Method and device for extracting work order text subject

By first clustering and then extracting topics, K-Means and DBSCAN algorithms are used to process electric heating text data, which solves the deviation problem of existing algorithms when extracting electric heating text data topics, and achieves accurate identification of customer needs.

CN112270191BActive Publication Date: 2024-12-03STATE GRID BEIJING ELECTRIC POWER CO +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011296432.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-18
Publication Date
2024-12-03
Estimated Expiration
2040-11-18

AI Technical Summary

Technical Problem

The existing LDA and PLSA algorithms are poor in extracting topics of electric heating complaint and consultation text data, resulting in a deviation in understanding customer needs and the inability to accurately identify customer actual needs.

Method used

采用先进行聚类的方法,通过K均值K-Means聚类算法将文本数据划分为多个集合,然后使用具有噪声的基于密度DBSCAN聚类算法提取主题信息,构建关键词库和风险主题库,确定主题词的频率和隶属度,生成用户的投诉和咨询热点。

Benefits of technology

It effectively solves the deviation problem of LDA and PLSA algorithms when extracting the topic of electric heating text data, improves the accuracy of identifying customer needs, and can more accurately identify customer actual needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112270191B_ABST
    Figure CN112270191B_ABST
Patent Text Reader

Abstract

The present application discloses a method and device for extracting the topic of a work order text. The method includes: obtaining the text data of the work order; clustering the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; for each text set, extracting the topic information of all texts in the text set according to a second predetermined clustering algorithm to obtain the subject words of the text set. The present application solves the technical problem that the related art uses LDA, PLSA and other algorithms to accurately extract the topics to be expressed by the electric heating complaint and consultation text data, resulting in a large deviation in the understanding of the customer's topic and the inability to accurately identify the customer's actual needs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of text recognition, and in particular, to a method and device for extracting the subject of a work order text. Background Art

[0002] At present, the main methods that can be used to extract text information topics include LDA and PLSA topic extraction. Algorithms such as LDA and PLSA are all unsupervised topic extraction machine learning technologies. Topics are extracted through a topic-based document generation algorithm. Only when there is text information with obvious differences in the target category, the automatic extraction of document topics will be effective. However, for electric heating complaints and consultation text data, each text has great similarities, and the differences in topic information and category divisions are not obvious. Therefore, the use of document topic generation models such as LDA to extract topics from electric heating complaints and consultation text data has little effect and cannot effectively extract the topics that the electric heating complaints and consultation text data want to express. That is, there is a certain deviation in understanding the topics expressed by customers, and customer needs cannot be accurately identified.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a method and device for extracting the subject of a work order text, so as to at least solve the technical problem that the LDA, PLSA and other algorithms used in the related technology cannot accurately extract the subject to be expressed in the electric heating complaint and consultation text data, resulting in a large deviation in the understanding of the subject to be expressed by the customer and an inability to accurately identify the customer's actual needs.

[0005] According to one aspect of an embodiment of the present application, a method for extracting the subject of a work order text is provided, comprising: obtaining text data of the work order; clustering each text corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; for each text set, extracting subject information from all texts in the text set according to a second predetermined clustering algorithm to obtain a subject word of the text set.

[0006] Optionally, the first predetermined clustering algorithm includes: a K-Means clustering algorithm; the second predetermined clustering algorithm includes: a density-based DBSCAN clustering algorithm with noise.

[0007] Optionally, before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the method includes: performing word segmentation on the text data to obtain a phrase sequence; performing part-of-speech tagging on each phrase in the phrase sequence to obtain a target phrase corresponding to each part-of-speech, where the parts of speech include nouns, verbs, and adverbs; determining the frequency of occurrence of the target phrase, and determining a keyword group based on the frequency; and constructing a keyword library based on the keyword group.

[0008] Optionally, determining a keyword group according to frequency includes: determining information entropy of a target phrase according to frequency, wherein information entropy is used to measure the amount of information; and taking a target phrase whose information entropy is greater than a first threshold as a keyword group.

[0009] Optionally, before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the method also includes: denoising the text data, and the denoising method at least includes: removing punctuation marks and numbers; converting English letters to lowercase and extracting roots of English words.

[0010] Optionally, when there are multiple subject words, after obtaining the subject words of the text set, the method also includes: determining multiple memberships corresponding to the multiple subject words of the text set; determining a target membership that is greater than a second threshold among the multiple memberships; using the subject word corresponding to the target membership as the target subject word, and using the sentence composed of the target subject word as the work order subject corresponding to the text set.

[0011] Optionally, the text data of the work order includes at least one of the following: electric heating business complaint data, electric heating business consultation data.

[0012] Optionally, after extracting topic information from all texts in the text set according to a second predetermined clustering algorithm to obtain the subject words of the text set, the method also includes: generating a risk subject library according to the subject words corresponding to the text set; determining the frequency of the subject words based on the risk subject library, and taking the subject words in the subject words whose frequency is greater than a third threshold as the user's complaint hotspots and / or consultation hotspots.

[0013] According to another aspect of an embodiment of the present application, a device for extracting the subject of a work order text is also provided, including: an acquisition module for acquiring text data of a work order; a clustering module for clustering the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; an extraction module for extracting subject information from all texts in the text set according to a second predetermined clustering algorithm for each text set to obtain a subject word of the text set.

[0014] According to another aspect of an embodiment of the present application, a processor is further provided, the processor being used to run a program stored in a memory, wherein when the program is run, any method of extracting a work order text subject is executed.

[0015] In an embodiment of the present application, a method is adopted in which the types corresponding to each text set are first obtained, and then the keywords of each text set are identified. The text data of the work order is obtained; the texts corresponding to the text data are clustered according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; for each text set, the subject information of all texts in the text set is extracted according to a second predetermined clustering algorithm to obtain the keywords of the text set, thereby achieving the technical effect of first clustering the texts corresponding to the text data using the first predetermined algorithm to obtain multiple text sets, and then identifying the topics corresponding to each text set of the text set based on the second predetermined algorithm, thereby solving the technical problem that the related technology using algorithms such as LDA and PLSA cannot accurately extract the topics to be expressed by the electric heating complaint and consultation text data, resulting in a large deviation in the understanding of the topic to be expressed by the customer, and cannot accurately identify the actual needs of the customer. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 It is a flowchart of an optional method for extracting the subject of a work order text according to an embodiment of the present application;

[0018] Figure 2 It is a structural diagram of an optional device for extracting the subject of a work order text according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0020] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations 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 necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] According to an embodiment of the present application, a method embodiment for extracting the subject of a work order text is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0022] Figure 1 A method for extracting the subject of a work order text according to an embodiment of the present application is as follows: Figure 1 As shown, the method comprises the following steps:

[0023] Step S102, obtaining text data of the work order;

[0024] Step S104, clustering the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain a plurality of text sets, wherein different sets correspond to different types;

[0025] Step S106: for each text set, extracting topic information from all the texts in the text set according to a second predetermined clustering algorithm to obtain the topic words of the text set.

[0026] In the method for extracting the topic of the work order text, first, the text data of the work order is obtained; then, the texts corresponding to the text data are clustered according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; finally, for each text set, the topic information of all the texts in the text set is extracted according to a second predetermined clustering algorithm to obtain the subject words of the text set, thereby achieving the technical effect of first clustering the texts corresponding to the text data using the first predetermined algorithm to obtain multiple text sets, and then identifying the topics corresponding to each text set in the text set based on the second predetermined algorithm, thereby solving the technical problem that the related technology using algorithms such as LDA and PLSA cannot accurately extract the topics to be expressed by the electric heating complaint and consultation text data, resulting in a large deviation in the understanding of the topic to be expressed by the customer, and cannot accurately identify the actual needs of the customer.

[0027] It should be noted that the first predetermined clustering algorithm can be a K-means clustering algorithm; the second predetermined clustering algorithm can be a density-based DBSCAN clustering algorithm with noise. It should also be noted that K-Means clustering uses Euclidean distance as a measurement standard and specifies the number of classes ultimately formed in the cluster. It can first initialize the category centers of the specified number of categories and eventually form the specified number of clusters after continuous iterations. Optionally, K-means clustering can use the tf-idf weight of each word in the electric heating business complaint and consultation text data as a sample feature, perform text clustering on the electric heating business complaint and consultation text data, and obtain K-category texts of the subject terms to be extracted.

[0028] In some optional embodiments of the present application, embedded feature selection can be performed through supervised learning algorithms, such as random forest algorithms, to automatically discover which word in the text has a high degree of influence on the classification when learning which category a sentence belongs to. Random forest feature importance is a type of embedded feature selection. Through the importance of random forest features, the subject words of the accepted content can be obtained, but the extracted subject words may be short in length and poor in interpretability. Therefore, it is necessary to extract the context fragments of the corresponding subject words in the original text. Specifically, after obtaining the subject words of the accepted content using the importance of random forest features, the obtained subject fragments have fragments with high semantic similarity, but it will cause information redundancy. Therefore, DBSCAN clustering based on cos similarity can be used to extract the subject of the fragment. Using DBSCAN clustering based on cos similarity, the subject fragments with high similarity are clustered into one category, and complaint risk themes with strong interpretability, clear semantics, and high conciseness are automatically extracted.

[0029] In some optional embodiments of the present application, before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the text data may be segmented to obtain a phrase sequence; each phrase in the phrase sequence may be tagged with a part of speech to obtain a target phrase corresponding to each part of speech, the parts of speech including noun, verb and adverb; the frequency of occurrence of the target phrase is determined, and a keyword group is determined based on the frequency; a keyword library is constructed based on the keyword group. Specifically, a Chinese word segmentation algorithm, for example, the Jieba word segmentation algorithm, may be used to recombine a continuous sequence of characters into a word sequence according to certain specifications, and the word segmentation process may be performed on the recognizable text content that has been formed.

[0030] Specifically, determining the keyword group according to the frequency can be achieved through the following steps: determining the information entropy of the target phrase according to the frequency, wherein the information entropy is used to measure the amount of information; and taking the target phrase whose information entropy is greater than a first threshold as the keyword group. Through the professional word library construction method based on information entropy, words that cannot be further segmented can be automatically found. It should be noted that information entropy can measure the uncertainty of the left and right adjacent characters of a candidate word. The greater the uncertainty, the more information its adjacent characters contain, and the higher the probability of forming a word. It can be understood that topic extraction can be performed directly based on the constructed keyword library. Specifically, the corresponding texts in the keyword library can be clustered according to the first predetermined clustering algorithm to obtain multiple text sets. It should be noted that different sets correspond to different types, and then for each text set, topic information is extracted from all texts in the text set according to the second predetermined clustering algorithm to obtain the subject words of the text set. It is easy to notice that the first predetermined clustering algorithm can be a K-means K-Means clustering algorithm; the second predetermined clustering algorithm can be a density-based DBSCAN clustering algorithm with noise. Topic extraction is performed directly based on the constructed keyword library, which can greatly reduce the amount of data calculation, save a lot of time, and improve the accuracy of topic identification.

[0031] In some embodiments of the present application, before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the text data can be denoised, and the denoising methods include but are not limited to: removing punctuation marks and numbers; converting English letters to lowercase and extracting the roots of English words.

[0032] In some optional embodiments of the present application, when there are multiple keywords, after obtaining the keywords of the text set, the work order subject can be determined in the following manner, specifically: determining multiple affiliations corresponding to the multiple keywords of the text set; determining a target affiliation that is greater than a second threshold value among the multiple affiliations; using the keyword corresponding to the target affiliation as the target keyword, and using the sentence composed of the target keyword as the work order subject corresponding to the text set.

[0033] It should be noted that the text data of the work order includes but is not limited to: electric heating business complaint data, electric heating business consultation data. It is easy to notice that the above electric heating business complaint data and electric heating business consultation data can come from the 95598 system.

[0034] In some embodiments of the present application, after extracting the subject information of all texts in the text set according to the second predetermined clustering algorithm and obtaining the subject words of the text set, a risk subject library can be generated according to the subject words corresponding to the text set; the frequency of the subject words is determined based on the risk subject library, and the subject words with a frequency greater than the third threshold in the subject words are used as the user's complaint hotspots and / or consultation hotspots. For example, there are complaints about staff members "smoking, untimely response to messages, and poor attitude" in the subject words, where the frequency of complaints-staff members-untimely response to messages is greater than the third threshold, then "untimely response to messages" is used as a complaint hotspot, which can be immediate. The above-mentioned first threshold, second threshold, and third threshold can be set automatically or manually according to the specific application scenario.

[0035] Figure 2 A device for extracting the subject of a work order text according to an embodiment of the present application is provided. Figure 2 As shown, the device for extracting the work order text subject includes:

[0036] An acquisition module 40 is used to acquire text data of a work order;

[0037] A clustering module 42, configured to cluster the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain a plurality of text sets, wherein different sets correspond to different types;

[0038] The extraction module 44 is used for extracting topic information from all the texts in each text set according to a second predetermined clustering algorithm to obtain the topic words of the text set.

[0039] In the device for extracting the topic of the work order text, an acquisition module 40 is used to obtain the text data of the work order; a clustering module 42 is used to cluster the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, wherein different sets correspond to different types; an extraction module 44 is used to extract topic information from all texts in the text set according to a second predetermined clustering algorithm for each text set to obtain the subject words of the text set, thereby achieving the technical effect of first clustering the texts corresponding to the text data using the first predetermined algorithm to obtain multiple text sets, and then identifying the topics corresponding to each text set based on the second predetermined algorithm, thereby solving the technical problem that the related technology using algorithms such as LDA and PLSA cannot accurately extract the topics to be expressed by the electric heating complaint and consultation text data, resulting in a large deviation in the understanding of the topic to be expressed by the customer, and cannot accurately identify the actual needs of the customer.

[0040] It should be noted that the first predetermined clustering algorithm can be a K-means clustering algorithm; the second predetermined clustering algorithm can be a density-based DBSCAN clustering algorithm with noise. It should also be noted that K-Means clustering uses Euclidean distance as a measurement standard and specifies the number of classes ultimately formed in the cluster. It can first initialize the category centers of the specified number of categories and eventually form the specified number of clusters after continuous iterations. Optionally, K-means clustering can use the tf-idf weight of each word in the electric heating business complaint and consultation text data as a sample feature, perform text clustering on the electric heating business complaint and consultation text data, and obtain K-category texts of the subject terms to be extracted.

[0041] In some optional embodiments of the present application, embedded feature selection can be performed through supervised learning algorithms, such as random forest algorithms, to automatically discover which word in the text has a high degree of influence on the classification when learning which category a sentence belongs to. Random forest feature importance is a type of embedded feature selection. Through the importance of random forest features, the subject words of the accepted content can be obtained, but the extracted subject words may be short in length and poor in interpretability. Therefore, it is necessary to extract the context fragments of the corresponding subject words in the original text. Specifically, after obtaining the subject words of the accepted content using the importance of random forest features, the obtained subject fragments have fragments with high semantic similarity, but it will cause information redundancy. Therefore, DBSCAN clustering based on cos similarity can be used to extract the subject of the fragment. Using DBSCAN clustering based on cos similarity, the subject fragments with high similarity are clustered into one category, and complaint risk themes with strong interpretability, clear semantics, and high conciseness are automatically extracted.

[0042] In some optional embodiments of the present application, before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the text data may be segmented to obtain a phrase sequence; each phrase in the phrase sequence may be tagged with a part of speech to obtain a target phrase corresponding to each part of speech, the parts of speech including noun, verb and adverb; the frequency of occurrence of the target phrase is determined, and a keyword group is determined based on the frequency; a keyword library is constructed based on the keyword group. Specifically, a Chinese word segmentation algorithm, for example, the Jieba word segmentation algorithm, may be used to recombine a continuous sequence of characters into a word sequence according to certain specifications, and the word segmentation process may be performed on the recognizable text content that has been formed.

[0043] It can be understood that topic extraction can be performed directly based on the constructed keyword library. Specifically, the corresponding texts in the keyword library can be clustered according to the first predetermined clustering algorithm to obtain multiple text sets. It should be noted that different sets correspond to different types. Then, for each text set, topic information is extracted from all texts in the text set according to the second predetermined clustering algorithm to obtain the subject words of the text set. It is easy to notice that the first predetermined clustering algorithm can be the K-Means clustering algorithm; the second predetermined clustering algorithm can be the density-based DBSCAN clustering algorithm with noise.

[0044] Specifically, determining a keyword group based on frequency can be achieved through the following steps: determining the information entropy of a target phrase based on frequency, wherein the information entropy is used to measure the amount of information; and taking a target phrase whose information entropy is greater than a first threshold as a keyword group. By using a professional vocabulary construction method based on information entropy, words that cannot be further segmented can be automatically discovered. It should be noted that information entropy can measure the uncertainty of the left and right adjacent characters of a candidate word. The greater its uncertainty, the more information its adjacent characters contain, and the higher the probability of its forming a word.

[0045] According to another aspect of an embodiment of the present application, there is also a non-volatile storage medium, which includes a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any method for extracting the work order text subject.

[0046] Specifically, the above storage medium is used to store program instructions for executing the following functions to achieve the following functions:

[0047] Acquire text data of the work order; cluster the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, where different sets correspond to different types; for each text set, extract topic information from all texts in the text set according to a second predetermined clustering algorithm to obtain the subject words of the text set.

[0048] According to another aspect of an embodiment of the present application, a processor is further provided, the processor being used to run a program stored in a memory, wherein when the program is run, any method of extracting a work order text subject is executed.

[0049] Specifically, the processor is used to call the program instructions in the memory to implement the following functions:

[0050] Acquire text data of the work order; cluster the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain multiple text sets, where different sets correspond to different types; for each text set, extract topic information from all texts in the text set according to a second predetermined clustering algorithm to obtain the subject words of the text set.

[0051] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0052] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0053] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0054] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0055] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0056] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the relevant technology or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0057] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for extracting the subject of a work order text, characterized in that: include: Get the text data of the work order; Clustering the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain a plurality of text sets, wherein different sets correspond to different types; For each text set, extracting subject information from all texts in the constructed text set according to a second predetermined clustering algorithm to obtain subject words of the text set; The first predetermined clustering algorithm includes: K-means K-Means clustering algorithm, K-Means clustering uses Euclidean distance as a measurement standard and specifies the number of classes finally formed in the clustering, and the second predetermined clustering algorithm includes: a density-based DBSCAN clustering algorithm with noise; Before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the method includes: Performing word segmentation processing on the text data to obtain a phrase sequence; Performing part-of-speech tagging on each phrase in the phrase sequence to obtain a target phrase corresponding to each part-of-speech, wherein the parts-of-speech include: noun, verb, and adverb; Determine the frequency of occurrence of the target phrase, and determine the keyword phrase according to the frequency; Building a keyword library based on the keyword group; Determining a keyword group according to the frequency includes determining the information entropy of the target phrase according to the frequency, wherein the information entropy is used to measure the amount of information; taking the target phrase whose information entropy is greater than a first threshold as the keyword group; When there are multiple subject words, after obtaining the subject words of the text set, the method further includes: Determining a plurality of membership degrees corresponding to a plurality of subject words of the text collection; Determining a target membership degree among the plurality of membership degrees that is greater than a second threshold; The subject word corresponding to the target membership is used as the target subject word, and the sentence composed of the target subject word is used as the work order subject corresponding to the text set.

2. The method according to claim 1, characterized in that Before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the method further includes: The text data is subjected to denoising processing, wherein the denoising processing method at least includes: removing punctuation marks and numbers; converting English letters into lowercase letters, and extracting roots of English words.

3. The method according to claim 1, characterized in that The text data of the work order includes at least one of the following: electric heating business complaint data and electric heating business consultation data.

4. The method according to claim 3, characterized in that After extracting subject information from all texts in the text set according to a second predetermined clustering algorithm to obtain subject words of the text set, the method further includes: Generate a risk theme library according to the subject words corresponding to the text collection; The frequency of the subject words is determined based on the risk subject library, and the subject words whose frequency is greater than a third threshold value among the subject words are used as complaint hotspots and / or consultation hotspots of users.

5. A device for extracting the subject of a work order text, characterized in that: include: The acquisition module is used to obtain the text data of the work order; A clustering module, used for clustering the texts corresponding to the text data according to a first predetermined clustering algorithm to obtain a plurality of text sets, wherein different sets correspond to different types; An extraction module, for extracting subject information from all texts in each text set according to a second predetermined clustering algorithm to obtain subject words of the text set; The first predetermined clustering algorithm includes: K-Means clustering algorithm, K-Means clustering uses Euclidean distance as a measurement standard and specifies the number of classes finally formed in the clustering, and the second predetermined clustering algorithm includes: a density-based DBSCAN clustering algorithm with noise; Before clustering the texts corresponding to the text data according to the first predetermined clustering algorithm, the method includes: Performing word segmentation processing on the text data to obtain a phrase sequence; Performing part-of-speech tagging on each phrase in the phrase sequence to obtain a target phrase corresponding to each part-of-speech, wherein the parts-of-speech include: noun, verb, and adverb; Determine the frequency of occurrence of the target phrase, and determine the keyword phrase according to the frequency; Building a keyword library based on the keyword group; Determining a keyword group according to the frequency includes determining the information entropy of the target phrase according to the frequency, wherein the information entropy is used to measure the amount of information; taking the target phrase whose information entropy is greater than a first threshold as the keyword group; When there are multiple subject words, after obtaining the subject words of the text set, the method further includes: Determining a plurality of membership degrees corresponding to a plurality of subject words of the text collection; Determining a target membership degree among the plurality of membership degrees that is greater than a second threshold; The subject word corresponding to the target membership is used as the target subject word, and the sentence composed of the target subject word is used as the work order subject corresponding to the text set.

6. A processor, characterized in that: The processor is used to run a program stored in the memory, wherein: When the program is running, the method for extracting the subject of the work order text as described in any one of claims 1 to 4 is executed.