data of the processing post in the historical work order matches the target responsibility post).

[0093] Next, based on each work order data corresponding to a processing post, a training sample is constructed respectively.

[0094] For example, first, the feedback content text of the processing post in each work order data is subjected to word segmentation processing to obtain a word segmentation result; then, the name of the processing post in the work order data is obtained; then, the name of the processing post is merged with the word segmentation result to obtain a word list corresponding to the processing post; finally, the word list is subjected to word denoising processing, and the word list obtained after denoising processing is taken as the sample data of a training sample corresponding to the processing post.

[0095] The specific implementation of the word segmentation processing of the feedback content text in the work order data of the processing post is described in the foregoing, which will not be repeated here. The specific implementation of merging the name of the processing post with the word segmentation result to obtain a word list corresponding to the processing post is described in the foregoing, which will not be repeated here. The specific implementation of the word denoising processing of the word list is described in the foregoing, which will not be repeated here.

[0096] For a work order handled by multiple posts, only one post plays a major role in the process, which is the target responsibility post of the work order. If a certain processing post is the target responsibility post of a certain work order, when generating a training sample according to the work order data of the processing post in the work order, the value of the sample label of the training sample can be set as the target responsibility post identifier (such as 1), which is used to indicate that the sample data corresponds to the target responsibility post; if a certain processing post is not the target responsibility post of a certain work order, when generating a training sample according to the work order data of the processing post in the work order, the value of the sample label of the training sample can be set as the non-target responsibility post identifier (such as 0), which is used to indicate that the sample data corresponds to the non-target responsibility post.

[0097] According to the foregoing method, based on the work order data of each historical work order, at least one training sample can be obtained respectively. In this way, based on a plurality of historical work orders, a plurality of training samples can be obtained.

[0098] In step 320, for each of the training samples, the words in the corresponding sample data are word vectorized, and the text vector corresponding to the training sample is obtained according to the result of the word vectorization of the words.

[0099] In some embodiments of the present application, the words in the corresponding sample data are word vectorized, and the text vector corresponding to the training sample is obtained according to the result of the word vectorization of the words, which includes: obtaining the term frequency-inverse document frequency of each word in the sample data based on the preset work order sample set as the word vector corresponding to each word; and splicing the word vectors corresponding to each word into the text vector corresponding to the corresponding training sample. The specific implementation of the term frequency-inverse document frequency of each word in the sample data based on the preset work order sample set is described above in the implementation of the term frequency-inverse document frequency of each word in the word list corresponding to the processing post based on the preset work order sample set, which will not be described here. After obtaining the word vectors of the words in the sample data, the word vectors of the words are spliced to obtain the text vector corresponding to the training sample.

[0100] In step 330, the work order responsibility assignment model is trained based on the text vectors corresponding to each of the training samples and the sample labels.

[0101] The training of the work order responsibility assignment model based on the text vectors corresponding to each of the training samples and the sample labels includes: obtaining, by the work order responsibility assignment model, the prediction probability of the text vector corresponding to each of the training samples matching the target responsibility post, and iteratively optimizing the parameters of the work order responsibility assignment model to minimize the error between the prediction probability and the sample label of the corresponding training sample.

[0102] In the embodiments of the present application, the work order responsibility assignment model adopts a LightGBM architecture.

[0103] LightGBM (Light Gradient Boosting Machine) is a framework implementing the GBDT algorithm, supporting efficient parallel training and small memory occupation.

[0104] For a training sample, the input of the work order responsibility assignment model is a text vector corresponding to the training sample, and the output of the model is a prediction probability of the training sample matching a target responsibility post, that is, the model takes the text vector as the feature of the processing post and outputs the prediction probability of the text vector matching the target responsibility post. The prediction probability of the text vector matching the target responsibility post is the prediction probability of the sample data of the training sample matching the target responsibility post. When training the work order responsibility assignment model based on the obtained training samples, the work order responsibility assignment model makes decisions based on the text vectors corresponding to the training samples, obtains the prediction probability of the text vector of each training sample matching the target responsibility post, and iteratively optimizes the model parameters to reduce the error between the prediction probability and the sample label of the training sample until the error converges.

[0105] In the model training process, K-fold cross-validation can be used to iteratively optimize the initial work order responsibility assignment model, adjust the parameters of the initial work order responsibility assignment model according to the classification accuracy of the initial work order responsibility assignment model, determine the optimal parameters of the initial work order responsibility assignment model, and obtain the final work order responsibility assignment model.

[0106] In some embodiments of the present application, parameter updating is performed by a Gradient-based One-Side Sampling (GOSS) algorithm, and the parameter updating process is as follows:

[0107]

[0108] wherein ω is a model parameter.

[0109] In some embodiments of the present application, the loss function of the work order responsibility assignment model is a binarylogloss for binary classification, and the formula is as follows:

[0110]

[0111] wherein I represents the number of training samples, y represents the sample label of the i-th training sample, and y i represents the prediction probability of the i-th training sample corresponding to the processing post matching the target responsibility post.

[0112] In the model training process, the loss function value of the work order responsibility assignment model is calculated once per iteration of the model parameters. When the loss function value of the model is small enough (e.g., meets a preset convergence condition), the iteration is stopped, and the model parameters at this time are considered as the optimal parameters, and the training process of the work order responsibility assignment model is completed. The model parameters obtained at this time are used to initialize the work order responsibility assignment model, and the initialized work order responsibility assignment model is applied to estimate the probability of matching the target responsible post.

[0113] The work order processing method disclosed in the embodiments of the present application comprises the following steps: obtaining post description information and feedback content text of each processing post of a work order; obtaining a word list corresponding to each processing post according to words in the feedback content text of each processing post and words in the post description information; performing word vectorization on each word in each word list, and obtaining a text vector corresponding to each processing post according to the result of the word vectorization; taking the text vector corresponding to each processing post as an input of a preset work order responsibility assignment model, obtaining a probability that the text vector output by the work order responsibility assignment model matches a target responsible post, and taking the probability as a probability that each processing post matches the target responsible post; and determining the target responsible post of the work order according to the probability that each processing post matches the target responsible post, so as to automatically determine the target responsible post of the work order according to work order data, thereby enabling automatic archiving of the work order and improving work order processing efficiency.

[0114] In the work order responsibility assignment scenario, the accuracy of manual processing is limited, and the processing standard fluctuates. The work order processing method disclosed in the embodiments of the present application combines natural language processing technology and machine learning technology to automatically identify the target responsible post of the work order based on work order data, which not only improves the work order responsibility assignment efficiency, but also improves the standardization of work order responsibility assignment.

[0115] In the work order responsibility assignment scenario, there are many posts involved in work order processing, and the characteristics of the posts are crossed, and the data volume is usually insufficient to support modeling of the processing posts. The work order processing method disclosed in the embodiments of the present application converts the classification problem of hundreds of target processing posts into a binary classification problem of whether it is a target responsible post, performs cleaning on the feedback content text in the work order data through natural language processing technology, uses a machine learning model to perform modeling to realize binary classification prediction, and then further automatically determines the target responsible post through screening logic, which not only significantly improves the work order responsibility assignment efficiency, but also tests that the accuracy rate of determining the target responsible post using the work order processing method disclosed in the embodiments of the present application is more than 99%.

[0116] Embodiment Two

[0117] The work order processing device disclosed in the embodiments of the present application comprises the following steps: Figure 4 As shown in the figure, the device comprises:

[0118] The processing post text acquisition module 400 is configured to acquire post description information and feedback content text of each processing post of the work order;

[0119] The word list acquisition module 410 is configured to acquire, according to words in the feedback content text and words in the post description information of each processing post, a word list corresponding to the processing post;

[0120] The vectorization module 420 is configured to perform word vectorization on each word in each word list, and obtain a text vector corresponding to each processing post according to a result of the word vectorization on the words.

[0121] The target responsibility post prediction module 430 is configured to take the text vector corresponding to each processing post as an input of a preset work order responsibility model, acquire a probability that a text vector output by the work order responsibility model matches a target responsibility post as a probability that the processing post matches the target responsibility post.

[0122] The target responsibility post determination module 440 is configured to determine the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post.

[0123] In some embodiments of the present application, the target responsibility post determination module 440 is further configured to:

[0124] acquire the processing post that does not belong to the preset management and control post and that satisfies the preset probability condition;

[0125] In the acquired processing post, determine the processing post with the earliest feedback time of the work order as the target responsibility post of the work order.

[0126] In some embodiments of the present application, the work order responsibility model is obtained by training the following method:

[0127] acquire a plurality of training samples, wherein each training sample corresponds to work order data of a processing post in a historical work order, and sample data of the training sample includes words in feedback content text of the corresponding processing post in the corresponding historical work order and words in post description information of the corresponding processing post; and a sample label of the training sample is a true value of whether the corresponding sample data matches a target responsibility post;

[0128] For each training sample, perform word vectorization on each word in the corresponding sample data, and obtain a text vector corresponding to the training sample according to a result of the word vectorization on the words.

[0129] The ticket responsibility assignment model is trained based on the text vector and the sample label corresponding to each training sample, and includes: obtaining, by the ticket responsibility assignment model, a prediction probability of matching a target responsibility post for the text vector corresponding to each training sample, and iteratively optimizing parameters of the ticket responsibility assignment model to minimize an error between the prediction probability and a sample label of the corresponding training sample.

[0130] In some embodiments of the present application, the vectorization module 420 is further configured to:

[0131] Obtain, as a word vector corresponding to each word in the word list, a term frequency-inverse document frequency of each word in the word list based on a preset ticket sample set, respectively;

[0132] Concatenate the word vector corresponding to each word into a text vector corresponding to the processing post.

[0133] In some embodiments of the present application, the word list acquisition module 410 is further configured to:

[0134] For each processing post, the following text processing operations are performed to obtain a word list corresponding to each processing post:

[0135] Performing word segmentation processing on the feedback content text of the processing post to obtain a word segmentation result;

[0136] Obtaining words in the post description information of the processing post;

[0137] Merging the words in the post description information with the word segmentation result to obtain a word list corresponding to the corresponding processing post;

[0138] For each word list, word denoising processing is performed, wherein the denoising processing includes one or more of the following operations: deleting high-frequency common words, deleting stop words, deleting meaningless words, and deleting symbols.

[0139] In some embodiments of the present application, the word segmentation processing on the feedback content text of the processing post includes:

[0140] Performing word segmentation processing on the feedback content text of the processing post based on a preset word segmentation vocabulary; wherein the preset word segmentation vocabulary is obtained by the following method:

[0141] Performing knowledge sorting on a preset corpus to obtain a professional vocabulary in the communication field;

[0142] Obtaining login words based on historical ticket data;

[0143] Obtaining a preset word segmentation vocabulary by combining the login words and the professional vocabulary in the communication field.

[0144] In some embodiments of the present application, the work order allocation model adopts a LightGBM architecture.

[0145] The work order processing device disclosed in the embodiments of the present application is used to implement the work order processing method described in Embodiment One of the present application. The specific implementation of each module of the device will not be described again, and the specific implementation of the corresponding steps of the method embodiments can be referred to.

[0146] The work order processing device disclosed in the embodiments of the present application obtains the post description information and the feedback content text of each processing post of a work order; obtains a word list corresponding to each processing post according to the words in the feedback content text and the words in the post description information of each processing post; performs word vectorization on each word in each word list, and obtains a text vector corresponding to each processing post according to the result of the word vectorization on each word; takes the text vector corresponding to each processing post as the input of a preset work order allocation model, obtains the probability that the corresponding text vector output by the work order allocation model matches a target responsibility post as the probability that each processing post matches a target responsibility post; and determines the target responsibility post of the work order according to the probability that each processing post matches a target responsibility post, which realizes automatic determination of the target responsibility post of a work order according to work order data, thereby enabling automatic archiving of work orders and improving work order processing efficiency.

[0147] In the work order allocation scenario, the accuracy of manual processing is limited, and the processing standard fluctuates. The work order processing device disclosed in the embodiments of the present application realizes automatic identification of the target responsibility post of a work order based on work order data by combining natural language processing technology and machine learning technology, which not only improves the work order allocation efficiency, but also improves the standardization of work order allocation.

[0148] In the work order allocation scenario, there are many posts involved in work order processing, and the characteristics of the posts are crossed. The data volume is usually insufficient to support modeling of the processing posts respectively. The work order processing method disclosed in the embodiments of the present application converts the classification problem of hundreds of target processing posts into a binary classification problem of whether it is a target responsibility post, performs cleaning on the feedback content text in the work order data through natural language processing technology, uses a machine learning model to model and realize binary classification prediction, and then further automatically determines the target responsibility post through screening logic, which not only significantly improves the work order allocation efficiency, but also tests that the accuracy rate of determining the target responsibility post using the work order processing device disclosed in the embodiments of the present application is more than 99%.

[0149] Various embodiments are described herein with reference to the accompanying drawings. The use of the terms "example" and "exemplary" throughout the description can mean that a particular implementation is provided as a non-limiting example, as opposed to representing the only implementation. But the use of these terms is not intended to imply that feature is essential to implementation or that it will always be included in a particular implementation. Various embodiments can be implemented in numerous ways, including as a process, an apparatus, a system, a method, a computer readable medium such as a computer readable storage medium, or a computer program product such as a computer program.

[0150] The above describes in detail a work order processing method and device provided by the present application. The principles and implementation manners of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range can be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

[0151] The device embodiments described above are only schematic, and the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement without creative labor.

[0152] The various component embodiments of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that microprocessors or digital signal processors (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the electronic device according to the embodiments of the present application. The present application can also be implemented as a device or apparatus program (for example, a computer program and a computer program product) for executing part or all of the methods described herein. Such program implementing the present application can be stored on a computer readable medium, or can have one or more signals in the form. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0153] For example, Figure 5An electronic device is shown, in which the method according to the application can be implemented. The electronic device can be a PC, a mobile terminal, a personal digital assistant, a tablet computer, etc. The electronic device traditionally comprises a processor 510 and a memory 520 and a program code 530 stored on the memory 520 and executable on the processor 510, which, when executed by the processor 510, implements the method described in the above embodiments. The memory 520 can be a computer program product or a computer readable medium. The memory 520 can be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory 520 has a storage space 5201 for the program code 530 of the computer program for executing any of the method steps described above. For example, the storage space 5201 for the program code 530 can comprise individual computer programs for implementing the various steps in the above methods, respectively. The program code 530 is computer readable code. The computer programs can be read out of or written into one or more computer program products. The computer program products comprise program code carriers such as hard disks, compact discs (CDs), memory cards or floppy disks. The computer programs comprise computer readable code which, when executed on the electronic device, causes the electronic device to perform the method according to the above embodiments.

[0154] The embodiments of the application further disclose a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the steps of the work order processing method according to the embodiment one of the application.

[0155] Such a computer program product can be a computer readable storage medium, which can have a structure analogous to the storage means 520 in the electronic device shown. The program code can be stored in the computer readable storage medium, for example, in a compressed form. The computer readable storage medium is typically a portable or fixed storage unit, as described with reference to the memory 520. Figure 5 The program code can be stored in the computer readable storage medium, for example, in a compressed form. The computer readable storage medium is typically a portable or fixed storage unit, as described with reference to the memory 520. Figure 6 The computer readable storage medium typically comprises computer readable code 530' which is code that is readable by a processor, which, when executed by the processor, implements the various steps of the methods described above.

[0156] Reference herein to "one embodiment", "an embodiment" or "one or more embodiments” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in one embodiment" in various places in this specification are not necessarily all referring to the same embodiment.

[0157] In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the application can be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

[0158] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word 'comprising' does not exclude the presence of elements or steps other than those listed in a claim. The word 'a' or 'an' preceding an element does not exclude the presence of a plurality of such elements. The application can be implemented by means of both hardware and software, and any combination thereof. In a unitary claim, several devices or sub-claims can be joined by means of the word 'or'. The word 'first','second', 'third', etc. do not imply any order. The terms 'first','second', 'third', etc. are to be interpreted according to their meaning in the context and are not to be interpreted as a ranking.

[0159] It has to be noted that, while the above describes example embodiments of the application, these are merely given by way of non-limiting examples. Many variations and modifications of the embodiments described herein can become apparent to those skilled in the art once they are made aware of the general inventive concept. It is therefore contemplated that the application shall also cover any variations and modifications to the above described embodiments that fall within the scope of the claims.

Claims

1. A work order processing method, characterized in that: include: Obtain the job description information and feedback content text of each processing position of the work order; Obtaining a word list corresponding to each processing position according to the words in the feedback content text and the words in the position description information of each processing position; Performing word vectorization on each word in each of the word lists, and obtaining a text vector corresponding to the corresponding processing position based on the result of performing the word vectorization on each word; The text vector corresponding to each processing post is used as the input of the preset work order responsibility model, and the probability that the corresponding text vector output by the work order responsibility model matches the target responsibility post is obtained as the probability that the corresponding processing post matches the target responsibility post; Determine the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post; The work order responsibility model is trained using the following method: Obtain several training samples, where each training sample corresponds to work order data for a processing position in a historical work order. The sample data of the training sample includes: words in the feedback content text of the corresponding processing position in the corresponding historical work order, and words in the job description information of the corresponding processing position; the sample label of the training sample is: whether the corresponding sample data matches the true value of the target responsibility position; For each of the training samples, word vectorization is performed on each word in the corresponding sample data, and a text vector corresponding to the training sample is obtained based on the result of the word vectorization on each word; Based on the text vectors and sample labels corresponding to each of the training samples, a work order responsibility model is trained, including: obtaining the predicted probability that the text vector corresponding to each of the training samples matches the target responsibility post through the work order responsibility model, and iteratively optimizing the parameters of the work order responsibility model with the goal of minimizing the error between the predicted probability and the sample label of the corresponding training sample.

2. The method according to claim 1, characterized in that Determining the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post includes: Obtaining the processing post that does not belong to the preset control post and whose probability meets the preset probability condition; Among the acquired processing posts, the processing post with the earliest feedback time for the work order is determined as the target responsible post for the work order.

3. The method according to any one of claims 1 to 2, characterized in that The word vectorization is performed on each word in each word list, and a text vector corresponding to the corresponding processing position is obtained according to the result of the word vectorization on each word, including: Obtain the word frequency of each word in the word list based on a preset work order sample set minus the inverse document word frequency as the word vector corresponding to each word; The word vectors corresponding to each word are concatenated into the text vector corresponding to the corresponding processing position.

4. The method according to any one of claims 1 to 2, characterized in that The step of obtaining a word list corresponding to each processing position according to the words in the feedback content text and the words in the position description information of each processing position includes: For each processing post, perform the following text processing operations to obtain a word list corresponding to each processing post: Performing word segmentation processing on the feedback content text of the processing post to obtain a word segmentation result; Obtaining words from the job description information of the processing post; Merge the words in the job description information with the word segmentation result to obtain a word list corresponding to the corresponding processing position; For each of the word lists, word denoising processing is performed respectively, wherein the denoising processing includes one or more of the following operations: deleting high-frequency common words, deleting stop words, deleting meaningless words, and deleting symbols.

5. The method according to claim 4, characterized in that The word segmentation processing of the feedback content text of the processing post includes: The feedback content text of the processing post is segmented based on a preset segmentation vocabulary; wherein the preset segmentation vocabulary is obtained by the following method: Sorting out knowledge from the preset corpus to obtain professional vocabulary in the communications field; Get login words based on historical work order data; The login word and the professional vocabulary in the communication field are combined to obtain a preset word segmentation vocabulary.

6. The method according to any one of claims 1 to 2, characterized in that The work order responsibility model adopts the LightGBM architecture.

7. A work order processing device, characterized in that: include: The processing post text acquisition module is used to obtain the post description information and feedback content text of each processing post in the work order; A word list acquisition module, configured to acquire a word list corresponding to each processing position according to the words in the feedback content text and the words in the position description information of each processing position; A vectorization module, configured to perform word vectorization on each word in each of the word lists, and obtain a text vector corresponding to the corresponding processing position based on the result of performing the word vectorization on each word; A target responsibility post prediction module is configured to use the text vector corresponding to each processing post as input to a preset work order responsibility model, obtain the probability that the corresponding text vector output by the work order responsibility model matches the target responsibility post, and use this probability as the probability that the corresponding processing post matches the target responsibility post; A target responsibility post determination module is used to determine the target responsibility post of the work order according to the probability that each processing post matches the target responsibility post; The work order responsibility model is trained using the following method: Obtain several training samples, where each training sample corresponds to work order data for a processing position in a historical work order. The sample data of the training sample includes: words in the feedback content text of the corresponding processing position in the corresponding historical work order, and words in the job description information of the corresponding processing position; the sample label of the training sample is: whether the corresponding sample data matches the true value of the target responsibility position; For each of the training samples, word vectorization is performed on each word in the corresponding sample data, and a text vector corresponding to the training sample is obtained based on the result of the word vectorization on each word; Based on the text vectors and sample labels corresponding to each of the training samples, a work order responsibility model is trained, including: obtaining the predicted probability that the text vector corresponding to each of the training samples matches the target responsibility post through the work order responsibility model, and iteratively optimizing the parameters of the work order responsibility model with the goal of minimizing the error between the predicted probability and the sample label of the corresponding training sample.

8. An electronic device comprising a memory, a processor, and a program code stored in the memory and executable on the processor, wherein: When the processor executes the program code, the work order processing method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having program code stored thereon, characterized in that: When the program code is executed by a processor, the steps of the work order processing method described in any one of claims 1 to 6 are implemented.

Images