CN114168734A - Client event list classification method, device, equipment and storage medium - Google Patents
Client event list classification method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN114168734A CN114168734A CN202111489314.9A CN202111489314A CN114168734A CN 114168734 A CN114168734 A CN 114168734A CN 202111489314 A CN202111489314 A CN 202111489314A CN 114168734 A CN114168734 A CN 114168734A
- Authority
- CN
- China
- Prior art keywords
- target
- data
- historical
- event list
- attribute
- 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
Images
Classifications
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the invention provides a method, a device, equipment and a storage medium for classifying client event lists, and relates to the technical field of artificial intelligence. The specific implementation scheme comprises the following steps: acquiring incoming call content data, target attribute data and historical relevance data in a target client event list; extracting target text characteristics corresponding to the incoming call content data; determining target attribute characteristics corresponding to the target attribute data; determining historical relevance characteristics corresponding to the historical relevance data; and determining the corresponding category of the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic. The client event list can be automatically classified, and the classification efficiency of the client event list is effectively improved. And effectively improves the accuracy of classification.
Description
Technical Field
The embodiment of the invention relates to the technical field of artificial intelligence, in particular to a method, a device, equipment and a storage medium for classifying a client event list.
Background
With the development of information technology, the linkage processing process of the client event list increasingly needs an automatic and intelligent processing method to reduce the participation of manual links, thereby reducing the manual errors and accelerating the linkage processing speed.
Currently, when classifying the customer event list, business personnel generally obtain the customer event list based on specific business knowledge and element analysis in the customer event list. The efficiency of classifying the client event list is low, and according to different service capabilities, classification deviation is generated during manual classification, and the accuracy of classifying the client event list is low.
Disclosure of Invention
The embodiment of the invention provides a method, a device, equipment and a storage medium for classifying a client event list, which are used for solving the technical problems that in the prior art, the efficiency of classifying the client event list is low, and the accuracy of classifying the client event list is also low.
In a first aspect, an embodiment of the present invention provides a method for classifying a customer event list, including:
acquiring incoming call content data, target attribute data and historical relevance data in a target client event list;
extracting target text characteristics corresponding to the incoming call content data;
determining target attribute characteristics corresponding to the target attribute data;
determining historical relevance characteristics corresponding to the historical relevance data;
and determining the category corresponding to the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic.
In a second aspect, an embodiment of the present invention provides an electronic device, including: at least one processor and memory
The memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the method of any one of the first aspects.
In a third aspect, the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored, and when executed by a processor, the computer-executable instructions are used to implement the method according to any one of the first aspect.
According to the classification method, the classification device, the classification equipment and the classification storage medium of the client event list, incoming call content data, target attribute data and historical relevance data in a target client event list are obtained; extracting target text characteristics corresponding to the incoming call content data; determining target attribute characteristics corresponding to the target attribute data; determining historical relevance characteristics corresponding to the historical relevance data; and determining the category corresponding to the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic. The incoming call content data, the attribute data and the historical relevance data of the corresponding client related to the category of the target client event list are fully acquired. And acquiring target text characteristics, target attribute characteristics and historical relevance characteristics corresponding to the three data, and determining the category corresponding to the target client event list according to the target text characteristics, the target attribute characteristics and the historical relevance characteristics. The client event list can be automatically classified, and the classification efficiency of the client event list is effectively improved. And effectively improves the accuracy of classification.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure.
FIG. 1 is a network architecture diagram of a method of classifying customer event tickets in which embodiments of the present invention may be implemented;
FIG. 2 is a flowchart illustrating a method for classifying a customer event list according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a method for classifying a customer event ticket according to another embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating the extraction of target semantic detail features in the classification method of the customer event list according to another embodiment of the present invention;
FIG. 5 is a diagram illustrating the extraction of target context features in a classification method for a client event list according to another embodiment of the present invention;
FIG. 6 is a diagram illustrating classification in a method for classifying a customer event list according to another embodiment of the present invention;
FIG. 7 is a flowchart illustrating a method for classifying a customer event list according to another embodiment of the present invention;
FIG. 8 is a flowchart illustrating a method for classifying a customer event ticket according to another embodiment of the present invention;
FIG. 9 is a schematic structural diagram of an electronic device for implementing a method for classifying a customer event list according to an embodiment of the present invention;
fig. 10 is a block diagram of an electronic device for implementing a method for classifying a customer event ticket according to an embodiment of the present invention.
With the foregoing drawings in mind, certain embodiments of the disclosure have been shown and described in more detail below. These drawings and written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the concepts of the disclosure to those skilled in the art by reference to specific embodiments.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
First, terms related to embodiments of the present invention are explained:
a client event sheet: for recording and describing the reason of customer calling, the problems or expected results.
Word embedding algorithm: the method is an algorithm for learning word vectors corresponding to words through a neural network.
A convolutional neural network: the method is a feedforward neural network containing convolution operation and having a certain depth, is one of representative algorithms for deep learning, and is generally applied to the fields of computer vision, natural language processing and the like.
And (3) a gated loop unit algorithm: abbreviated as GRU. The artificial neural network is a kind of recurrent neural network algorithm, and has a tree-like hierarchical structure, and the network nodes recur the input information according to the connection order.
XGboost classifier: the classification model is based on a decision tree, continuously screens and combines features in an iterative process, and can integrate a plurality of weak classifiers together to form a strong classifier.
For a clear understanding of the technical solutions of the present application, the prior art solutions are described in detail.
The client event list mainly comprises a plurality of work order elements, one part of the work order elements is elements related to the client, mainly comprises client names, client incoming call numbers, client certificate numbers, client card numbers, province and city belonging elements, and the other part of the work order elements is elements related to events, mainly comprises event types, event sources, incoming call contents and the like. The "event type" mainly refers to the category to which the event belongs, and generally includes four categories, namely complaint, suggestion, raise and help.
When the customer event list is classified, business personnel generally obtain the customer event list based on specific business knowledge and element analysis in the customer event list, so the classification process is slow, and the classification efficiency of the customer event list is low. And the accuracy of the classification result is related to the business capability of business personnel, so that the business capability of some business personnel is lower, the classification deviation can be generated when the classification is carried out manually, and the classification accuracy rate of the customer event list is also lower.
Methods also exist for automatically classifying customer event tickets using deep learning models. The convolutional neural network model (CNN) and the long-term and short-term memory network model (LSTM) are common. These methods of automatically classifying the customer event tickets using the deep learning model generally automatically classify the customer event tickets based on incoming call content data. But the category to which the customer event ticket belongs is related not only to incoming call content data but also to other data. Therefore, the accuracy of the method for automatically classifying the client event list by adopting the deep learning model is lower at present.
Therefore, in the face of the technical problems in the prior art, the inventor finds, through creative research, that the category corresponding to the client event list is not only related to the incoming call content data, but also related to important attribute data in the client event list and historical relevance data corresponding to the client. It is necessary to fully consider the data of these three aspects when classifying the customer event tickets. Extracting target text characteristics corresponding to the incoming call content data; determining target attribute characteristics corresponding to the target attribute data; and finally, determining the category corresponding to the target client event list according to the target text feature, the target attribute feature and the historical relevance feature. The client event list can be automatically classified, and the classification efficiency of the client event list is effectively improved. And moreover, all data and characteristics related to the client event list category are fully considered, so that the classification accuracy is effectively improved.
Therefore, the inventor proposes a technical scheme of the embodiment of the invention based on the above creative discovery. The following describes a network architecture of a method for classifying a client event ticket according to an embodiment of the present invention.
Fig. 1 is a network architecture diagram that can implement the method for classifying a customer event ticket according to an embodiment of the present invention, and as shown in fig. 1, the network architecture of the method for classifying a customer event ticket according to this embodiment includes: an electronic apparatus 1 and a processing mechanism apparatus 2. The business personnel adopt the electronic equipment 1 to record the work order elements in the event list so as to generate a client event list. And may store the customer event ticket locally. The electronic device 1 acquires the target customer event list to be classified from the local. And acquiring the incoming call content data in the target client event list. And acquiring target attribute data from the work order elements. And acquiring a history client event list which is classified and corresponds to the target client and generates corresponding history correlation data. The classification method of the client event list provided by the invention is adopted to classify the target client event list so as to obtain the corresponding category of the target client event list. The processing mechanism and the processing method for acquiring the client event list of each type which is configured in advance. And sending the target client event list, the corresponding category and the corresponding processing method to the corresponding processing mechanism device 2, so that the corresponding processing mechanism device 2 can process the target client event list according to the processing method.
It is to be understood that the generated customer event list may also be stored in a certain electronic device or a database independently, which is not limited in this embodiment.
The following describes the technical solutions of the present invention and how to solve the above technical problems with specific embodiments. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.
Example one
Fig. 2 is a schematic flow chart of a method for classifying a customer event list according to an embodiment of the present invention, and as shown in fig. 2, an execution subject of the method for classifying a customer event list according to the embodiment is a device for classifying a customer event list. The sorting means for the customer event ticket is located in the electronic device. The method for classifying the customer event list provided by the embodiment includes the following steps:
step 201, obtaining incoming call content data, target attribute data and history correlation data in the target client event list.
In this embodiment, the target customer event list is a customer event list to be classified.
The work order elements included in the target customer event order may include: event list number, client name, client incoming call number, client certificate number, client gender, client card number, elements such as province and city, event source, event emergency degree, incoming call time, incoming call content and the like.
And the other work order elements in the target client event list except the incoming call content are the attribute data of the target client event list. And the attribute data which is relatively large in relation to the category corresponding to the target client event list in the attribute data of the target client event list is target attribute data.
In this embodiment, a client event list represents one appeal of a client, but the appeal of the client often has certain coherence, for example, a more urgent help-seeking client event list, the client often makes a call for multiple times in a short time for feedback, so that the client behavior class characteristics have certain auxiliary value for the classification of the client event list, but the characteristics cannot be extracted by single text classification, so that historical relevance data is obtained. The historical relevance data is data of the distribution condition of corresponding categories of the historical client event list belonging to the same client as the target client event list.
Specifically, in this embodiment, the placement position of each attribute data and incoming call content data in the target client event list can be configured in advance according to the format of the client event list. And the placement position of the target attribute data can be configured in advance. And acquiring the incoming call content data according to the placement position of the incoming call content data after acquiring the target client event list, and acquiring the target attribute data according to the placement position of the preset target attribute data. And acquiring attribute data representing the client identification information from the attribute data, and acquiring data of the distribution condition of the corresponding category of the historical client event list belonging to the same client as the target client event list from the historical client event list as historical relevance data.
Step 202, extracting target text features corresponding to the incoming call content data.
In this embodiment, a deep learning model or a machine learning model may be adopted to extract target text features corresponding to the incoming call content data according to the characteristics of the incoming call content data. Semantic detail features and context features can be expressed in the target text features.
The type of the deep learning model or the machine learning model is not limited.
Step 203, determining the target attribute characteristics corresponding to the target attribute data.
In this embodiment, the target attribute data may be subjected to normalized encoding processing, and a vector or a matrix after the normalized encoding processing is determined as the target attribute feature.
And step 204, determining historical relevance characteristics corresponding to the historical relevance data.
In this embodiment, data of category distribution conditions corresponding to historical client event sheets belonging to the same client as the target client event sheet is obtained. And representing the data of the category distribution condition corresponding to the historical client event list belonging to the same client as the target client event list by adopting a matrix or a vector so as to determine the corresponding historical relevance characteristic.
Step 205, determining the category corresponding to the target client event list according to the target text feature, the target attribute feature and the historical relevance feature.
In this embodiment, the target text feature, the target attribute feature, and the historical relevance feature may be input into a preset classifier, the preset classifier is adopted to classify the target client event list according to the target text feature, the target attribute feature, and the historical relevance feature, and a category corresponding to the target client event list is output.
The preset classifier type is not limited, and may be an XGBoost classifier, for example.
The category corresponding to the target client event list is any one of the following categories: complaining, suggesting, showing and seeking help.
In the method for classifying the client event list provided by the embodiment, incoming call content data, target attribute data and historical relevance data in a target client event list are acquired; extracting target text characteristics corresponding to the incoming call content data; determining target attribute characteristics corresponding to the target attribute data; determining historical relevance characteristics corresponding to the historical relevance data; and determining the corresponding category of the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic. The incoming call content data, the attribute data and the historical relevance data of the corresponding client related to the category of the target client event list are fully acquired. And acquiring target text characteristics, target attribute characteristics and historical relevance characteristics corresponding to the three data, and determining the category corresponding to the target client event list according to the target text characteristics, the target attribute characteristics and the historical relevance characteristics. The client event list can be automatically classified, and the classification efficiency of the client event list is effectively improved. And effectively improves the accuracy of classification.
In the prior art, when the content incoming call data is classified by adopting the CNN model, sequential information in a text is ignored in the convolution operation process, and semantic information cannot be learned for the text with a long text and an association relation between the text and a context.
When the LSTM model is adopted to classify the incoming call content data, compared with CNN, the method is more suitable for learning long-term dependency relationship, so that the method is more suitable for processing data of long text type, and is not suitable for short text, and when processing word sequences in time sequence, the influence degree of each input word vector on model training is different, and the influence of words appearing later on the model is stronger.
However, the text length of the incoming call content data is influenced by factors such as complexity and severity of the content fed back by the client, the text content description of the event list which is easier to process is shorter, and the text content description corresponding to the more complex event is longer. And important contents are not arranged behind in the incoming call content data. Therefore, when classifying the incoming call content data, the classification method is required to have good performance on text contents with different lengths, so that the CNN model and the LSTM model have certain limitations.
Therefore, in the embodiment of the invention, when the target text characteristic corresponding to the incoming call content data is extracted, the convolutional neural network from training to convergence is adopted to extract the target semantic detail characteristic corresponding to the incoming call content data, and the gated cyclic unit algorithm from training to convergence is adopted to extract the target context characteristic corresponding to the incoming call content data. When the two text features are extracted, the semantic detail features and the context features of the incoming call content data can be fully extracted without being influenced by the text length of the incoming call content data and the important content distribution of the incoming call content data.
Fig. 3 is a schematic flowchart of a method for classifying a customer event ticket according to another embodiment of the present invention, and as shown in fig. 3, the method for classifying a customer event ticket according to the present embodiment further refines steps 202 to 204 on the basis of the method for classifying a customer event ticket according to the first embodiment, and then the method for classifying a customer event ticket according to the present embodiment includes the following steps:
step 301, obtaining incoming call content data, target attribute data and history correlation data in the target client event list.
In this embodiment, the implementation manner of step 301 is similar to that of step 201 in the first embodiment, and is not described in detail here.
Step 302, performing text segmentation processing and word segmentation processing on the incoming call content data to obtain a word sequence corresponding to the incoming call content data.
In this embodiment, a natural language processing technology is adopted to perform text segmentation on incoming call content data according to semantics to form text sentences, so that each segmented text sentence accurately expresses the semantics. And then, performing word segmentation on each text sentence by adopting a word segmentation algorithm, removing stop words, shielding words and the like, and forming a word sequence corresponding to the incoming call content data by using the reserved words.
Wherein the word sequence may be represented as S ═ { x ═ x1,x2,...,xlWhere x denotes the word and l is the sequence length.
Step 303, performing word vector representation on words in the word sequence by using a word embedding algorithm to obtain a text matrix composed of a plurality of word vectors.
In this embodiment, before extracting the target text feature from the word sequence, the word sequence needs to be converted into data that can be processed by a computer, which is referred to as feature representation. In this step, a word embedding algorithm in the natural language processing technology is adopted, and the word sequence is input into the word embedding algorithm. Word embedding algorithms perform word vector representations on words in a sequence of words. After the word sequence of the incoming call content data is expressed by the word vector, a text matrix formed by a plurality of word vectors is formed.
Wherein the text matrix can be represented as: e is an element of RlnWhere l is the sequence length and n is the word vector dimension.
And step 304, extracting target semantic detail characteristics corresponding to the incoming call content data by adopting a convolutional neural network trained to be convergent.
In this embodiment, the target text features include: target semantic detail features and target context features. And the target semantic detail features are obtained by performing feature extraction on the incoming call content data by adopting a convolutional neural network trained to be convergent.
As an alternative implementation, in this embodiment, step 304 includes the following steps:
step 3041, the text matrix is input into a convolutional neural network trained to converge.
As shown in fig. 4, the text matrix E RlnInput into a convolutional neural network trained to converge.
Wherein the convolutional neural network trained to converge comprises a plurality of convolutional kernels. The plurality of convolution kernels have a plurality of different window sizes.
Step 3042, performing convolution operation on the text matrix by using a plurality of convolution kernels in the convolutional neural network trained to converge, respectively, to obtain a plurality of convolution feature data.
In this embodiment, the convolution operation performed on the text matrix by each convolution kernel is expressed as formula (1) and formula (2):
ci＝f(wei:i+h-1+ b) formula (1)
C＝[c1,c2,...cl-h+1]Formula (2)
Wherein, ciFor a convolution kernel on the ith window with a part e of the text matrixi:i+h-1And (5) performing convolution operation. Wherein the window size of the convolution kernel is [ i: i + h-1 ]]. f is the activation function, w is the parameter matrix, and b is the bias term. And C is convolution characteristic data formed after convolution operation of the convolution kernel and the text matrix according to the fixed window and the convolution operation of the convolution kernel and the text matrix is continued after the convolution kernel moves. l + h-1 represents the number of times that the convolution kernel performs convolution operations with a corresponding partial matrix in the text matrix.
Since there are multiple convolution kernels in the convolutional neural network trained to converge, if the number of convolution kernels is so m, the number of convolution feature data is m.
Step 3043, perform the maximum pooling operation on the convolution feature data corresponding to each convolution kernel to obtain the most significant convolution feature corresponding to each convolution kernel.
In this embodiment, the most significant convolution feature can be expressed as shown in equation (3):
PC＝max C＝max[c1,c2,...,cl-h+1]formula (3)
Where max represents the maximum pooling operation. PCRepresenting the most significant convolution characteristic for a certain convolution kernel.
Then the most significant feature corresponding to the mth convolution kernel can be expressed as
Step 3044, determining the most significant convolution features corresponding to each convolution kernel as the target semantic detail features.
In this embodiment, because the text matrix is respectively subjected to convolution operation by using the plurality of convolution kernels to obtain the plurality of convolution feature data, the convolution feature data can be enriched, and after the maximum pooling operation is performed on the convolution feature data corresponding to each convolution kernel, the most significant convolution feature can be obtained, and the most significant convolution feature can embody semantic details of the incoming call content data, so that the most significant convolution feature corresponding to each convolution kernel is determined as the target semantic detail feature.
The target semantic detail feature can be expressed as shown in equation (4):
wherein, FCIs a target semantic detail feature. Wherein, as shown in fig. 4, the target semantic detail feature is represented in a matrix form.
And 305, extracting the target context characteristics corresponding to the incoming call content data by adopting a gated cyclic unit algorithm trained to be convergent.
It should be noted that steps 304-305 are an alternative implementation of step 202 in the first embodiment.
As an alternative implementation, in this embodiment, step 305 includes the following steps:
step 3051, inputting the text matrix into a gated cyclic unit algorithm trained to converge.
As shown in FIG. 5, the text matrix E ∈ RlnInput into a gated round robin unit algorithm trained to converge.
Step 3052, extracting hidden layer state data corresponding to each word vector in the text matrix by adopting a gated cycle unit algorithm from training to convergence.
The process of extracting hidden layer state data corresponding to each word vector in the text matrix by the control cycle unit algorithm is represented by formula (5) to formula (8):
zi＝σ(WZei+Uzhi-1) Formula (5)
ri＝σ(Wrei+Urhi-1) Formula (6)
Wherein e isiFor the i-th word vector, the word vector,
is candidate hidden layer state data, hiIs the hidden layer state data corresponding to the ith word vector. z is a radical ofiIndicating that the gate is updated, the range is [0,1 ]]，riIs a reset gate, which can determine whether the state value of the candidate hidden layer abandons the state value of the hidden layer at the previous moment, and the other variables are parameters,
representing element multiplication.
After the hidden layer state data corresponding to each word vector in the text matrix is extracted by adopting a gated cyclic unit algorithm from training to convergence, the output hidden layer state data can be represented as shown in formula (9):
H＝[h1,h2,...,hl]formula (9)
Step 3053, performing maximum pooling operation on the hidden layer state data corresponding to all the word vectors to obtain the most significant hidden layer state data.
In this embodiment, the most significant hidden layer state data can be represented by equation (10):
Pr＝max H＝max[h1,h2,...,hl]formula (10)
Wherein, PrDenoted as the most significant hidden layer state data. max is expressed as the maximum pooling operation.
Step 3054, determining the most significant hidden layer state data as the target context feature.
In this embodiment, since the effect of the previous word vector is taken into consideration when determining the hidden layer state data corresponding to each word vector, the most significant hidden layer state data is determined as the target context feature. Wherein the target context feature may also represent Pr. So that as shown in figure 5 of the drawings,the target context characteristics are represented in the form of a matrix.
Step 306, encoding the target attribute data to obtain corresponding target attribute characteristics.
In this embodiment, step 305 is an alternative implementation of step 203.
Specifically, in this embodiment, each target attribute data is subjected to normalization coding processing, and after vectors or matrices subjected to normalization coding processing are spliced, a target attribute feature is formed.
Illustratively, the target attribute data includes gender, province and city, and urgency. And carrying out normalized coding on the three target attribute data, and splicing the vectors or matrixes subjected to normalized coding to form the whole target attribute characteristic.
And 307, performing matrix representation on the historical relevance data to obtain historical relevance characteristics.
Optionally, in this embodiment, the historical relevance data is category distribution data corresponding to historical customer event tickets of the target customer in different statistical periods.
Illustratively, the target client in the target client event ticket is client a. The different statistical periods were the last week, the last 1 month, the last 2 months. Counting the category distribution data corresponding to the historical event list of the client a in the last week, wherein the category distribution data are respectively as follows: complaints are: item 0, suggest: 1 piece, table raise: item 0, help: and (5) pieces. The statistical data of the category distribution corresponding to the historical event list of the client a in the last 1 month are respectively as follows: complaints are: item 0, suggest: 3 pieces, table raise: 1, seeking help: and (5) pieces. The statistical data of the category distribution corresponding to the historical event list of the client a in the last 2 months are respectively as follows: complaints are: item 0, suggest: 5 pieces, table raise: 3, seeking help: and (5) pieces. Then the historical relevance data can form a column vector of category distribution data according to each statistical period, and the column vectors of corresponding category distribution data formed in different statistical periods are spliced to form the historical relevance feature.
And 308, determining the corresponding category of the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic.
As an alternative implementation, in this embodiment, step 308 includes the following steps:
step 3081, the target text characteristics, the target attribute characteristics and the historical relevance characteristics are spliced to obtain target characteristics corresponding to the target client event list.
The target text feature can be expressed as shown in formula (11):
P＝[FcPr]formula (11)
Wherein P is a target text feature.
Specifically, in this embodiment, the target text feature, the target attribute feature, and the historical relevance feature may all be represented by vectors or matrices, so that the target text feature, the target attribute feature, and the historical relevance feature may be spliced to form a target matrix, and the target matrix is a target feature corresponding to the target customer event ticket.
3082, the target features are input into an XGboost classifier trained to converge.
3083, classifying the target client event list according to the target characteristics by using the XGboost classifier trained to be convergent, and outputting the category corresponding to the target client event.
In this embodiment, since the target features include target text features, target attribute features, and historical relevance features, the dimensions of the target features are large, and as shown in fig. 6, after the target features are input into the XGBoost classifier trained to converge, the XGBoost classifier trained to converge is used to screen and combine the target features, thereby determining the category corresponding to the target client event ticket.
As shown in fig. 6, the output target customer event list corresponds to any one of complaints, suggestions, praise and help.
In the method for classifying a customer event list provided by this embodiment, the target text features include: when extracting the target text characteristics corresponding to the incoming call content data, extracting the target semantic detail characteristics corresponding to the incoming call content data by adopting a convolutional neural network trained to be convergent, and extracting the target context characteristics corresponding to the incoming call content data by adopting a gated cyclic unit algorithm trained to be convergent.
According to the classification method of the client event list provided by the embodiment, when the category corresponding to the target client event list is determined according to the target text feature, the target attribute feature and the historical relevance feature, the target text feature, the target attribute feature and the historical relevance feature are spliced to obtain the target feature corresponding to the target client event list; inputting the target features into an XGboost classifier trained to converge; and classifying the target client event list according to the target characteristics by adopting an XGboost classifier trained to be convergent, and outputting the category corresponding to the target client event. The target features have larger dimensionality, and the XGboost classifier trained to be convergent can effectively screen and combine the target features, so that the XGboost classifier is a strong classifier, and the accuracy of the classification result of the client event list can be effectively improved.
EXAMPLE III
Fig. 7 is a schematic flow chart of a method for classifying a client event ticket according to another embodiment of the present invention, and as shown in fig. 7, the method for classifying a client event ticket according to this embodiment further includes a step of training a preset convolutional neural network, a preset gated cyclic unit algorithm, and a preset XGBoost classifier based on the above embodiment. The method for classifying a customer event list provided by this embodiment further includes the following steps:
step 401, a training sample is obtained, where the training sample includes a historical customer event list and a category label to which the historical customer event list belongs.
In this embodiment, the process of classifying the client event list includes three models, which are a preset convolutional neural network, a preset gate control cycle unit algorithm, and a preset XGBoost classifier. So that three models need to be trained. Before training the three models, training samples were obtained. The training sample comprises a historical client event list marked with a category label to which the historical client event list belongs.
Before the model is trained by adopting the historical client event list, incoming call content data, target historical attribute data and historical relevance data in the historical client event list also need to be acquired. The manner of obtaining the incoming call content data, the target attribute data, and the history association data in the target client event list is similar to that in the above embodiment, and details are not repeated here.
Step 402, training a preset convolutional neural network, a preset gating cycle unit algorithm and a preset XGboost classifier by using training samples to obtain the convolutional neural network trained to be convergent, the gating cycle unit algorithm trained to be convergent and the XGboost classifier trained to be convergent.
In this embodiment, a preset convolutional neural network and a preset gate control cycle unit algorithm in the three models may be trained, while parameters in the preset XGBoost classifier are kept unchanged. After the parameters in the preset convolutional neural network and the preset gating cycle unit algorithm are trained to be convergent, keeping the parameters in the convolutional neural network trained to be convergent and the gating cycle unit algorithm trained to be convergent unchanged, and continuing to train the parameters in the preset XGboost classifier. And finally, fine-tuning the parameters according to the training result to minimize the loss function corresponding to each trained model, thereby obtaining a convolutional neural network trained to be convergent, a gating cycle unit algorithm trained to be convergent and an XGboost classifier trained to be convergent.
When a preset convolutional neural network and a preset gate control cycle unit algorithm are trained, semantic detail features corresponding to incoming call content data in a historical client event list are extracted by adopting a preset convolutional neural network model, and context features corresponding to the incoming call content data in the historical client event list are extracted by adopting the preset gate control cycle unit algorithm. And determining the category corresponding to the target client event list by adopting a preset XGboost classifier according to the text characteristics, the target attribute characteristics and the historical relevance characteristics corresponding to the historical client event list. The specific implementation manner is similar to the implementation manner of the corresponding steps in the above embodiments, and details are not repeated here.
Wherein, when the number of the training samples is M and there are n class labels, there are
Definition of
And (3) representing the predicted score value of the ith historical client event list on the nth class, and representing a loss function corresponding to the preset XGboost classifier as shown in an equation (12):
where L represents the cross entropy loss function, yiDenotes class i tags and Ω (f) is a penalty term.
In the method for classifying the client event tickets provided by the embodiment, training samples are obtained, wherein the training samples comprise historical client event tickets and class labels to which the historical client event tickets belong; training a preset convolutional neural network, a preset gating cycle unit algorithm and a preset XGboost classifier by using training samples to obtain the convolutional neural network from training to convergence, the gating cycle unit algorithm from training to convergence and the XGboost classifier from training to convergence. The three models from training to convergence can be adopted to classify the client event list, and the classification accuracy is further improved.
Example four
Fig. 8 is a flowchart illustrating a method for classifying a customer event ticket according to still another embodiment of the present invention, and as shown in fig. 8, the method for classifying a customer event ticket according to this embodiment further includes, on the basis of the method for classifying a customer event ticket according to the third embodiment, before step 402, a step of determining target historical attribute data in a historical customer event ticket. The method for classifying the customer event list provided by this embodiment further includes the following steps:
step 501, all attribute data and corresponding belonging category labels in each historical client event list are obtained.
All the attribute data in each historical client event list can include the number of the event list, the name of the client, the incoming call number of the client, the certificate number of the client, the gender of the client, the number of the card of the client, the province and city elements, the source of the event, the emergency degree of the event, the incoming call time and the like. The category label corresponding to each historical customer event ticket may be any one of the following category labels: complaint labels, suggestion labels, word labels, help labels.
Step 502, all attribute data are encoded respectively to obtain an attribute feature matrix.
In this embodiment, each attribute data is normalized and encoded, and an attribute feature matrix corresponding to each attribute data is obtained. Wherein the attribute feature matrix corresponding to each attribute data can be represented as xi。
Illustratively, when the gender of the customer is normalized, the male may be coded as 000, the female may be coded as 001, and the unknown gender may be coded as 011.
Step 503, encoding the class label to obtain a class label matrix.
In this embodiment, the category label is also subjected to normalized encoding, and the matrix obtained by encoding is a category label matrix. The category label matrix may be denoted as y.
Step 504, calculating the correlation coefficient of each attribute feature matrix and the category label matrix.
Wherein, each attribute feature matrix and the correlation coefficient can be expressed as shown in formula (13):
wherein, muxRepresenting the mean of the attribute feature matrix. Mu.syMeans, σ, representing class label matrixxThe mean square error of the attribute feature matrix is represented. SigmayThe mean square error of the class label matrix is represented. ρ represents a correlation coefficient.
And 505, determining target historical attribute data in the historical client event list according to the correlation coefficient, so as to determine a feature matrix corresponding to the target historical attribute data as the historical attribute feature corresponding to the training sample.
In this embodiment, the target historical attribute data in the customer event list is historical according to the magnitude of the correlation coefficient. The larger the correlation coefficient is, the stronger the correlation between the attribute data and the tag is, and the smaller the correlation coefficient is, the weaker the correlation between the attribute data and the tag is.
Specifically, a correlation coefficient threshold may be set, the correlation coefficient corresponding to each attribute data is compared with a preset correlation coefficient threshold, and if it is determined that the correlation coefficient is greater than or equal to the correlation coefficient threshold, the attribute is determined to be the target historical attribute data that needs to be retained. And if the attribute data is less than the correlation coefficient threshold value, determining that the attribute data is not the target historical attribute data.
It should be noted that the feature matrix corresponding to the target historical attribute data is determined as the historical attribute feature corresponding to the training sample. Then the obtained target attribute data is also the attribute data of the same type as the target historical attribute data when the target client event ticket is classified.
In the method for classifying the client event tickets provided by the embodiment, before training a preset convolutional neural network, a preset gating cycle unit algorithm and a preset XGBoost classifier by using training samples, all attribute data and corresponding class labels in each historical client event ticket are obtained; respectively encoding all attribute data to obtain an attribute feature matrix; encoding the class label to obtain a class label matrix; calculating a correlation coefficient of each attribute feature matrix and the category label matrix; and determining target historical attribute data in the historical client event list according to the correlation coefficient so as to determine a feature matrix corresponding to the target historical attribute data as the historical attribute feature corresponding to the training sample. Because more attribute data exist in the historical client event list, and the relevance between some attribute data and the category corresponding to the historical client event list is smaller, the target historical attribute data in the historical client event list is determined according to the correlation coefficient of the attribute feature matrix and the category label matrix, so that the determined target historical attribute data is the attribute data with the larger relevance between the categories corresponding to the historical client event list. The dimension of the characteristic can be effectively reduced, and the interference of less-relevant attribute data on the determination of the category corresponding to the historical client event list can be reduced.
As an optional implementation manner, the method for classifying a client event ticket provided in this embodiment, after determining a category corresponding to a target client event ticket according to a target text feature, a target attribute feature and a history relevance feature on the basis of the method for classifying a client event ticket provided in any one of the above embodiments, further includes the following technical solutions:
and acquiring a mapping relation between the type corresponding to each preset customer event list and the processing mechanism and the processing method. And determining a processing mechanism and a processing method corresponding to the target client event list according to the mapping relation. And sending an event list processing instruction to the processing mechanism equipment, wherein the event list processing instruction can comprise a target client event list, a corresponding category and a processing method. The processing mechanism equipment can process the target client event list according to the processing method, and the intellectualization of processing the client event list is improved. And then promote the efficiency that customer's incident list was handled, reduce the manual work link, practice thrift the human cost.
EXAMPLE five
Fig. 9 is a schematic structural diagram of an electronic device for implementing the method for classifying a customer event list according to the embodiment of the present invention, and as shown in fig. 9, an electronic device 60 provided in this embodiment includes: at least one processor 62 and memory 61.
The memory 61 stores computer execution instructions.
The at least one processor 62 executes the computer-executable instructions stored by the memory, causing the at least one processor to perform a method as provided in any one of the first through fourth embodiments.
EXAMPLE six
Fig. 10 is a block diagram of an electronic device, such as a computer, a messaging device, a tablet device, a personal digital assistant, a server cluster, etc., that is used to implement the method for classifying a client event ticket according to an embodiment of the present invention, as shown in fig. 10.
The processing component 702 generally controls overall operation of the electronic device 700, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing components 702 may include one or more processors 720 to execute instructions to perform all or a portion of the steps of the methods described above. Further, the processing component 702 may include one or more modules that facilitate interaction between the processing component 702 and other components. For example, the processing component 702 may include a multimedia module to facilitate interaction between the multimedia component 708 and the processing component 702.
The memory 704 is configured to store various types of data to support operations at the electronic device 700. Examples of such data include instructions for any application or method operating on the electronic device 700, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 704 may be implemented by any type or combination of volatile or non-volatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.
The power supply component 706 provides power to the various components of the electronic device 700. The power components 706 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device 700.
The multimedia component 708 includes a screen that provides an output interface between the electronic device 700 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 708 includes a front facing camera and/or a rear facing camera. The front camera and/or the rear camera may receive external multimedia data when the electronic device 700 is in an operation mode, such as a photographing mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.
The audio component 710 is configured to output and/or input audio signals. For example, the audio component 710 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 700 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal may further be stored in the memory 704 or transmitted via the communication component 716. In some embodiments, audio component 710 also includes a speaker for outputting audio signals.
The I/O interface 712 provides an interface between the processing component 702 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.
The sensor assembly 714 includes one or more sensors for providing various aspects of status assessment for the electronic device 700. For example, the sensor assembly 714 may detect an open/closed state of the electronic device 700, the relative positioning of components, such as a display and keypad of the electronic device 700, the sensor assembly 714 may also detect a change in position of the electronic device 700 or a component of the electronic device 700, the presence or absence of user contact with the electronic device 700, orientation or acceleration/deceleration of the electronic device 700, and a change in temperature of the electronic device 700. The sensor assembly 714 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 714 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
The communication component 716 is configured to facilitate wired or wireless communication between the electronic device 700 and other devices. The electronic device 700 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 716 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 716 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
In an exemplary embodiment, the electronic device 700 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described methods.
In an exemplary embodiment, a non-transitory computer readable storage medium comprising instructions, such as the memory 704 comprising instructions, executable by the processor 720 of the electronic device 700 to perform the above-described method is also provided. 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.
A non-transitory computer readable storage medium, wherein instructions of the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the method provided by any one of the first to fourth embodiments.
In an exemplary embodiment, a computer program product is further provided, which includes a computer program that is executed by a processor to perform the method provided in any one of the first to fourth embodiments.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure 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 present disclosure is limited only by the appended claims.
Claims (11)
1. A method for classifying a customer event ticket, comprising:
acquiring incoming call content data, target attribute data and historical relevance data in a target client event list;
extracting target text characteristics corresponding to the incoming call content data;
determining target attribute characteristics corresponding to the target attribute data;
determining historical relevance characteristics corresponding to the historical relevance data;
and determining the category corresponding to the target client event list according to the target text characteristic, the target attribute characteristic and the historical relevance characteristic.
2. The method of claim 1, wherein the target text feature comprises: target semantic detail features and target context features;
the extracting of the target text features corresponding to the incoming call content data includes:
extracting target semantic detail features corresponding to the incoming call content data by adopting a convolutional neural network trained to be convergent;
and extracting the target context characteristics corresponding to the incoming call content data by adopting a gated loop unit algorithm from training to convergence.
3. The method according to claim 2, wherein before extracting the target text feature corresponding to the incoming call content data, further comprising:
performing text segmentation processing and word segmentation processing on the incoming call content data to obtain a word sequence corresponding to the incoming call content data;
and performing word vector representation on words in the word sequence by adopting a word embedding algorithm to obtain a text matrix consisting of a plurality of word vectors.
4. The method of claim 3, wherein the extracting the target semantic detail features corresponding to the incoming call content data by using the convolutional neural network trained to converge comprises:
inputting the text matrix into a convolutional neural network trained to converge;
performing convolution operation on the text matrix by adopting a plurality of convolution kernels in the convolution neural network from training to convergence to obtain a plurality of convolution characteristic data;
performing maximum pooling operation on the convolution characteristic data corresponding to each convolution kernel to obtain the most significant convolution characteristic corresponding to each convolution kernel;
and determining the most significant convolution features corresponding to each convolution kernel as the target semantic detail features.
5. The method according to claim 3, wherein the extracting the target context feature corresponding to the incoming call content data by adopting the gated loop unit algorithm trained to converge comprises:
inputting the text matrix into a gated cyclic unit algorithm trained to converge;
extracting hidden layer state data corresponding to each word vector in the text matrix by adopting a gated cycle unit algorithm from training to convergence;
performing maximum pooling operation on hidden layer state data corresponding to all word vectors to obtain most obvious hidden layer state data;
determining the most significant hidden layer state data as the target context feature.
6. The method according to any one of claims 1-5, wherein the determining the target attribute characteristics corresponding to the target attribute data comprises:
encoding the target attribute data to obtain corresponding target attribute characteristics;
the determining the historical relevance characteristics corresponding to the historical relevance data includes:
performing matrix representation on the historical relevance data to obtain historical relevance characteristics;
the historical relevance data is category distribution data corresponding to historical client event lists of the target client in different statistical periods.
7. The method according to any one of claims 1-5, wherein the determining the category corresponding to the target customer event ticket according to the target text feature, the target attribute feature and the historical relevance feature comprises:
splicing the target text features, the target attribute features and the historical relevance features to obtain target features corresponding to the target client event list;
inputting the target features into the trained to converge XGboost classifier;
and classifying the target client event list by adopting the XGboost classifier trained to be convergent according to the target characteristics, and outputting the category corresponding to the target client event.
8. The method of any one of claims 1-5, further comprising:
obtaining a training sample, wherein the training sample comprises a historical customer event list and a category label to which the historical customer event list belongs;
and training the preset convolutional neural network, the preset gating cycle unit algorithm and the preset XGboost classifier by using the training samples to obtain the convolutional neural network trained to be convergent, the gating cycle unit algorithm trained to be convergent and the XGboost classifier trained to be convergent.
9. The method of claim 8, wherein before training a predetermined convolutional neural network, a predetermined gated cyclic unit algorithm, and a predetermined XGBoost classifier using the training samples, the method further comprises:
acquiring all attribute data in each historical client event list and corresponding belonging category labels;
respectively encoding all attribute data to obtain an attribute feature matrix;
encoding the class label to obtain a class label matrix;
calculating a correlation coefficient of each attribute feature matrix and the category label matrix;
and determining target historical attribute data in the historical client event list according to the correlation coefficient so as to determine a feature matrix corresponding to the target historical attribute data as the historical attribute feature corresponding to the training sample.
10. An electronic device, comprising: at least one processor and memory
The memory stores computer-executable instructions;
the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the method of any one of claims 1-9.
11. A computer-readable storage medium having computer-executable instructions stored thereon, which when executed by a processor, perform the method of any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111489314.9A CN114168734A (en) | 2021-12-07 | 2021-12-07 | Client event list classification method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111489314.9A CN114168734A (en) | 2021-12-07 | 2021-12-07 | Client event list classification method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114168734A true CN114168734A (en) | 2022-03-11 |
Family
ID=80484246
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111489314.9A Pending CN114168734A (en) | 2021-12-07 | 2021-12-07 | Client event list classification method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114168734A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115618100A (en) * | 2022-09-15 | 2023-01-17 | 中航信移动科技有限公司 | Data processing method, storage medium and electronic device for associated event recommendation |
| CN116886424A (en) * | 2023-08-15 | 2023-10-13 | 哈尔滨雷风恒科技开发有限公司 | Digital transmission security analysis system and method based on big data of computer |
-
2021
- 2021-12-07 CN CN202111489314.9A patent/CN114168734A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115618100A (en) * | 2022-09-15 | 2023-01-17 | 中航信移动科技有限公司 | Data processing method, storage medium and electronic device for associated event recommendation |
| CN115618100B (en) * | 2022-09-15 | 2024-02-06 | 中航信移动科技有限公司 | Data processing method, storage medium and electronic equipment for associated event recommendation |
| CN116886424A (en) * | 2023-08-15 | 2023-10-13 | 哈尔滨雷风恒科技开发有限公司 | Digital transmission security analysis system and method based on big data of computer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110580290B (en) | Method and device for optimizing training set for text classification | |
| CN106202330B (en) | Junk information judgment method and device | |
| CN110602527B (en) | Video processing method, device and storage medium | |
| RU2643500C2 (en) | Method and device for training classifier and recognizing type | |
| AU2016256764A1 (en) | Semantic natural language vector space for image captioning | |
| WO2022166069A1 (en) | Deep learning network determination method and apparatus, and electronic device and storage medium | |
| CN114168734A (en) | Client event list classification method, device, equipment and storage medium | |
| CN111753895A (en) | Data processing method, device and storage medium | |
| CN111210844B (en) | Method, device and equipment for determining speech emotion recognition model and storage medium | |
| CN113095085A (en) | Text emotion recognition method and device, electronic equipment and storage medium | |
| CN110889489A (en) | Neural network training method, image recognition method and device | |
| CN111753917A (en) | Data processing method, device and storage medium | |
| CN113849723A (en) | Search method and search device | |
| CN112948565A (en) | Man-machine conversation method, device, electronic equipment and storage medium | |
| CN112328809A (en) | Entity classification method, device and computer readable storage medium | |
| CN114902249A (en) | Method, system, classification method, system, and medium for generating image recognition model | |
| CN111274389A (en) | Information processing method and device, computer equipment and storage medium | |
| CN116912478A (en) | Object detection model construction, image classification method and electronic equipment | |
| CN111538998A (en) | Text encryption method and device, electronic equipment and computer readable storage medium | |
| CN113486978B (en) | Training method and device for text classification model, electronic equipment and storage medium | |
| CN115718801A (en) | Text processing method, model training method, device, equipment and storage medium | |
| CN115146633A (en) | Keyword identification method and device, electronic equipment and storage medium | |
| CN111400443B (en) | Information processing method, device and storage medium | |
| CN114943336A (en) | Model pruning method, device, equipment and storage medium | |
| CN117648986B (en) | Task processing and code processing method, computing device, medium, and program product |
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 |