A work order feature classification method and system based on natural language processing

Abstract

A work order feature classification method and system based on natural language processing, the method comprising: in response to a work order processing instruction, determining a business identifier corresponding to a target work order; obtaining a historical post identifier and a current post identifier corresponding to the business identifier; when the historical post identifier and the current post identifier are different, obtaining a common ambiguous feature set commonly corresponding to the current post identifier and the historical post identifier; through a natural language processing algorithm, performing text recognition on target work order data to obtain a plurality of text features; and for each text feature, if the text feature belongs to the common ambiguous feature set, determining at least one adjacent feature corresponding to the text feature; determining an available text meaning corresponding to the text feature according to the text feature and the at least one adjacent feature, and classifying the text feature in the target work order based on the available text meaning. The application improves the accuracy of work order feature recognition by focusing on personnel post flow and adjacent feature correlation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of language processing, and in particular to a work order feature classification method and system based on natural language processing. BACKGROUND

[0002] Work orders carry important information such as customer demands and business handling instructions. In the context of enterprise architecture updates, changes in collaboration patterns between business departments and reorganization of business processes have raised higher requirements for accurate recording and classification of work orders to ensure efficient and accurate information transmission between different positions and functions. Currently, enterprises generally use work order classification systems based on natural language processing technology. These systems extract and identify key words and semantic structures in work order text content through text analysis algorithms to automatically classify work order features. At the same time, when the enterprise architecture is updated, the position information of business personnel is also updated in a timely manner. The system can classify work orders based on the current position of business personnel, such as classifying sales department work orders into the sales category and after-sales department work orders into the after-sales category, so that work order classification is to some extent consistent with business processes, providing some convenience for subsequent business processing.

[0003] With the continuous updating of enterprise architecture and the rapid development of business, the positions of business personnel change frequently. Due to the change of business personnel's position, although the work order classification system can classify based on position information, business personnel may subconsciously use professional terms from previous positions to describe the work content of the current position when recording work orders. However, existing systems often have difficulty accurately identifying and distinguishing such cross-position term mixing situations. This results in the incorrect classification of work order features of the current position into the feature category of the previous position, leading to work order classification errors.

[0004] The prior art (CN114528399A) discloses a work order text classification method, device, storage medium and computer equipment. However, the prior art cannot accurately distinguish between term mixing situations when business personnel record work orders using professional terms from previous positions to describe the work content of the current position after the change of their positions, resulting in low accuracy of work order feature recognition.

[0005] Therefore, how to improve the accuracy of work order feature recognition is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0006] To solve the problems in the prior art, the present application provides a work order feature classification method and system based on natural language processing to solve the technical problem of inaccurate work order feature classification.

[0007] The present application adopts the following technical solutions.

[0008] In a first aspect, the application provides a work order feature classification method based on natural language processing, comprising:

[0009] In response to a work order processing instruction, based on target work order data, determining the business representative identifier corresponding to the target work order, wherein the work order processing instruction represents the start of processing the target work order data included in the target work order;

[0010] Obtain the historical post identifier and the current post identifier corresponding to the business representative identifier; when the historical post identifier and the current post identifier are different, obtain the common ambiguous feature set commonly corresponding to the current post identifier and the historical post identifier, wherein the common ambiguous feature set includes a plurality of common ambiguous features with the same text characters and different text meanings, and the different text meanings represent different text meanings corresponding to the common ambiguous feature in different posts;

[0011] Through a natural language processing algorithm, text recognition is performed on the target work order data to obtain a plurality of text features; and for each text feature, if the text feature belongs to the common ambiguous feature set, at least one adjacent feature corresponding to the text feature is determined;

[0012] According to the text feature and the at least one adjacent feature, determine the available text meaning corresponding to the text feature, and classify the text feature in the target work order based on the available text meaning.

[0013] Optionally, the common ambiguous feature set commonly corresponding to the current post identifier and the historical post identifier is obtained, comprising:

[0014] Determine the current feature identifier set corresponding to the current post identifier and the historical feature identifier set corresponding to the historical post identifier, wherein the current feature identifier set and the historical feature identifier set each include a plurality of corresponding relationships between text characters and text meanings;

[0015] Based on the current feature identifier set and the historical feature identifier set, determine a plurality of initial ambiguous features with the same text characters, and obtain a plurality of text meanings corresponding to each initial ambiguous feature;

[0016] For each initial ambiguous feature, when the plurality of text meanings corresponding to the initial ambiguous feature are different, the initial ambiguous feature is determined to be a common ambiguous feature.

[0017] Optionally, the at least one adjacent feature corresponding to the text feature is determined, comprising:

[0018] determine a natural language sentence corresponding to the text feature according to the plurality of text features, wherein the natural language sentence comprises a plurality of text words, and the text words are the text feature or any adjacent entity;

[0019] combine the text words two by two to obtain at least one entity discrimination group, and obtain the adjacent feature based on the entity discrimination group having an actual meaning.

[0020] Optionally, the determining of the at least one adjacent feature corresponding to the text feature comprises:

[0021] determine a natural language sentence corresponding to the text feature according to the plurality of text features, wherein the natural language sentence comprises a plurality of text words, and the text words are the text feature or any adjacent entity;

[0022] randomly combine the text words in a number M to obtain an entity discrimination group corresponding to the number M; the number M = 3, …, N, and N is a positive integer;

[0023] according to the entity discrimination group corresponding to each number M or number 2, the entity discrimination group having an actual meaning is taken as the adjacent feature.

[0024] Optionally, the determining of the at least one adjacent feature corresponding to the text feature comprises:

[0025] expand the text feature to a preset range before and after the position of the text feature to define a search interval of the adjacent feature, and preliminarily screen out candidate adjacent features having semantic association with the text feature in the search interval based on the grammatical structure, semantic representation and logical relationship in the business scenario of the text;

[0026] use a semantic analysis model to quantitatively evaluate the semantic association between the candidate adjacent feature and the text feature, calculate the collaborative appearance probability between the candidate adjacent feature and the text feature, and determine the candidate adjacent feature as the adjacent feature when the collaborative appearance probability is not less than a preset semantic association strength threshold.

[0027] Optionally, the preliminarily screening out of the candidate adjacent features having semantic association with the text feature in the search interval based on the grammatical structure, semantic representation and logical relationship in the business scenario of the text comprises:

[0028] perform feature extraction on the text feature and the candidate feature text in the search interval respectively to obtain grammatical features and semantic features of the text, the grammatical features are TF-IDF feature vectors, and the semantic features are semantic vector representations of the text obtained by using a pre-trained language model;

[0029] respectively calculate the similarity of the text features and the candidate feature texts in the text syntax structure, semantic representation and business logic, to obtain the syntax structure similarity, semantic representation similarity and business logic similarity;

[0030] weighting and fusing the syntax structure similarity, semantic representation similarity and business logic similarity and their corresponding weights to obtain a comprehensive similarity score of the candidate feature text;

[0031] When the comprehensive similarity score is greater than a preset score threshold, the candidate feature text corresponding to the comprehensive similarity score is confirmed as a candidate adjacent feature.

[0032] Optionally, before determining the current feature identifier set corresponding to the current post identifier, the method further includes:

[0033] obtaining an initial feature identifier set, wherein the initial feature identifier set includes a historical current feature identifier set;

[0034] determining whether there is a new text feature compared with the initial feature identifier set;

[0035] If yes, updating the initial feature identifier set according to the new text feature to obtain the current feature identifier set.

[0036] Optionally, the natural language processing algorithm is based on a word vector model algorithm.

[0037] In a second aspect, the present application provides a work order feature classification system based on natural language processing, comprising:

[0038] A first identification module is configured to determine a business representative identifier corresponding to a target work order based on target work order data in response to a work order processing instruction, wherein the work order processing instruction represents starting to process the target work order data included in the target work order;

[0039] An acquisition module is configured to acquire a historical post identifier and a current post identifier corresponding to the business representative identifier, and acquire a common ambiguous feature set commonly corresponding to the current post identifier and the historical post identifier when the historical post identifier and the current post identifier are different, wherein the common ambiguous feature set includes a plurality of common ambiguous features with the same text characters and different text meanings, and the different text meanings represent different text meanings of the common ambiguous features corresponding to different posts;

[0040] A second identification module is configured to perform text recognition on the target work order data through a natural language processing algorithm to obtain a plurality of text features, and determine at least one adjacent feature corresponding to each text feature if the text feature belongs to the common ambiguous feature set.

[0041] a processing module configured to determine an available text meaning corresponding to the text feature according to the text feature and the at least one adjacent feature, and classify the text feature in the target work order based on the available text meaning.

[0042] Optionally, the obtaining module is further configured to:

[0043] determine a current feature identifier set corresponding to the current post identifier and a historical feature identifier set corresponding to the historical post identifier, the current feature identifier set and the historical feature identifier set each including a plurality of corresponding relationships between text characters and text meanings;

[0044] determine a plurality of initial ambiguous features with identical text characters based on the current feature identifier set and the historical feature identifier set, and obtain a plurality of text meanings corresponding to each of the initial ambiguous features;

[0045] for each of the initial ambiguous features, if the plurality of text meanings corresponding to the initial ambiguous feature are different, determine the initial ambiguous feature as a common ambiguous feature.

[0046] Optionally, the second identifying module is further configured to:

[0047] determine a natural sentence corresponding to the text feature according to the plurality of text features, wherein the natural sentence includes a plurality of text words, and the text words are the text feature or any adjacent entity;

[0048] combine the text words two by two to obtain at least one entity discrimination group, and obtain the adjacent feature based on an entity discrimination group with an actual meaning.

[0049] Optionally, the system further includes:

[0050] an updating module configured to:

[0051] obtain an initial feature identifier set, and the initial feature identifier set includes a historical current feature identifier set;

[0052] determine whether there is a new text feature in the plurality of text features compared with the initial feature identifier set;

[0053] if yes, update the initial feature identifier set according to the new text feature to obtain the current feature identifier set.

[0054] Optionally, the second identifying module is further configured to:

[0055] According to the plurality of text features, a natural language sentence corresponding to the text feature is determined, wherein the natural language sentence includes a plurality of text words, and the text words are the text feature or any adjacent entity;

[0056] The text words are randomly combined in a number M to obtain an entity discrimination group corresponding to the number M; the number M = 3,…, N, and N is a positive integer;

[0057] According to the entity discrimination group corresponding to each number M or number 2, the entity discrimination group with actual meaning is taken as the adjacent feature.

[0058] Optionally, the natural language processing algorithm is based on a word vector model algorithm.

[0059] In a third aspect, the present application provides an electronic device, comprising a processor and a memory connected to the processor in communication;

[0060] The memory stores computer execution instructions;

[0061] The processor executes the computer execution instructions stored in the memory to implement the natural language processing-based work order feature classification method according to any one of the first aspect.

[0062] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by a processor to implement the natural language processing-based work order feature classification method according to any one of the first aspect.

[0063] Compared with the prior art, the present application has at least the following beneficial effects:

[0064] The natural language processing-based work order feature classification method provided by the present application can accurately identify the post change of a business staff by judging whether the historical post identifier and the current post identifier are the same, and determine the common ambiguous feature set corresponding to the current post and the historical post according to the judgment. After the natural language processing algorithm is used to recognize the text of the work order, the text features belonging to the ambiguous feature set are comprehensively judged in combination with their adjacent features, and the classification is performed according to the meaning. This process can not only timely discover the post change and focus on the key ambiguous features, but also accurately classify based on the context and deep understanding of the text, and at the same time, identify other texts when determining the business staff, improve the text analysis efficiency, effectively solve the problem of incorrect classification of work orders caused by the use of old terms due to the post change of business staff after the update of enterprise architecture, and ensure the accuracy and efficiency of work order processing. BRIEF DESCRIPTION OF DRAWINGS

[0065] The accompanying drawings, which are incorporated herein and constitute part of the specification, illustrate embodiments consistent with the application and, together with the description, further serve to explain the principles of the application.

[0066] Figure 1 A flowchart of a work order feature classification method based on natural language processing provided by an embodiment of the application is shown in the figure.

[0067] Figure 2 A structural diagram of a work order feature classification system based on natural language processing provided by an embodiment of the application is shown in the figure.

[0068] Figure 3 A structural diagram of an electronic device provided by an embodiment of the application is shown in the figure.

[0069] The specific embodiments of the application have been shown and described in the above drawings, and will be described in more detail in the following. These drawings and the written description are not intended to limit the scope of the inventive concept in any way, but to illustrate the inventive concept to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0070] The exemplary embodiments will be described in detail herein with reference to the attached drawings. Unless otherwise specified, the same numbers in different drawings indicate the same or similar elements. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the application. Rather, they are merely examples of systems and methods consistent with some aspects of the application as detailed in the appended claims.

[0071] The technical solutions of the application and how the technical solutions of the application solve the above technical problems will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the application will be described below with reference to the drawings.

[0072] As Figure 1 shown, Figure 1 A flowchart of a work order feature classification method based on natural language processing provided by an embodiment of the application is shown in the figure. The execution subject of the embodiment of the application can be an electronic device. Embodiment 1 of the application provides a work order feature classification method based on natural language processing, which can specifically include steps S201 to S204, wherein:

[0073] S201, in response to a work order processing instruction, based on target work order data, determining the business identifier corresponding to the target work order, wherein the work order processing instruction represents starting to process the target work order data included in the target work order.

[0074] The work order processing instruction should explicitly include the unique identifier of the target work order, the business staff identification information, and the storage location of the work order data, and other key information. The business staff identification can be identified from the preset packet header position of the target work order data packet. The packet header adopts a fixed-length field design, and the byte sequence of the identifier field strictly follows the network byte sequence specification. The identification method can be to parse the business staff identification field in the packet header of the target work order data packet, decode the byte sequence of the business staff identification field according to the preset identifier coding rule, and convert it into a business staff identification format recognizable by the system, to ensure accurate extraction of the business staff information.

[0075] S202, obtaining the historical post identifier and the current post identifier corresponding to the business staff identification determined in S201; when the historical post identifier and the current post identifier are different, obtaining a common ambiguity feature set corresponding to the current post identifier and the historical post identifier, wherein the common ambiguity feature set includes a plurality of common ambiguity features with the same text characters and different text meanings, and the different text meanings represent different text meanings corresponding to the common ambiguity feature when the common ambiguity feature corresponds to different posts.

[0076] The acquisition of the historical post identifier and the current post identifier requires accessing the electronic archives of the business staff in the enterprise human resource management system or the business operation record database to extract the identification information. The electronic archives should include the complete career records of the business staff to ensure that the historical and current post identifiers can be accurately obtained.

[0077] In addition, the text characters refer to specific characters that constitute text words, including letters, numbers, Chinese characters, and special symbols as basic units; the text meaning refers to the meaning of the text characters in a specific context, for example, "IT" in the technical research and development post means "information technology", involving software development, etc.; in the customer service post, it means "information technology support", focusing on solving customer technical problems. For example, "legal affairs", in the legal compliance department, means "legal affairs processing", covering contract review, etc.; in the sales department, it means "legal support", mainly involving sales compliance review of contract terms.

[0078] In one implementation manner, the obtaining of the common ambiguity feature set corresponding to the current post identifier and the historical post identifier includes:

[0079] determining a current feature identifier set corresponding to the current post identifier and a historical feature identifier set corresponding to the historical post identifier, the current feature identifier set and the historical feature identifier set each including a plurality of corresponding relationships between text characters and text meanings;

[0080] based on the current feature identifier set and the historical feature identifier set, determining a plurality of initial ambiguity features with the same text characters, and obtaining a plurality of text meanings corresponding to each of the initial ambiguity features.

[0081] For each of the initial ambiguous features, if there is a difference in the text meanings of the initial ambiguous features corresponding to the several text meanings, it is determined that the initial ambiguous feature is a common ambiguous feature.

[0082] For example, in a certain enterprise, it is assumed that a business personnel has served in a "software development" post and then transferred to an "IT support" post. In the current feature identifier set corresponding to the current post "IT support", "code" refers to "program instructions", and "debug" refers to "troubleshooting equipment or system failures"; in the historical feature identifier set corresponding to the historical post "software development", "code" refers to "software programming language instructions", and "debug" refers to "testing and correcting software". By comparing the two feature identifier sets, it is found that the two text characters "code" and "debug" exist in both posts but have different text meanings, so they are determined to be common ambiguous features.

[0083] In an implementable manner, before the determination of the current feature identifier set corresponding to the current post identifier, the method further comprises:

[0084] obtaining an initial feature identifier set, wherein the initial feature identifier set comprises the historical current feature identifier set;

[0085] determining whether there is a new text feature compared to the initial feature identifier set;

[0086] If yes, updating the initial feature identifier set according to the new text feature to obtain the current feature identifier set.

[0087] The method for classifying work order features based on natural language processing provided by the embodiments of the present application can update the feature identifier set when a new undefined text appears. In this way, by dynamically updating the feature identifier set, the system can incorporate the new text feature into the feature identifier set without manual annotation, which is used for subsequent determination of the common ambiguous feature set, thereby improving the accuracy and efficiency of work order feature classification.

[0088] S203, performing text recognition on the target work order data by a natural language processing algorithm to obtain a plurality of text features; and for each of the text features, if the text feature belongs to the common ambiguous feature set obtained in S202, determining at least one adjacent feature corresponding to the text feature.

[0089] The natural language processing algorithm is based on a word vector model algorithm. The target work order data is input into the word vector model for recognition to obtain a plurality of text features.

[0090] It should be noted that the identification process of S201 in the embodiment and the identification process of S203 can be performed simultaneously. On the one hand, the efficiency of text analysis can be ensured, and on the other hand, the accuracy of text analysis can be improved.

[0091] It can be understood that based on the linear sequence characteristics of the text, the search range of the adjacent features is defined by expanding the position of the text feature to the front and back text content within a reasonable range.

[0092] Generally, A and B are positive integers determined in advance according to the semantic coherence of the text and experimental experience. In this way, the range in which the adjacent features may appear is framed, ensuring that both the content closely related to the text feature in terms of semantics and the irrelevant information introduced by excessively expanding the search range can be avoided.

[0093] In the determined search range, the initially framed adjacent text features are examined one by one. According to the multi-dimensional standards of the text's grammatical structure, semantic representation, and logical relationship in the business scenario, the candidate adjacent features that may have semantic association with the text feature are preliminarily screened out to eliminate those features that have no obvious semantic association or very weak semantic association with the text feature, thereby forming a more focused list of candidate adjacent features.

[0094] In one possible implementation, the above-mentioned multi-dimensional standards of the text's grammatical structure, semantic representation, and logical relationship in the business scenario, which are used to preliminarily screen out the candidate adjacent features that may have semantic association with the text feature, include:

[0095] Feature extraction is performed on the target text and the candidate feature text to be analyzed respectively to obtain the grammatical features and semantic features of the text.

[0096] The grammatical features are TF-IDF (Term Frequency-Inverse Document Frequency) feature vectors, which can be obtained by performing word segmentation, part-of-speech tagging, and core word extraction on the text. The semantic features can use a pre-trained language model to obtain the semantic vector representation of the text.

[0097] The similarity of the target text and the candidate feature text in the three dimensions of grammatical structure, semantic representation, and business logic is calculated respectively.

[0098] The syntax structure similarity is core word overlap and TF-IDF vector cosine similarity. The core word overlap is the ratio of the number of core words common to the target text and the candidate feature text to the total number of core words in the target text and the candidate feature text. The semantic representation similarity is the cosine similarity of semantic vectors. The business logic similarity is the correlation between business keywords and business processes. The correlation can be the coincidence frequency between the business keywords and the preset keywords included in the business processes.

[0099] According to the preset weight parameter, the similarity scores of the three dimensions are weighted and fused to obtain a comprehensive similarity score between the texts. The comprehensive similarity score is compared with a preset score threshold to screen out candidate features with a score higher than the preset score threshold as adjacent features that may have semantic association with the target text.

[0100] Further, for the candidate adjacent features, a semantic analysis model in a natural language processing algorithm, such as a word vector model or a pre-trained language model, is used to quantitatively evaluate the semantic association between the candidate adjacent features and the text features. By calculating the probability index of the co-occurrence of the two, it is verified whether it meets a preset semantic association strength threshold. Specifically, the co-occurrence probability can be obtained by inputting the candidate adjacent features and the text features into a co-occurrence statistical model. After obtaining the co-occurrence probability, the co-occurrence probability is compared with the preset semantic association strength threshold. If it is not less than the preset semantic association strength threshold, it is determined to meet the threshold requirement. For the candidate adjacent features that meet the threshold requirement, the adjacent features of the text features are formally determined. For those that do not meet the requirement, they are excluded, so as to ensure that the determined adjacent features can truly reflect the semantic association of the text features in the context and provide a reliable basis for accurately determining the available text meaning of the text features.

[0101] In one possible case, in the text processing process, it is usually necessary to position mark the words to identify adjacent features. Generally, each word is assigned a position identifier, and the words corresponding to adjacent position identifiers are considered to be adjacent. However, in actual application, there is a special case where adjacent entities are not equivalent to adjacent features. For example, in the text "apple pear a box", the adjacent entities include "apple", "pear", "apple pear", and "a box", and the real adjacent features should be "apple pear" and "a box". At this time, if the simple judgment according to the position identifier is still used, the adjacent feature may be misjudged, leading to deviation of text semantic understanding and affecting the accuracy of subsequent text analysis. Therefore, a more accurate method is needed to determine adjacent features to avoid such misjudgment.

[0102] In one possible implementation, the determining of the at least one adjacent feature corresponding to the text feature includes:

[0103] determine a natural language sentence corresponding to the text feature according to the plurality of text features, wherein the natural language sentence comprises a plurality of text words, and the text words are the text feature or any adjacent entity;

[0104] combine the text words two by two to obtain at least one entity discrimination group, and obtain the adjacent feature based on the entity discrimination group having an actual meaning.

[0105] For example, there is a natural language sentence "Today the weather is rainy, and I perform maintenance on device A", wherein the text features include "today", "weather", "rainy", "I", "to", "device A", and "maintenance". According to the text features, it is determined that the natural language sentence corresponding to the text features is "Today the weather is rainy, and I perform maintenance on device A", wherein the natural language sentence comprises a plurality of text words, which can be the text features themselves, such as "today", "weather", and the like, or adjacent entities, such as "device A rain" and "maintenance". Then, the text words are combined two by two to obtain a plurality of entity discrimination groups, such as "today weather", "weather is rainy", "rainy I", "I to device A", "to device A perform maintenance", and the like. Then, the entity discrimination groups are screened based on whether there is an actual meaning. For example, "weather is rainy" indicates weather conditions, and "to device A perform maintenance" indicates an action object and an action type, and these combinations have an actual meaning, so they are determined as adjacent features. However, a combination such as "rainy I" has no actual meaning and is excluded. Finally, the adjacent features "weather is rainy" and "to device A perform maintenance" are obtained, which can more accurately reflect the semantic information of the text and help subsequent text analysis and processing.

[0106] In another possible implementation, the determining of the at least one adjacent feature corresponding to the text feature comprises:

[0107] determine a natural language sentence corresponding to the text feature according to the plurality of text features, wherein the natural language sentence comprises a plurality of text words, and the text words are the text feature or any adjacent entity;

[0108] randomly combine the text words in a number M to obtain an entity discrimination group corresponding to the number M; and the number M = 3, …, N, and N is a positive integer.

[0109] determine the entity discrimination group having an actual meaning as the adjacent feature according to the entity discrimination group corresponding to each number M or number 2.

[0110] The method for classifying work order features based on natural language processing provided by the embodiments of the present application can directly and quickly exhaust the adjacent combination possibilities between text words in a two-by-two combination manner, is particularly suitable for short texts or scenarios with high real-time requirements, can give results in a short time, and is suitable for preliminary text feature mining; adjacent words combined in pairs can directly reflect the close semantic correlation between words, which is helpful for identifying common phrase structures and has natural advantages for identifying fixed collocations and common phrases, and can effectively capture basic semantic units in the text. The combination manner with a random number can flexibly adapt to text analysis requirements of different lengths and complexities. When M takes different values, different levels of semantic information can be mined, from the phrase level to the paragraph level, which has stronger adaptability and expansibility and can more comprehensively capture text features.

[0111] S204, determining the available text meaning corresponding to the text feature according to the text feature and the at least one adjacent feature obtained in S203, and classifying the text feature in the target work order based on the available text meaning.

[0112] It should be noted that the above technical solution determines whether the job personnel has recently changed positions, if there is a position change, the historical position information and the current position information of the job personnel are obtained, and a common ambiguous word set shared by the historical position and the current position is identified and determined. Subsequently, the word recognition program is executed for the current work order text, each word in the text is compared with the aforementioned ambiguous word set one by one, if the text word belongs to the ambiguous word set, at least one adjacent word of the ambiguous word in the text is further determined, and the ambiguous word and its adjacent words are comprehensively considered to accurately determine to which position the ambiguous word belongs to.

[0113] It should be noted that for the above-mentioned method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0114] It is further noted that, while the steps of the flow diagram are shown in a particular order, the steps are not necessarily performed in the order shown by the arrows. Unless otherwise explicitly stated herein, the steps are not necessarily performed in the order shown by the arrows. The steps can be performed in other orders. Moreover, at least some of the steps of the flow diagram can include multiple sub-steps or stages, which are not necessarily performed at the same time, but can be performed at different times, and which are not necessarily performed sequentially, but can be performed in rotation or alternation with other steps or sub-steps or stages of other steps.

[0115] Figure 2 A structural schematic diagram of a work order feature classification system based on natural language processing provided by an embodiment of the present application is shown in FIG. 2. As shown in FIG. 2, the embodiment 2 of the present application provides a work order feature classification system 40 based on natural language processing, which includes: Figure 2

[0116] A first identification module 401 is configured to determine a business representative identifier corresponding to a target work order based on target work order data in response to a work order processing instruction, wherein the work order processing instruction represents starting to process the target work order data included in the target work order.

[0117] An acquisition module 402 is configured to acquire a historical post identifier and a current post identifier corresponding to the business representative identifier, and acquire a common ambiguity feature set commonly corresponding to the current post identifier and the historical post identifier when the historical post identifier and the current post identifier are different, wherein the common ambiguity feature set includes a plurality of common ambiguity features with the same text characters and different text meanings, and the different text meanings represent that the common ambiguity feature corresponds to different post text meanings.

[0118] A second identification module 403 is configured to perform text identification on the target work order data by a natural language processing algorithm to obtain a plurality of text features, and determine at least one adjacent feature corresponding to each text feature if the text feature belongs to the common ambiguity feature set.

[0119] A processing module 404 is configured to determine a usable text meaning corresponding to the text feature according to the text feature and the at least one adjacent feature, and classify the text feature in the target work order based on the usable text meaning.

[0120] Optionally, the acquisition module 402, when acquiring the common ambiguity feature set commonly corresponding to the current post identifier and the historical post identifier, is configured to:

[0121] ​determine a current feature identifier set corresponding to the current post identifier and a historical feature identifier set corresponding to the historical post identifier, the current feature identifier set and the historical feature identifier set each including a plurality of corresponding relationships between text characters and text meanings;

[0122] based on the current feature identifier set and the historical feature identifier set, determine a plurality of initial ambiguous features with the same text characters, and obtain a plurality of text meanings corresponding to each of the initial ambiguous features;

[0123] for each of the initial ambiguous features, if the plurality of text meanings corresponding to the initial ambiguous feature are different, determine that the initial ambiguous feature is a common ambiguous feature.

[0124] Optionally, the second identification module 403, in the execution of determining the at least one adjacent feature corresponding to the text feature, is configured to:

[0125] determine a natural sentence corresponding to the text feature according to the plurality of text features, wherein the natural sentence includes a plurality of text words, and the text words are the text feature or any adjacent entity;

[0126] combine the text words two by two to obtain at least one entity discrimination group, and obtain the adjacent feature based on the entity discrimination group with actual meaning.

[0127] Optionally, the system 40 further includes:

[0128] an updating module configured to:

[0129] obtain an initial feature identifier set, wherein the initial feature identifier set includes a historical current feature identifier set;

[0130] determine whether there is a new text feature in the plurality of text features compared with the initial feature identifier set;

[0131] if yes, update the initial feature identifier set according to the new text feature to obtain the current feature identifier set.

[0132] Optionally, the second identification module 403, in the execution of determining the at least one adjacent feature corresponding to the text feature, is configured to:

[0133] determine a natural sentence corresponding to the text feature according to the plurality of text features, wherein the natural sentence includes a plurality of text words, and the text words are the text feature or any adjacent entity;

[0134] randomly combine the text words in a number M to obtain an entity discrimination group corresponding to the number M; the number M = 3,…, N, and N is a positive integer;

[0135] According to the entity discrimination group corresponding to each quantity M or quantity 2, the entity discrimination group having an actual meaning is taken as the adjacent feature.

[0136] Optionally, the natural language processing algorithm is based on a word vector model algorithm.

[0137] The natural language processing-based work order feature classification system provided in this embodiment can execute the method provided in the method embodiment, and has similar implementation principles and technical effects. Details are not described herein again.

[0138] It should be understood that the system embodiments described above are only illustrative, and the system of the present application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and another division mode can be used in actual implementation. For example, multiple units / modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0139] In addition, unless otherwise specified, each functional unit / module in each embodiment of the present application can be integrated in one unit / module, or each unit / module can exist physically, or two or more units / modules can be integrated together. The integrated unit / module can be implemented in the form of hardware or software program module.

[0140] If the integrated unit / module is implemented in the form of hardware, the hardware can be a digital circuit, an analog circuit, etc. The physical implementation of the hardware structure includes but is not limited to transistors, memristors, etc.

[0141] Figure 3 The structural schematic diagram of the electronic device provided in the present application is shown in FIG. 3. As shown in FIG. 3, the electronic device 50 provided in this embodiment 3 includes at least one processor 501 and a memory 502. Optionally, the electronic device 50 further includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected through a bus. Figure 3

[0142] In the specific implementation process, the at least one processor 501 executes the computer execution instructions stored in the memory 502, so that the at least one processor 501 executes the natural language processing-based work order feature classification method described in embodiment 1.

[0143] The specific implementation process of the processor 501 can refer to the above method embodiments, which have similar implementation principles and technical effects. Details are not described herein again.

[0144] ​Unless specifically stated, the processor 501 can be any suitable hardware processor such as a CPU, a GPU, a FPGA, a DSP, an ASIC, and so on. Unless specifically stated, the memory 502 can be any suitable magnetic or magneto-optical storage medium such as, for example, Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random-Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), and so on.

[0145] The integrated units / modules, if implemented in the form of software program modules and sold or used as independent products, can be stored in a computer readable memory. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a number of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0146] Embodiment 4 of the present application provides a computer readable storage medium, the computer readable storage medium stores computer execution instructions, when the processor executes the computer execution instructions, the method for classifying work order features based on natural language processing according to embodiment 1 is realized.

[0147] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments. The technical features of the above embodiments can be combined arbitrarily, in order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combination of these technical features does not exist contradictory, it should be considered as the scope of the present application.

[0148] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0149] It is to be understood that the application is not limited to the precise construction herein disclosed and shown in the drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is limited only by the claims that follow.

Claims

1. A work order feature classification method based on natural language processing, characterized in that, include: In response to a work order processing instruction, the salesperson identifier corresponding to the target work order is determined based on the target work order data, wherein the work order processing instruction indicates that the processing of the target work order data included in the target work order has begun; Obtain the historical job identifier and current job identifier corresponding to the salesperson identifier; when the historical job identifier and the current job identifier are different, obtain the common ambiguous feature set that is commonly associated with the current job identifier and the historical job identifier, wherein the common ambiguous feature set includes multiple common ambiguous features with the same text characters but different text meanings, and the different text meanings indicate that the text meanings of the common ambiguous features are different when they correspond to different jobs; The target work order data is subjected to text recognition using a natural language processing algorithm to obtain multiple text features; and for each text feature, if the text feature belongs to the common ambiguous feature set, at least one adjacent feature corresponding to the text feature is determined. Based on the text feature and the at least one adjacent feature, determine the available text meaning corresponding to the text feature, so as to classify the text feature in the target work order based on the available text meaning; Determining at least one adjacent feature corresponding to the text feature includes: Centered on the location of the text feature, the search range of adjacent features is defined by expanding outwards and in front of it within a predetermined range. Within the search range, candidate adjacent features that are semantically related to the text feature are initially screened based on the grammatical structure, semantic representation, and logical relationship in the business scenario. Using a semantic analysis model, the semantic correlation between candidate adjacent features and the text feature is quantitatively evaluated. The probability of co-occurrence between candidate adjacent features and the text feature is calculated. When the probability of co-occurrence is not less than a preset semantic correlation strength threshold, it is determined to be an adjacent feature.

2. The work order feature classification method based on natural language processing according to claim 1, characterized in that: The common ambiguous feature set obtained by combining the current job identifier and the historical job identifier includes: Determine the current feature identifier set corresponding to the current job identifier and the historical feature identifier set corresponding to the historical job identifier. Both the current feature identifier set and the historical feature identifier set include the correspondence between multiple text characters and text meanings. Based on the current feature identifier set and the historical feature identifier set, several initial ambiguous features with the same text characters are determined, and several text meanings corresponding to each initial ambiguous feature are obtained; For each of the initial ambiguity features, if the meanings of several texts corresponding to the initial ambiguity feature are different, the initial ambiguity feature is determined to be a common ambiguity feature.

3. The work order feature classification method based on natural language processing according to claim 1, characterized in that: The preliminary screening of candidate adjacent features that are semantically related to the text features within the search interval based on the text's syntactic structure, semantic representation, and logical relationships in the business scenario includes: Feature extraction is performed on the text features and the candidate feature texts within the search range to obtain the syntactic features and semantic features of the text. The syntactic features are TF-ID feature vectors, and the semantic features are semantic vector representations of the text obtained using a pre-trained language model. Calculate the similarity between the text feature and the candidate feature text in terms of syntactic structure, semantic representation, and logical relationship in the business scenario, respectively, to obtain syntactic structure similarity, semantic representation similarity, and business logic similarity; The similarity scores of syntactic structure, semantic representation, and business logic, along with their corresponding weights, are weighted and fused to obtain the comprehensive similarity score of the candidate feature texts. When the overall similarity score is greater than the preset score threshold, the candidate feature text corresponding to the overall similarity score is confirmed as a candidate adjacent feature.

4. The work order feature classification method based on natural language processing according to claim 2, characterized in that: Before determining the current feature identifier set corresponding to the current job identifier, the method further includes: Obtain an initial feature identifier set, which includes the historical current feature identifier set; Determine whether there are any new text features compared to the initial feature identifier set; If so, the initial feature identifier set is updated according to the new text features to obtain the current feature identifier set.

5. The work order feature classification method based on natural language processing according to claim 1, characterized in that, The natural language processing algorithm is based on a word vector model algorithm.

6. A work order feature classification system based on natural language processing, utilizing the work order feature classification method according to any one of claims 1 to 5, characterized in that, The system includes: The first identification module is used to respond to a work order processing instruction and determine the salesperson identifier corresponding to the target work order based on the target work order data, wherein the work order processing instruction indicates that the processing of the target work order data included in the target work order has begun; The acquisition module is used to acquire the historical job identifier and the current job identifier corresponding to the salesperson identifier; when the historical job identifier and the current job identifier are different, the module acquires the common ambiguous feature set that is common to both the current job identifier and the historical job identifier. The common ambiguous feature set includes multiple common ambiguous features with the same text characters but different text meanings. The different text meanings indicate that the text meanings of the common ambiguous features are different when they correspond to different jobs. The second recognition module is used to perform text recognition on the target work order data using a natural language processing algorithm to obtain multiple text features; and for each text feature, if the text feature belongs to the common ambiguous feature set, then determine at least one adjacent feature corresponding to the text feature. The processing module is used to determine the available text meaning corresponding to the text feature based on the text feature and the at least one adjacent feature, so as to classify the text feature in the target work order based on the available text meaning.

7. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the work order feature classification method based on natural language processing as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the work order feature classification method based on natural language processing as described in any one of claims 1 to 5.

Images