CN112434501A - Work order intelligent generation method and device, electronic equipment and medium - Google Patents
Work order intelligent generation method and device, electronic equipment and medium Download PDFInfo
- Publication number
- CN112434501A CN112434501A CN202011150090.4A CN202011150090A CN112434501A CN 112434501 A CN112434501 A CN 112434501A CN 202011150090 A CN202011150090 A CN 202011150090A CN 112434501 A CN112434501 A CN 112434501A
- Authority
- CN
- China
- Prior art keywords
- work order
- service
- data
- label model
- classification label
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000008569 process Effects 0.000 claims description 25
- 238000012545 processing Methods 0.000 claims description 19
- 238000005516 engineering process Methods 0.000 abstract description 9
- 230000000875 corresponding effect Effects 0.000 description 64
- 238000010586 diagram Methods 0.000 description 7
- 238000013145 classification model Methods 0.000 description 6
- 230000006854 communication Effects 0.000 description 4
- 239000013604 expression vector Substances 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000013526 transfer learning Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a method and a device for intelligently generating a work order, electronic equipment and a medium. By applying the technical scheme of the application, a first classification label model and a second classification label model for judging the service category of the session and a work order information label model for filling the specific content of the work order can be obtained according to the dialogue text data training of the customer service and the user in the historical time period. And further, the problem that in the related technology, in each business session between customer service and a user, the customer service staff is required to manually search for a proper template from the template library, so that the working efficiency is reduced is solved.
Description
Technical Field
The present application relates to data processing technologies, and in particular, to a method, an apparatus, an electronic device, and a medium for intelligently generating a work order.
Background
Due to the rise of the communication era and society, users often use various services on service platforms through intelligent terminals.
Further, it is a trend to use a telephone to contact a customer service to obtain corresponding services. The customer service lines generally exist in various industries, the user service lines generally provide services such as business query, recharging and payment, business transaction, information subscription and the like for users, and the quality of service of the customer service lines represents the enterprise image. In order to ensure high quality of service, after-sales service of many enterprises is implemented 24 hours all day by 24 hours of manual service online. The customer service staff needs to record the personal information and the user appeal while answering the user call, and manually loads the work order template to fill in the work order after the call is finished. Sometimes, the customer service personnel still need to listen to the telephone for recording to complete the work order filling because the customer service personnel forget the important information.
In the related technology, dozens of hundreds of work order templates are required to be preset in a work order system used by customer service, and customer service personnel need to browse and find the most appropriate work order template in a system catalog while answering a call. However, such a processing method usually requires a great deal of effort and time for the customer service staff, which reduces the work efficiency.
Disclosure of Invention
The embodiment of the application provides a method, a device, electronic equipment and a medium for intelligently generating a work order, and is used for solving the problem that in the related technology, in each business session between customer service and a user, the customer service staff is required to manually search a proper template from a template library, so that the work efficiency is reduced.
According to an aspect of an embodiment of the present application, a method for intelligently generating a work order is provided, which includes:
acquiring first conversation text data and corresponding first work order data generated by customer service and a user in a service conversation process in a historical time period;
training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data;
when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model;
and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation.
Optionally, in another embodiment based on the foregoing method of the present application, after the obtaining the first dialogue text data and the corresponding first work order data generated by the customer and the user during the business dialogue in the historical time period, the method further includes:
performing data association on the first dialogue text data and the corresponding first work order data to obtain associated historical data;
and storing the associated historical data into a comma separated value CSV file through a unique identification bit.
Optionally, in another embodiment based on the foregoing method of the present application, after the storing the associated history data in a comma separated value CSV file by using a unique identification bit, the method further includes:
dividing the associated historical data into user text data and customer service text data;
respectively obtaining the maximum word numerical values of the user text data and the customer service text number, and performing preset processing on the sentence patterns of which the word numerical values do not meet the maximum word numerical values in the same sentence pattern, wherein the preset processing comprises any one of zero padding completion or interception processing to obtain a first processed text;
acquiring a maximum sentence value of the number of the customer service texts, and performing preset processing on the texts of which the sentence values do not meet the maximum sentence value in the first processed text to obtain a second processed text;
and taking the second processed text as sample training data.
Optionally, in another embodiment based on the foregoing method of the present application, after the combining the pattern history data and the sentence history data as sample training data, the method further includes:
acquiring the sample training data;
and training a first-level attention model initialized randomly by using the sample training data to obtain the first classification label model meeting preset conditions, wherein the first classification label model is used for determining the service type of the dialogue text data.
Optionally, in another embodiment based on the foregoing method of the present application, after the first classification label model meeting a preset condition, the method further includes:
obtaining model parameters of a cleared output layer in the first classification label model and the sample training data;
and training a randomly initialized second-level attention model by using the model parameters and the sample training data to obtain a second classification label model meeting preset conditions, wherein the second classification label model is used for determining sub-service labels under the service type of the dialogue text data.
Optionally, in another embodiment based on the foregoing method of the present application, after obtaining the second classification label model that meets a preset condition, the method further includes:
extracting service categories corresponding to a third amount of work order information from the associated historical data, and respectively establishing corresponding category labels;
removing the output layers of the first classification label model, and after the hidden variables output by the second classification label model, performing class label judgment on the state of each work order information aiming at the first number of output layers to obtain a corresponding judgment result;
and generating the work order information label model based on the judgment result and the second classification label model of the removal output layer.
Optionally, in another embodiment based on the foregoing method of the present application, the determining, based on the first classification label model and the second classification label model, a work order template corresponding to the current service session includes:
inputting the second dialogue text data into the first classification label model to obtain a main business label corresponding to the second dialogue text data;
inputting the service class label and the second dialogue text data into the second class label model to obtain a sub-service label corresponding to the second dialogue text data;
and selecting a target work order template matched with the main service label and the sub service labels from a plurality of pre-stored work order templates of different service labels.
Optionally, in another embodiment based on the foregoing method of the present application, the main service label includes at least one of: fault service, consultation service, complaint service and service handling service; and the number of the first and second groups,
the sub-service label includes at least one of: fee service, personnel service, website service, equipment service, network service.
According to another aspect of the embodiments of the present application, there is provided an apparatus for intelligently generating a work order, including:
the acquisition module is configured to acquire first conversation text data and corresponding first work order data generated by customer service and a user in a business conversation process in a historical time period;
the generating module is configured to train to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data;
the determining module is configured to determine a target work order template corresponding to the business session based on the first classification label model and the second classification label model when second conversation text data generated by the customer service and the user in the business session process is acquired;
and the filling module is configured to fill the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation.
According to another aspect of the embodiments of the present application, there is provided an electronic device including:
a memory for storing executable instructions; and
and the display is used for displaying with the memory to execute the executable instructions so as to complete the operation of any one of the intelligent work order generation methods.
According to yet another aspect of the embodiments of the present application, a computer-readable storage medium is provided for storing computer-readable instructions, which when executed perform the operations of any one of the above-mentioned methods for intelligent work order generation.
In the application, first conversation text data and corresponding first work order data generated by a customer service and a user in a business conversation process in a historical time period can be obtained; training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data; when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model; and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation. By applying the technical scheme of the application, the first classification label model and the second classification label model for judging the service category of the session and the work order information label model for filling the specific content of the work order can be obtained according to the dialogue text data training of the customer service and the user in the historical time period. And further, the problem that in the related technology, in each business session between customer service and a user, the customer service staff is required to manually search for a proper template from the template library, so that the working efficiency is reduced is solved.
The technical solution of the present application is further described in detail by the accompanying drawings and examples.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description, serve to explain the principles of the application.
The present application may be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:
FIG. 1 is a schematic diagram of the intelligent generation of work orders as set forth in the present application;
FIG. 2 is an architecture diagram of a first category label model and a second category label model proposed in the present application;
FIG. 3 is an architecture diagram of a work order information tag model as presented in the present application;
FIG. 4 is a schematic structural diagram of an electronic device for intelligently generating a work order according to the present application;
fig. 5 is a schematic view of an electronic device according to the present application.
Detailed Description
Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that: the relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise.
Meanwhile, it should be understood that the sizes of the respective portions shown in the drawings are not drawn in an actual proportional relationship for the convenience of description.
The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the application, its application, or uses.
Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.
In addition, technical solutions between the various embodiments of the present application may be combined with each other, but it must be based on the realization of the technical solutions by a person skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination of technical solutions should be considered to be absent and not within the protection scope of the present application.
It should be noted that all the directional indicators (such as upper, lower, left, right, front and rear … …) in the embodiment of the present application are only used to explain the relative position relationship between the components, the motion situation, etc. in a specific posture (as shown in the drawings), and if the specific posture is changed, the directional indicator is changed accordingly.
A method for intelligent generation of work orders according to an exemplary embodiment of the present application is described below in conjunction with fig. 1-3. It should be noted that the following application scenarios are merely illustrated for the convenience of understanding the spirit and principles of the present application, and the embodiments of the present application are not limited in this respect. Rather, embodiments of the present application may be applied to any scenario where applicable.
The application also provides a method, a device, a target terminal and a medium for intelligently generating the work order.
Fig. 1 schematically shows a flowchart of a method for intelligently generating a work order according to an embodiment of the present application. As shown in fig. 1, the method includes:
s101, first conversation text data and corresponding first work order data generated by customer service and a user in a business conversation process in a historical time period are obtained.
First, the historical time period is not specifically limited in the present application, and may be, for example, one week, one month, or the like. The first dialogue text data can be text data generated by voice data conversion in the communication process of the user and the customer service. And the first work order data is the corresponding work order data manually selected by the customer service in the previous communication process.
Further, in the process of converting the voice data of the user and the customer service in the communication process into the text data, a voice recognition technology can be firstly performed to convert the vocabulary contents in the human voice into computer readable input, such as keys, binary codes or character sequences, and the like. The field that the speech recognition technology relates to includes wherein: signal processing, pattern recognition, probability and information theory, sound and hearing mechanisms, artificial intelligence, and the like.
Still further, the data obtained by speech recognition may be subjected to character recognition by Natural speech Processing (NLP). Among them, NLP is an important branch in the computer field and the artificial intelligence field. Due to the fact that data are greatly enhanced, computing power is greatly improved, and deep learning is achieved end-to-end training.
S102, training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data.
Further, the first classification label model in the present application may be a model for distinguishing the service types in the session process, for example, the first classification label model is generally classified into several categories, such as failure, consultation, complaint, and service handling, and the first class of the work order is extracted as a label of the first class classification model of the work order.
In addition, the second classification label model in the present application may take the first class of service as an example for consultation, and the corresponding work order template may be further divided into: and extracting the secondary work order template types as labels of a secondary classification model.
Moreover, the work order information label model in the present application may be a model for filling work order information in a work order template. The work order information may be information to be filled in the work order template, and may be, for example, user information, a service address, a work order serial number, and the like.
Further, taking the work order template with the service type as the "information query" type as an example, the user can query the desired information by informing any one of the address, the username, and the identification number of the user. And after the user is confirmed to inquire the archive information through the identity card number, the identity card information of the user is input to the corresponding work order information position in the work order template.
In one embodiment, the label model is not specifically limited, and may be, for example, a convolutional neural network model or a Hierarchical Attention network model (HAN).
Among them, the Hierarchical Attention Network (HAN) is a kind of neural network for document classification. The model has two distinct features: i.e. having a hierarchical structure (words forming sentences, which form documents), which may reflect the hierarchical structure of the document, the document representation may be constructed by first constructing a representation of the sentences and then aggregating them into a document representation. In addition, two levels of attention mechanisms may also be applied at the word and sentence level, enabling it to participate in increasingly important content separately in building the document representation.
S103, when second conversation text data generated by the customer service and the user in the business conversation process is collected, a target work order template corresponding to the business conversation is determined based on the first classification label model and the second classification label model.
Further, in the embodiment of the present application, the preprocessed dialog text data may be input into the first classification tag model and the second classification tag model, and then, according to the selection output by the two classification tag models, the target work order template corresponding to the current service session is selected from the preset multiple work order templates.
For example, after the preprocessed first conversation text data is input into the first classification label model and the second classification label model, the business conversation corresponding to the business conversation is obtained as the fee inquiry business, and then the work order template corresponding to the fee inquiry is selected from the plurality of work order templates to be used as the target work order model.
And S104, filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation.
Further, after the target work order template corresponding to the current service session is determined, the work order information extraction of the current service session can be implemented, specifically, the second preprocessed dialogue text data is input into the multi-task learning-based work order information label model to realize the work order information extraction, and then the extracted work order information is filled into the target work order template initialized at random.
That is to say, in the embodiment of the present application, the target work order template initialized at random can be obtained through the primary classification and the secondary classification of the dialog text. And then, a corresponding work order content template is established in advance for each service type, and a work order information slot is reserved, so that the work order information extracted from the second dialogue text is filled into the corresponding slot at this time, and the intelligent generation of the work order content is completed.
In the application, first conversation text data and corresponding first work order data generated by a customer service and a user in a business conversation process in a historical time period can be obtained; training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data; when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model; and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation. By applying the technical scheme of the application, the first classification label model and the second classification label model for judging the service category of the session and the work order information label model for filling the specific content of the work order can be obtained according to the dialogue text data training of the customer service and the user in the historical time period. And further, the problem that in the related technology, in each business session between customer service and a user, the customer service staff is required to manually search for a proper template from the template library, so that the working efficiency is reduced is solved.
Optionally, in a possible implementation manner of the present application, after S101 (obtaining the first dialog text data and the corresponding first work order data generated by the customer service and the user during the service session within the historical time period), the following steps may be implemented:
performing data association on the first dialogue text data and the corresponding first work order data to obtain associated historical data;
and storing the associated historical data into a comma separated value CSV file through the unique identification bit.
Further, as generally speaking, the translated conversation text of the voice call between the user and the customer service in the historical period is stored separately from the manual input work order corresponding to the conversation text. For example, a work order manually entered by the customer service can be stored in a relational database, and the call translation text is compressed and directly stored in a server disk due to more text contents.
In order to obtain data of model training, in the embodiment of the present application, first, data association is performed between a first service dialog text and a historical work order, and the dialog text and a corresponding work order are associated through a unique identification bit and then written into a CSV file.
Among them, the CSV file is a general, relatively simple file format. It is widely used by users, businesses and science. The most widespread application is the transfer of tabular data between programs that themselves operate on incompatible formats (often proprietary and/or non-canonical formats). Because a large number of programs support some CSV variant, the use of CSV file storage may be an alternative input/output format.
Optionally, in a possible embodiment of the present application, after storing the correlated history data in the comma separated value CSV file through the unique identification bit, the following steps may be implemented:
dividing the associated historical data into user text data and customer service text data;
respectively obtaining maximum word values of user text data and customer service text numbers, and performing preset processing on the sentence patterns of which the word values do not meet the maximum word values in the same sentence pattern, wherein the preset processing comprises any one of zero filling completion or interception processing to obtain a first processed text;
acquiring a maximum sentence value of the number of the customer service texts, and presetting the texts of which the sentence values do not meet the maximum sentence value in the first processed text to obtain a second processed text;
the second processed text is used as sample training data.
Furthermore, before training the classification label model, the embodiment of the application can also input the preprocessed data after truncation and completion into the model to serve as training data. For example, for each sentence in the conversation text, the conversation between the customer service and the user needs to retain the identity of the speaker (i.e. divided into user text data and customer service text data). And then intercepting the maximum word number T (first number), if the word number of each sentence is less than T (first number), filling and completing with zero, and if the word number of each sentence is more than T (first number), intercepting to ensure that the word number is L, and further obtaining a second processed text.
Furthermore, for each dialog in the dialog text, it is also necessary to intercept the maximum number of sentences L (the second number), if the number of sentences is less than L (the second number), zero padding completion needs to be performed, and if the number of sentences is greater than L (the second number), it is necessary to perform an interception operation to ensure that the number of sentences is L, and then obtain a second processed text, and use the second processed text as sample training data.
The first quantity and the second quantity are hyper-parameters, and can be selected preferentially according to the distribution of the dialogue text data. This is not particularly limited in this application.
Optionally, in a possible implementation manner of the present application, after combining the pattern history data and the sentence history data as the sample training data, the following steps may be further implemented:
acquiring sample training data;
and training the randomly initialized first-level attention model by using sample training data to obtain a first classification label model meeting a preset condition, wherein the first classification label model is used for determining the service type of the dialogue text data.
According to the method and the device, a randomly initialized first-level attention model can be trained through sample training data acquired based on a dialog text and corresponding work order first-level class labels, and then a trained first class label model is obtained; and training a dialogue text secondary classification model through the dialogue text and the work order secondary class label. The first classification label model and the second classification label model can both adopt HAN models (hierarchical attention models). The HAN model gives different weights to each word expression vector in a sentence to sum to obtain a sentence expression vector, and gives different weights to each sentence expression vector to sum to obtain a text expression vector.
Optionally, in a possible implementation manner of the present application, after obtaining the first classification label model satisfying the preset condition, the following steps may be further implemented:
obtaining model parameters and sample training data of a first classification label model;
and training the randomly initialized second-level attention model by using the model parameters and the sample training data to obtain a second classification label model meeting the preset conditions, wherein the second classification label model is used for determining sub-service labels under the service type of the dialogue text data.
Further, as shown in fig. 2, the model structures of the first classification label model and the second classification label model are shown. Because the structure of the second classification label model is consistent with that of the first classification label model, the input and the output of the model are similar, and because the data volume of the second classification is smaller than that of the first classification label model, a transfer learning mechanism is utilized when the second classification label model of the dialog text is trained, and model parameters except for an output layer of the first classification model of the dialog text are loaded when the model is initialized. Model convergence can be faster by loading parameters of the primary classification model, and the overfitting problem caused by insufficient training data can be relieved.
Further optionally, in a possible embodiment of the present application, after obtaining the second classification label model satisfying the preset condition, the following steps may be further performed:
extracting service categories corresponding to the third amount of work order information from the associated historical data, and respectively establishing corresponding category labels;
removing the output layers of the first classification label model, and after the hidden variables output by the second classification label model, performing class label judgment on the state of each work order information by using a multi-task learning mode aiming at the first number of output layers to obtain a corresponding judgment result;
and generating a work order information label model based on the judgment result and the second classification label model of the eliminated output layer.
Furthermore, the multi-task learning work order information extraction model is trained through the conversation text and the work order information, key information elements in the conversation text are extracted, and the work order content generation task is converted into a classification task, so that task simplification is achieved, and the accuracy and the normalization of the work order generation content are improved.
Specifically, the service categories corresponding to the third amount of work order information may be extracted from the associated historical data, and corresponding category labels may be respectively established to construct the multitask learning model. The work order information label model is configured as shown in fig. 3, an output layer of the second classification label model can be removed and parameters are fixed by using a transfer learning idea, according to the number K of the work order information, K output layers are connected behind the last output hidden variable of the model to classify and judge the state of each work order information, and each output layer is composed of a full connection layer and softmax. And completing the multi-task learning model networking of the work order information state judgment, and only updating the parameters of the output layer during model training.
Furthermore, the multi-task learning mode can be used for carrying out state classification on the K (third number) work order information at the same time, the condition that the K (third number) classification models are independently constructed and trained to independently judge the state of each work order information is avoided, the workload of model construction and training can be effectively reduced, and computer resources are saved. Meanwhile, the problem of model under-fitting caused by insufficient classification data of certain work order information states or over-fitting caused by unbalanced data distribution can be solved to a certain extent by utilizing the transfer learning idea.
Further optionally, in a possible implementation manner of the present application, determining, based on the first classification tag model and the second classification tag model, a work order template corresponding to the current service session includes:
inputting the second dialogue text data into the first classification label model to obtain a main service label corresponding to the second dialogue text data;
inputting the main service label and the second dialogue text data into a second classification label model to obtain a sub service label corresponding to the second dialogue text data;
and selecting a target work order template matched with the main service label and the sub-service labels from the pre-stored work order templates of a plurality of different service labels.
In the process of determining the target work order template, the preprocessed conversation text data can be input into a first classification label model to obtain a main service label (such as a fault service label, a consultation service label, a complaint service label, a service handling label and the like).
Furthermore, the main service label and the dialogue text data are also required to be input into a second classification label model to obtain sub-service labels (such as a cost service label, a personnel service label, a website service label, an equipment service, a network service, and the like), so as to determine the service attribute of the session, and thus, the work order template of the service attribute is selected.
In another embodiment of the present application, as shown in fig. 4, the present application further provides a device for intelligently generating a work order. The method comprises an acquisition module 201, a generation module 202, a determination module 203 and a filling module 204, wherein,
an obtaining module 201, configured to obtain first conversation text data and corresponding first work order data generated by a customer service and a user in a business conversation process in a historical time period;
the generating module 202 is configured to train to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data;
the determining module 203 is configured to determine a target work order template corresponding to the service session based on the first classification label model and the second classification label model when second conversation text data generated by the customer service and the user in the service session process is acquired;
and the filling module 204 is configured to fill the work order information in the target work order template based on the work order information tag model and the second conversation text data to obtain the target work order of the current conversation.
In the application, first conversation text data and corresponding first work order data generated by a customer service and a user in a business conversation process in a historical time period can be obtained; training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data; when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model; and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation. By applying the technical scheme of the application, the first classification label model and the second classification label model for judging the service category of the session and the work order information label model for filling the specific content of the work order can be obtained according to the dialogue text data training of the customer service and the user in the historical time period. And further, the problem that in the related technology, in each business session between customer service and a user, the customer service staff is required to manually search for a proper template from the template library, so that the working efficiency is reduced is solved.
In another embodiment of the present application, the obtaining module 201 further includes:
the obtaining module 201 is configured to perform data association on the first dialogue text data and the corresponding first work order data to obtain associated historical data;
and the acquisition module 201 is configured to store the associated history data into a comma separated value CSV file through the unique identification bit.
In another embodiment of the present application, the obtaining module 201 further includes:
the acquisition module 201 is configured to distinguish the associated historical data into user text data and customer service text data;
an obtaining module 201, configured to obtain maximum word values of the user text data and the customer service text number, respectively, and perform preset processing on a sentence pattern in which the word value does not satisfy the maximum word value in the same sentence pattern, where the preset processing includes any one of zero padding completion or truncation processing, so as to obtain a first processed text;
an obtaining module 201, configured to obtain a maximum sentence value of the number of customer service texts, and perform the preset processing on a text of which a sentence value does not satisfy the maximum sentence value in the first processed text to obtain a second processed text;
an obtaining module 201 configured to take the second processed text as sample training data.
In another embodiment of the present application, the obtaining module 201 further includes:
an acquisition module 201 configured to acquire sample training data;
the obtaining module 201 is configured to train a first-level attention model initialized at random by using sample training data, and obtain a first classification label model meeting a preset condition, where the first classification label model is used to determine a service type of the dialog text data.
In another embodiment of the present application, the obtaining module 201 further includes:
an obtaining module 201 configured to obtain model parameters of the first classification label model and sample training data;
the obtaining module 201 is configured to train the randomly initialized second-level attention model by using the model parameters and the sample training data to obtain a second classification label model meeting the preset condition, where the second classification label model is used to determine sub-service labels under the service type of the dialog text data.
In another embodiment of the present application, the obtaining module 201 further includes:
the obtaining module 201 is configured to extract service categories corresponding to a third amount of work order information from the associated historical data, and respectively establish corresponding category labels;
the obtaining module 201 is configured to remove the output layers of the first classification label model, and after the hidden variables output by the second classification label model, perform class label judgment on the states of each work order information by using a multi-task learning mode for the first number of output layers to obtain corresponding judgment results;
and the obtaining module 201 is configured to generate a work order information label model based on the judgment result and the second classification label model of the output layer.
In another embodiment of the present application, the obtaining module 201 further includes:
the obtaining module 201 is configured to input the second dialogue text data into the first classification tag model to obtain a main service tag corresponding to the second dialogue text data;
the obtaining module 201 is configured to input the main service tag and the second dialogue text data into the second classification tag model to obtain a sub-service tag corresponding to the second dialogue text data;
the obtaining module 201 is configured to select a target work order template matched with the main service label and the sub-service labels from pre-stored work order templates of a plurality of different service labels.
In another embodiment of the present application, the master service tag includes at least one of: fault service, consultation service, complaint service and service handling; and, the sub-service label includes at least one of: fee service, personnel service, website service, equipment service, network service.
Fig. 5 is a block diagram illustrating a logical structure of an electronic device in accordance with an exemplary embodiment. For example, the electronic device 300 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, an exercise device, a personal digital assistant, and the like.
In an exemplary embodiment, there is also provided a non-transitory computer-readable storage medium, such as a memory, including instructions executable by an electronic device processor to perform the method for intelligent generation of a work order, the method comprising: acquiring first conversation text data and corresponding first work order data generated by customer service and a user in a service conversation process in a historical time period; training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data; when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model; and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation. Optionally, the instructions may also be executable by a processor of the electronic device to perform other steps involved in the exemplary embodiments described above. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
In an exemplary embodiment, there is also provided an application/computer program product including one or more instructions executable by a processor of an electronic device to perform the above method of intelligent generation of a work order, the method comprising: acquiring first conversation text data and corresponding first work order data generated by customer service and a user in a service conversation process in a historical time period; training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data; when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model; and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation. Optionally, the instructions may also be executable by a processor of the electronic device to perform other steps involved in the exemplary embodiments described above.
Fig. 5 is an exemplary diagram of the computer device 30. Those skilled in the art will appreciate that the schematic diagram 5 is merely an example of the computer device 30 and does not constitute a limitation of the computer device 30 and may include more or less components than those shown, or combine certain components, or different components, e.g., the computer device 30 may also include input output devices, network access devices, buses, etc.
The Processor 302 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. The general purpose processor may be a microprocessor or the processor 302 may be any conventional processor or the like, the processor 302 being the control center for the computer device 30 and connecting the various parts of the overall computer device 30 using various interfaces and lines.
The modules integrated by the computer device 30 may be stored in a computer-readable storage medium if they are implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, all or part of the flow of the method according to the embodiments of the present invention may also be implemented by hardware related to computer readable instructions, which may be stored in a computer readable storage medium, and when the computer readable instructions are executed by a processor, the steps of the method embodiments may be implemented.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (11)
1. A method for intelligently generating a work order is characterized by comprising the following steps:
acquiring first conversation text data and corresponding first work order data generated by customer service and a user in a service conversation process in a historical time period;
training to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data;
when second conversation text data generated by the customer service and the user in the business conversation process is collected, determining a target work order template corresponding to the business conversation based on the first classification label model and the second classification label model;
and filling the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation.
2. The method of claim 1, wherein after the obtaining the first dialog text data and the corresponding first work order data generated by the customer and the user during the business session within the historical time period, further comprising:
performing data association on the first dialogue text data and the corresponding first work order data to obtain associated historical data;
and storing the associated historical data into a comma separated value CSV file through a unique identification bit.
3. The method of claim 2, wherein after storing the associated history data to a comma separated value CSV file via a unique identification bit, further comprising:
dividing the associated historical data into user text data and customer service text data;
respectively obtaining the maximum word numerical values of the user text data and the customer service text number, and performing preset processing on the sentence patterns of which the word numerical values do not meet the maximum word numerical values in the same sentence pattern, wherein the preset processing comprises any one of zero padding completion or interception processing to obtain a first processed text;
acquiring a maximum sentence value of the number of the customer service texts, and performing preset processing on the texts of which the sentence values do not meet the maximum sentence value in the first processed text to obtain a second processed text;
and taking the second processed text as sample training data.
4. The method of claim 3, wherein after said combining said lexical history data and said sentence history data as sample training data, further comprising:
acquiring the sample training data;
and training a first-level attention model initialized randomly by using the sample training data to obtain the first classification label model meeting preset conditions, wherein the first classification label model is used for determining the service type of the dialogue text data.
5. The method according to claim 3 or 4, wherein after obtaining the first classification label model satisfying a preset condition, the method further comprises:
obtaining model parameters of a cleared output layer in the first classification label model and the sample training data;
and training a randomly initialized second-level attention model by using the model parameters and the sample training data to obtain a second classification label model meeting preset conditions, wherein the second classification label model is used for determining sub-service labels under the service type of the dialogue text data.
6. The method of claim 5, wherein after obtaining the second classification label model satisfying a preset condition, further comprising:
extracting service categories corresponding to a third amount of work order information from the associated historical data, and respectively establishing corresponding category labels;
removing the output layers of the first classification label model, and after the hidden variables output by the second classification label model, performing classification label judgment on the state of each work order information by using a multi-task learning mode aiming at the first number of output layers to obtain a corresponding judgment result;
and generating the work order information label model based on the judgment result and the second classification label model of the removal output layer.
7. The method of claim 1, wherein the determining the work order template corresponding to the current service session based on the first classification label model and the second classification label model comprises:
inputting the second dialogue text data into the first classification label model to obtain a main business label corresponding to the second dialogue text data;
inputting the main service label and the second dialogue text data into the second classification label model to obtain a sub service label corresponding to the second dialogue text data;
and selecting a target work order template matched with the main service label and the sub service labels from a plurality of pre-stored work order templates of different service labels.
8. The method of claim 7, wherein the master service label comprises at least one of: fault service, consultation service, complaint service and service handling; and the number of the first and second groups,
the sub-service label includes at least one of: fee service, personnel service, website service, equipment service, network service.
9. An apparatus for intelligent generation of work orders, comprising:
the acquisition module is configured to acquire first conversation text data and corresponding first work order data generated by customer service and a user in a business conversation process in a historical time period;
the generating module is configured to train to obtain a first classification label model, a second classification label model and a work order information label model based on the first dialogue text data and the corresponding first work order data;
the determining module is configured to determine a target work order template corresponding to the business session based on the first classification label model and the second classification label model when second conversation text data generated by the customer service and the user in the business session process is acquired;
and the filling module is configured to fill the work order information in the target work order template based on the work order information label model and the second conversation text data to obtain the target work order of the current conversation.
10. An electronic device, comprising:
a memory for storing executable instructions; and the number of the first and second groups,
a processor for display with the memory to execute the executable instructions to perform the operations of the method of intelligent generation of work orders of any of claims 1-8.
11. A computer-readable storage medium storing computer-readable instructions that, when executed, perform the operations of the method for intelligent generation of work orders of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011150090.4A CN112434501A (en) | 2020-10-23 | 2020-10-23 | Work order intelligent generation method and device, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011150090.4A CN112434501A (en) | 2020-10-23 | 2020-10-23 | Work order intelligent generation method and device, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112434501A true CN112434501A (en) | 2021-03-02 |
Family
ID=74696004
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011150090.4A Pending CN112434501A (en) | 2020-10-23 | 2020-10-23 | Work order intelligent generation method and device, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112434501A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113779216A (en) * | 2021-09-02 | 2021-12-10 | 携程计算机技术（上海）有限公司 | Event recording method, system, device and storage medium based on multi-party conversation |
| CN113836864A (en) * | 2021-08-26 | 2021-12-24 | 广西电网有限责任公司来宾供电局 | Development work order content comparison method and system |
| CN114979723A (en) * | 2022-02-14 | 2022-08-30 | 杭州脸脸会网络技术有限公司 | Virtual intelligent customer service method, device, electronic device and storage medium |
| CN116644145A (en) * | 2023-07-26 | 2023-08-25 | 北京仁科互动网络技术有限公司 | Session data processing method, device, equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102572134A (en) * | 2010-12-30 | 2012-07-11 | 上海博泰悦臻电子设备制造有限公司 | Work order processing method, work order building system and work order processing system |
| US8898779B1 (en) * | 2012-02-03 | 2014-11-25 | Symantec Corporation | Data loss prevention of information using structured document templates and forms |
| CN109726372A (en) * | 2017-10-31 | 2019-05-07 | 百度在线网络技术（北京）有限公司 | Generation method, device and the computer-readable medium of work order based on message registration |
| CN110888968A (en) * | 2019-10-15 | 2020-03-17 | 浙江省北大信息技术高等研究院 | Customer service dialogue intention classification method and device, electronic equipment and medium |
| CN111581390A (en) * | 2019-02-19 | 2020-08-25 | 北京嘀嘀无限科技发展有限公司 | Knowledge graph construction method and device and electronic equipment |
| CN111680497A (en) * | 2019-02-25 | 2020-09-18 | 北京嘀嘀无限科技发展有限公司 | Session feature extraction method, session recognition model training method and device |
-
2020
- 2020-10-23 CN CN202011150090.4A patent/CN112434501A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102572134A (en) * | 2010-12-30 | 2012-07-11 | 上海博泰悦臻电子设备制造有限公司 | Work order processing method, work order building system and work order processing system |
| US8898779B1 (en) * | 2012-02-03 | 2014-11-25 | Symantec Corporation | Data loss prevention of information using structured document templates and forms |
| CN109726372A (en) * | 2017-10-31 | 2019-05-07 | 百度在线网络技术（北京）有限公司 | Generation method, device and the computer-readable medium of work order based on message registration |
| CN111581390A (en) * | 2019-02-19 | 2020-08-25 | 北京嘀嘀无限科技发展有限公司 | Knowledge graph construction method and device and electronic equipment |
| CN111680497A (en) * | 2019-02-25 | 2020-09-18 | 北京嘀嘀无限科技发展有限公司 | Session feature extraction method, session recognition model training method and device |
| CN110888968A (en) * | 2019-10-15 | 2020-03-17 | 浙江省北大信息技术高等研究院 | Customer service dialogue intention classification method and device, electronic equipment and medium |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113836864A (en) * | 2021-08-26 | 2021-12-24 | 广西电网有限责任公司来宾供电局 | Development work order content comparison method and system |
| CN113779216A (en) * | 2021-09-02 | 2021-12-10 | 携程计算机技术（上海）有限公司 | Event recording method, system, device and storage medium based on multi-party conversation |
| CN114979723A (en) * | 2022-02-14 | 2022-08-30 | 杭州脸脸会网络技术有限公司 | Virtual intelligent customer service method, device, electronic device and storage medium |
| CN114979723B (en) * | 2022-02-14 | 2023-08-29 | 杭州脸脸会网络技术有限公司 | Virtual intelligent customer service method, device, electronic device and storage medium |
| CN116644145A (en) * | 2023-07-26 | 2023-08-25 | 北京仁科互动网络技术有限公司 | Session data processing method, device, equipment and storage medium |
| CN116644145B (en) * | 2023-07-26 | 2023-10-13 | 北京仁科互动网络技术有限公司 | Session data processing method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112346567B (en) | Virtual interaction model generation method and device based on AI (Artificial Intelligence) and computer equipment | |
| CN112434501A (en) | Work order intelligent generation method and device, electronic equipment and medium | |
| CN111917878B (en) | Message processing method, device, equipment and storage medium | |
| CN112380870A (en) | User intention analysis method and device, electronic equipment and computer storage medium | |
| CN111666400B (en) | Message acquisition method, device, computer equipment and storage medium | |
| CN108305050A (en) | Information of reporting a case to the security authorities and the extracting method of service requirement information, device, equipment and medium | |
| CN111651571A (en) | Man-machine cooperation based session realization method, device, equipment and storage medium | |
| CN110598070A (en) | Application type identification method and device, server and storage medium | |
| CN114549241A (en) | Contract examination method, device, system and computer readable storage medium | |
| CN112632244A (en) | Man-machine conversation optimization method and device, computer equipment and storage medium | |
| CN117251547A (en) | User question response method and device, equipment and medium thereof | |
| CN110399473B (en) | Method and device for determining answers to user questions | |
| CN110059178A (en) | Problem distributing method and device | |
| CN111309882B (en) | Method and device for realizing intelligent customer service question and answer | |
| CN113360654A (en) | Text classification method and device, electronic equipment and readable storage medium | |
| CN117520503A (en) | Financial customer service dialogue generation method, device, equipment and medium based on LLM model | |
| CN117313736A (en) | Implicit demand recognition method and device for comment text | |
| CN114528851B (en) | Reply sentence determination method, reply sentence determination device, electronic equipment and storage medium | |
| CN115953080A (en) | Engineer service level determination method, apparatus and storage medium | |
| CN110046233A (en) | Problem distributing method and device | |
| CN114239602A (en) | Session method, apparatus and computer program product | |
| CN114186024A (en) | Recommendation method and device | |
| CN112163075A (en) | Information recommendation method and device, computer equipment and storage medium | |
| CN112328871B (en) | Reply generation method, device, equipment and storage medium based on RPA module | |
| CN112558913A (en) | Conversation method and device based on aggregated card, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |