CN113779241A

CN113779241A - Information acquisition method and device, computer readable storage medium and electronic equipment

Info

Publication number: CN113779241A
Application number: CN202110265856.1A
Authority: CN
Inventors: 郑吉星
Original assignee: Beijing Jingdong Century Trading Co Ltd; Beijing Wodong Tianjun Information Technology Co Ltd
Current assignee: Beijing Jingdong Century Trading Co Ltd; Beijing Wodong Tianjun Information Technology Co Ltd
Priority date: 2021-03-11
Filing date: 2021-03-11
Publication date: 2021-12-10

Abstract

The disclosure relates to the field of computers, and provides an information acquisition method, an information acquisition device, a computer-readable storage medium and an electronic device, wherein the method comprises the following steps: acquiring a multi-dimensional feature data set corresponding to a service request, wherein the multi-dimensional feature data set comprises a plurality of feature data of a plurality of dimensions; acquiring a feature vector set of a multi-dimensional feature data set, and inputting the feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category; and obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category. According to the method and the device, the target category information of the service request is acquired through the characteristic data of multiple dimensions, and the accuracy of information acquisition is improved.

Description

Information acquisition method and device, computer readable storage medium and electronic equipment

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to an information acquisition method, an information acquisition apparatus, a computer-readable storage medium, and an electronic device.

Background

With the development of computer technology, many organizations such as large-scale enterprises and government departments submit forms of related problems by using computer technology when processing client problems or performing cross-department cooperative processing on problems. The method can clearly record and store the related problems and the processing scheme, and avoids omission of the related problems.

In the prior art, the problem form is filled in by pure hands, or a default option with the most filling times built in mechanically is adopted to help the manual selection. When the options are more and the form filler is not familiar with the options of the related content, a lot of time is spent, and the accuracy is low.

In view of the above, there is a need in the art to develop a new information acquisition method and apparatus.

It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present disclosure, and thus may include information that does not constitute prior art known to those of ordinary skill in the art.

Disclosure of Invention

An object of the present disclosure is to provide an information acquisition method, an information acquisition apparatus, a computer-readable storage medium, and an electronic device, which can improve accuracy and efficiency of information acquisition at least to a certain extent.

Additional features and advantages of the disclosure will be set forth in the detailed description which follows, or in part will be obvious from the description, or may be learned by practice of the disclosure.

According to an aspect of the present disclosure, there is provided an information acquisition method, the method including: acquiring a multi-dimensional feature data set corresponding to a service request, wherein the multi-dimensional feature data set comprises a plurality of feature data of a plurality of dimensions; acquiring a feature vector set of a multi-dimensional feature data set, and inputting the feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category; and obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

In some exemplary embodiments of the present disclosure, the category prediction models include a department category prediction model and a business category prediction model, and the prediction categories include a predicted department category and a predicted business category; inputting a feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category, including: inputting a feature vector set of the multi-dimensional feature data set into the department category prediction model to obtain a prediction department category corresponding to the multi-dimensional feature data set and a first prediction probability corresponding to the prediction department category; and inputting the feature vector set of the multi-dimensional feature data set and the prediction department category into the service category prediction model to obtain a predicted service category corresponding to the multi-dimensional feature data set and a second prediction probability corresponding to the predicted service category.

In some exemplary embodiments of the present disclosure, the predicted department category includes a plurality; inputting the feature vector set of the multi-dimensional feature data set and the prediction department category into the business category prediction model, including: sequencing the plurality of prediction department categories according to a first prediction probability corresponding to the prediction department categories, and acquiring the prediction department categories with the preset number as target prediction department categories; and acquiring a data vector of the target prediction department category, fusing the data vector of the target prediction department category and the feature vector set of the multi-dimensional feature data set to obtain a fused feature vector set, and inputting the fused feature vector set into the service category prediction model.

In some exemplary embodiments of the present disclosure, the predicted traffic class includes a plurality; obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category, wherein the target category information comprises: obtaining a target prediction service class according to a second prediction probability corresponding to each prediction service class; judging whether the target prediction service class and a target prediction department class corresponding to the target prediction service class have an affiliated relationship according to a preset affiliated relationship; and if so, configuring the target prediction business category and the target prediction department category as target category information.

In some exemplary embodiments of the present disclosure, obtaining a feature vector set of a multi-dimensional feature data set includes: and inputting the multi-dimensional feature data set into a vector extraction model to obtain a feature vector set corresponding to the multi-dimensional feature data set.

In some exemplary embodiments of the present disclosure, the feature data in the multi-dimensional feature data set includes discrete feature data, continuous feature data, multi-valued discrete feature data, text feature data; the method further comprises the following steps: if the feature data are the discrete feature data, the continuous feature data and the multi-valued discrete feature data, configuring the vector extraction model as a feature outline; and if the feature data are the text feature data, configuring the vector extraction model into a word vector model.

In some exemplary embodiments of the present disclosure, inputting the multidimensional feature data set into a vector extraction model to obtain a feature vector set corresponding to the multidimensional feature data set, includes: judging whether the feature data in the multi-dimensional feature data set is the text feature data or not; if so, inputting the feature data into the word vector model to obtain a first feature vector corresponding to the feature data; if not, inputting the feature data into the feature outline to obtain a second feature vector corresponding to the feature data; and carrying out vector splicing on the first characteristic vector and the second characteristic vector to obtain a characteristic vector set corresponding to the multi-dimensional characteristic data set.

In some exemplary embodiments of the present disclosure, the word vector model includes three convolutional neural network layers, three maximum pooling layers, and one fully-connected layer, respectively.

In some exemplary embodiments of the present disclosure, the method further comprises: acquiring all text data in a resource conversion platform, and configuring the text data into text data samples; and training a word vector model to be trained through the text data sample to obtain the word vector model.

In some exemplary embodiments of the present disclosure, the method further comprises: acquiring a historical multi-dimensional feature data set corresponding to a historical service request, and acquiring a historical feature vector set corresponding to the historical multi-dimensional feature data set according to a vector extraction model; configuring the historical feature vector set as a feature vector set sample, and configuring target category information corresponding to the historical service request as a category label of the feature vector set sample; and training the class prediction model to be trained through the feature vector set sample and the class label of the feature vector set sample to obtain the class prediction model.

In some exemplary embodiments of the present disclosure, training the to-be-trained class prediction model through the feature vector set sample and the class label of the feature vector set sample to obtain the class prediction model includes: inputting the feature vector sample into the class prediction model to be trained to obtain a sample class corresponding to the feature vector sample; and determining a loss function according to the sample class and the class label, and adjusting the parameters of the class prediction model to be trained until the loss function reaches the minimum value to obtain the class prediction model.

In some exemplary embodiments of the present disclosure, the class prediction model to be trained includes a department class prediction model to be trained and a business class prediction model to be trained; inputting the feature vector sample into the class prediction model to be trained to obtain a sample class corresponding to the feature vector sample, including: inputting the feature vector set sample into the department category prediction model to be trained to obtain a plurality of sample department categories corresponding to the feature vector set sample and first probabilities corresponding to the sample department categories; determining a target sample department category from the plurality of sample department categories according to each of the first probabilities; and inputting the feature vector set sample and the target sample department category into the service category prediction model to be trained to obtain a plurality of sample service categories corresponding to the feature vector set sample and second probabilities corresponding to the sample service categories.

In some exemplary embodiments of the present disclosure, the category label includes a department category label and a business category label; determining a loss function according to the sample class and the class label, and obtaining the class prediction model by adjusting parameters of the class prediction model to be trained until the loss function reaches the minimum, wherein the method comprises the following steps: determining a first loss function according to the sample department category and the department category label, and obtaining the department category prediction model by adjusting parameters of the to-be-trained department category prediction model until the first loss function reaches the minimum; and determining a second loss function according to the sample service class and the service class label, and adjusting the parameters of the service class prediction model to be trained until the second loss function reaches the minimum value to obtain the service class prediction model.

In some exemplary embodiments of the present disclosure, the method further comprises: sending target category information corresponding to the service request to a service window of a terminal, and receiving a correction operation aiming at the target category information, wherein the correction operation comprises real category information; determining a prediction accuracy according to the target category information and the real category information, and judging whether the prediction accuracy is smaller than a first accuracy threshold value; if so, taking the multi-dimensional feature data set as a training sample, taking the real category information as a label corresponding to the training sample, and training the category prediction model through the training sample and the label corresponding to the training sample to obtain an updated category prediction model.

In some exemplary embodiments of the present disclosure, the method further comprises: configuring a flow control switch through which a plurality of the multi-dimensional feature data sets are assigned to the class prediction model and the updated class prediction model; and monitoring the prediction accuracy of the updated category prediction model in real time, and replacing the category prediction model with the updated category prediction model when the prediction accuracy of the updated category prediction model is greater than a second accuracy threshold.

In some exemplary embodiments of the present disclosure, obtaining a multidimensional feature dataset corresponding to a service request includes: and calling real-time data interfaces with multiple dimensions, and acquiring multiple characteristic data corresponding to the service request through each real-time data interface.

According to an aspect of the present disclosure, there is provided an information acquisition apparatus including: the data acquisition module is used for acquiring a multi-dimensional feature data set corresponding to the service request, wherein the multi-dimensional feature data set comprises a plurality of feature data of a plurality of dimensions; the category prediction model is used for acquiring a feature vector set of a multi-dimensional feature data set and inputting the feature vector set of the multi-dimensional feature data set into the category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category; and the information acquisition module is used for acquiring target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

According to an aspect of the present disclosure, there is provided a computer-readable medium on which a computer program is stored, the program, when executed by a processor, implementing the information acquisition method as described in the above embodiments.

According to an aspect of the present disclosure, there is provided an electronic device including: one or more processors; a storage device for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the information acquisition method as described in the above embodiments.

As can be seen from the foregoing technical solutions, the information obtaining method and apparatus, the computer-readable storage medium, and the electronic device in the exemplary embodiments of the present disclosure have at least the following advantages and positive effects:

the information acquisition method comprises the steps of firstly, acquiring a multidimensional characteristic data set corresponding to a service request, wherein the multidimensional characteristic data set comprises a plurality of characteristic data of a plurality of dimensions; then, acquiring a feature vector set of the multi-dimensional feature data set, and inputting the feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category; and finally, obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category. The information acquisition method can acquire the multi-dimensional feature data corresponding to the service request, and obtain the target category information corresponding to the multi-dimensional feature data through the category prediction model, so that the multi-classification problem is solved, the model coverage is enlarged, and the information acquisition efficiency and accuracy are improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

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. It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those drawings by one of ordinary skill in the art without the exercise of inventive faculty.

Fig. 1 schematically shows a flow diagram of an information acquisition method according to an embodiment of the present disclosure;

FIG. 2 schematically shows a flowchart of a method of deriving a feature vector set according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a flow diagram of a method of deriving a prediction class according to an embodiment of the present disclosure;

FIG. 4 schematically shows a flowchart of a method of obtaining target category information according to an embodiment of the present disclosure;

FIG. 5 schematically illustrates a flowchart of a method of deriving a class prediction model according to an embodiment of the present disclosure;

FIG. 6 schematically illustrates a flowchart of a method of training a class prediction model to be trained, according to an embodiment of the present disclosure;

FIG. 7 schematically illustrates a flow diagram of a method of obtaining a sample category according to an embodiment of the present disclosure;

FIG. 8 schematically illustrates a flowchart of a method of obtaining an updated category prediction model, according to an embodiment of the present disclosure;

FIG. 9 schematically shows an architectural diagram of an information acquisition system according to an embodiment of the present disclosure;

fig. 10 schematically shows a block diagram of an information acquisition apparatus according to an embodiment of the present disclosure;

FIG. 11 schematically shows a block schematic diagram of an electronic device according to an embodiment of the present disclosure;

FIG. 12 schematically shows a program product schematic according to an embodiment of the disclosure.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the subject matter of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations, or operations have not been shown or described in detail to avoid obscuring aspects of the disclosure.

The block diagrams shown in the figures are functional entities only and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.

The flow charts shown in the drawings are merely illustrative and do not necessarily include all of the contents and operations/steps, nor do they necessarily have to be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.

In one embodiment of the present disclosure, an information acquisition method is proposed, which can be applied to, but is not limited to, the following scenarios: in the e-commerce platform, pushing a consultation problem generation form of the user to a corresponding production department; or in an enterprise or government complaint platform, relevant departments are determined according to complaint contents of users, and the like, the specific application scenario of the information acquisition method is not specifically limited by the disclosure, and changes of the specific application scenario are understood to all belong to the protection scope of the disclosure. In the embodiment of the present disclosure, taking an application to an e-commerce platform as an example, fig. 1 shows a flow diagram of an information acquisition method, and as shown in fig. 1, the information acquisition method at least includes the following steps:

step S110: acquiring a multidimensional characteristic data set corresponding to the service request, wherein the multidimensional characteristic data set comprises a plurality of characteristic data of a plurality of dimensions;

step S120: acquiring a feature vector set of a multi-dimensional feature data set, and inputting the feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category;

step S130: and obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

The information acquisition method in the embodiment of the disclosure can acquire the multi-dimensional feature data corresponding to the service request, and obtain the target category information corresponding to the multi-dimensional feature data through the category prediction model, thereby solving the multi-classification problem, expanding the coverage of the model, and improving the efficiency and accuracy of information acquisition.

It should be noted that the information acquisition method according to the exemplary embodiment of the present disclosure may be executed by a server, and an information acquisition apparatus corresponding to the information acquisition method may also be configured in the server. In addition, it should be understood that a terminal device (e.g., a mobile phone, a tablet, etc.) may also implement the steps of the information acquisition method, and corresponding apparatuses may also be configured in the terminal device.

In order to make the technical solution of the present disclosure clearer, each step of the information acquisition method is explained next.

In step S110, a multidimensional feature data set corresponding to the service request is obtained, where the multidimensional feature data set includes a plurality of feature data of a plurality of dimensions.

In an exemplary embodiment of the present disclosure, the service request includes a service request triggered by a user at a customer service center of the e-commerce platform, and the service request may be before sale, during sale, or in an after-sale process. For example, after the user completes a certain order, a service request is issued to the customer service center for the order, and the service request may be a question for any aspect of the order, which is not specifically limited by the present disclosure.

If the user presents a plurality of different problems to the customer service center with respect to the same order, a plurality of service requests are generated, and each service request corresponds to an order number. It can thus be seen that the same order or the same goods can correspond to one or more service requests.

In addition, the service request can be triggered and generated by customer service personnel of the customer service center, and when the user provides a problem to the customer service center, if the customer service personnel can provide a reasonable answer to the problem provided by the user, the service request is not triggered; if the customer service staff can not directly provide reasonable answers to the questions posed by the users, the customer service staff needs to query a database corresponding to the target category information of the questions or send the questions to the staff corresponding to the target category information of the questions for processing, and then the customer service staff can trigger the service request. Wherein the database may be a distributed file system.

In an exemplary embodiment of the present disclosure, the multi-dimensional feature data set includes a plurality of feature data of a plurality of dimensions. The multiple dimensions include dimensions related to orders and issues on the e-commerce platform, for example, the multiple dimensions may include an order dimension, an after-market dimension, an invoice dimension, an event dimension, and a logistics dimension. The event dimension is based on the questions posed by the user and the communication process between the user and the customer service staff.

In addition, the feature data corresponding to each dimension respectively includes: the order dimension comprises a sales order type code, an item price, an order comprehensive state code, an order mark, order completion time, order delivery timeliness, whether an order is printed or not, whether the order is picked up or not, whether the order is packaged or not, whether the order is delivered from a warehouse or not, three items recently purchased by a user and the like; the after-sales dimension comprises an after-sales service single link code, an after-sales service single state code, an audit result code, a processing result code, an arbitration state, dispute or not and the like; the invoice dimensionality comprises an invoice state, an invoice type, an invoice head raising type, invoice quality increasing verification, whether the invoice is mailed or not, whether an invoice issuing mechanism is the same as a selling mechanism or not and the like; the event summary comprises an event summary, an event type, an event state, an event source, a special question ID, an event summary log, and the like; the logistics dimension comprises a waybill category code, a warehouse number, a delivery date expected by a user, a delivery event, a routing node where the logistics is located, whether to sort and deliver goods for the first time, a real operation point, a real operation type and the like.

In an exemplary embodiment of the disclosure, real-time data interfaces of multiple dimensions are called, and multiple feature data corresponding to a service request are acquired through each real-time data interface.

Specifically, a service request is received, and a multi-dimensional feature data set is obtained in real time according to the service request. When a user purchases goods and consults a problem on an e-commerce platform, the server stores the feature data of multiple dimensions of the same order into the storage units in real time, and if the feature data of a certain dimension changes, the feature data of a certain dimension corresponding to the order can be updated in real time. For example, after the order is completed, the characteristic data of the logistics dimension can be updated in real time along with the goods transportation condition; or in the event dimension, the characteristic of the event summary is updated in real time in the communication process of the user and the customer service personnel.

Therefore, in the online application stage, the service request is responded, the real-time data interfaces of multiple dimensions are called according to the order number corresponding to the service request, the multiple feature data of the multiple dimensions corresponding to the order number are obtained through the real-time data interfaces of the multiple dimensions respectively, and the multiple feature data of the multiple dimensions generate the multi-dimensional feature data set.

In addition, the service request corresponds to a work order number, the work order number is the unique identifier of the service request, a user or customer service personnel can send out the service request at one or more moments aiming at one order number, and one order number can correspond to a plurality of service requests. Therefore, the work order number may be composed of the order number and the current timestamp of the service request, and may be any character string that can uniquely identify the service request.

And after the multidimensional characteristic data set corresponding to the service request is obtained, correspondingly storing the work order number corresponding to the service request and the corresponding multidimensional characteristic data set into a database. So that the multidimensional feature data set corresponding to the service request can be queried through the work order number and further can be used for training the category prediction model.

In an exemplary embodiment of the present disclosure, after obtaining the multidimensional feature dataset corresponding to the service request, the multidimensional feature dataset may be preprocessed. Wherein the pre-processing may comprise: and judging whether the characteristic data corresponding to the event dimensionality in the multi-dimensional characteristic data set is empty, and if the characteristic data corresponding to the event dimensionality in the multi-dimensional characteristic data set is empty, discarding the multi-dimensional characteristic data set. In addition, the pre-processing may further include: the data with empty feature data is supplemented, and the supplement mode of the feature data and the preprocessing mode of the feature data are not particularly limited in the present disclosure.

In step S120, a feature vector set of the multi-dimensional feature data set is obtained, and the feature vector set of the multi-dimensional feature data set is input into the category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category.

In an exemplary embodiment of the present disclosure, the prediction category refers to a form category corresponding to the service request, the form category may include a plurality of department categories, and the form category may also include a plurality of business categories. In addition, the form categories may also include major categories and minor categories, each major category including a plurality of minor categories, the major categories referring to department categories, and each department category may include a plurality of business categories.

For example, in an e-commerce platform, department categories may include logistics, spare part warehouse, sales, large electronic, billing, and so on. The logistics distribution class can comprise business classes such as customer order promotion, goods damage, consultation self-service and the like; the spare part library class may include: business categories such as few pieces, wrong pieces, damaged pieces, order cancellation, order distribution and wrong delivery; the warehouses may include: service categories such as damaged, few products B, hasten bills A and the like; the mining and marketing classes may include: the manufacturer directly sends service categories such as order problems, three bags after sale, maintenance, goods arrival consultation and the like; large electrical classes may include: business categories such as a forward order prompting problem, an opening and checking machine, a reverse order prompting problem and the like; the tickets may include: the invoice self-service order entry is unsuccessful, and the electronic invoice, the historical invoicing common ticket and other business categories are provided. Of course, the specific department category and business category may be set according to the application scenario, which is not specifically limited by the present disclosure.

In an exemplary embodiment of the present disclosure, the feature data in the multi-dimensional feature data set includes discrete feature data, continuous feature data, multi-valued discrete feature data, text feature data. The discrete characteristic data, the continuous characteristic data, the multi-valued discrete characteristic data and the text characteristic data are in a data format of the characteristic data.

For example, in the feature data of multiple dimensions listed in the above-described embodiment, the audit result code of the after-sale dimension belongs to discrete feature data, three items recently purchased by the user of the order dimension belong to multi-valued discrete feature data, the item price of the order dimension belongs to continuous feature data, the summary of the event in the event dimension belongs to text feature data, and the like. The embodiment of the present disclosure exemplifies the above examples for discrete feature data, continuous feature data, multi-valued discrete feature data, and text feature data, but the present disclosure is not limited to this, as a matter of course, the plurality of feature data further includes a plurality of feature data which are discrete feature data, continuous feature data, multi-valued discrete feature data, and text feature data, respectively.

In addition, since the feature data of multiple dimensions includes multiple data formats, the feature data of multiple dimensions needs to be converted into a specific format, and different data conversion methods are provided for different data formats. The specific format comprises the step of uniformly converting each data format into a vector form.

In an exemplary embodiment of the present disclosure, a multi-dimensional feature data set is input into a vector extraction model to obtain a feature vector set corresponding to the multi-dimensional feature data set. The feature data in the multi-dimensional feature data set comprises several data formats, namely discrete feature data, continuous feature data, multi-valued discrete feature data and text feature data. Therefore, the vector extraction model can be configured according to different data formats of the feature data, and if the feature data are discrete feature data, continuous feature data and multi-valued discrete feature data, the vector extraction model is configured into a feature outline; and if the feature data are text feature data, configuring the vector extraction model into a word vector model.

Specifically, fig. 2 shows a schematic flow chart of the method for obtaining the feature vector set, as shown in fig. 2, the flow at least includes steps S210 to S240, and the following is described in detail:

in step S210, it is determined whether the feature data in the multi-dimensional feature data set is text feature data.

In an exemplary embodiment of the present disclosure, whether a plurality of feature data in the multi-dimensional feature data set are text feature data may be respectively determined according to a data format of the feature data.

In step S220, if the feature data is text feature data, the feature data is input into the word vector model to obtain a first feature vector corresponding to the feature data.

In an exemplary embodiment of the disclosure, the vector dimension of the text feature data may be configured by using the feature outline, and after the text feature data is processed by using the feature outline, the word vector model is input to obtain the feature vector corresponding to the text feature data.

In an exemplary embodiment of the present disclosure, the word vector model may be a word2vec (word to vector) model. The word2vec model can be trained by using an encyclopedia data set, and is trained based on abundant corpora in the encyclopedia, but most corpora are unrelated to the service scene aiming at the specific service scene, so that when the word vector model to be trained is too large, the time for loading the model when the word vector model to be trained is called is consumed, and the word vector model obtained by training cannot better focus on the service scene.

Therefore, in the exemplary embodiment of the present disclosure, when the word vector model to be trained is trained, all text data in the resource conversion platform are acquired, and the text data is configured as a text data sample; and training the word vector model to be trained through the text data sample to obtain the word vector model.

The resource conversion platform comprises an e-commerce platform and acquires all text data in the e-commerce platform. All text data may cover all texts in the business scenario, for example, a search text of the user on the e-commerce platform, an evaluation text of the user for an article or a merchant, a description text of the article when the merchant is registered, a rule text of the e-commerce platform, text feature data in a history multi-dimensional feature data set, and the like, which is not specifically limited by the present disclosure.

In an exemplary embodiment of the present disclosure, the word vector model may be composed of three convolutional neural network layers, three maximum pooling layers, and one full connection layer, and a drop operation may also be added in the word vector model. The word vector model in this embodiment is used to process text feature data, and a feature vector corresponding to the text feature data is obtained after processing through a multi-layer network structure. According to the method, useless data are deleted to a great extent, the small data size of the trained word vector model is guaranteed, the time for calling the word vector model is shorter, meanwhile, the word vector model is enabled to focus on text feature data in business, and the feature vector corresponding to the obtained text feature data is more accurate.

In step S230, if the feature data is not text feature data, the feature data is input into the feature outline to obtain a second feature vector corresponding to the feature data.

In the exemplary embodiment of the present disclosure, in order to facilitate format conversion of feature data in different data formats, after the multi-dimensional feature data set is obtained, data conversion is performed on the feature data in the multi-dimensional feature data set by using the feature schema.

Specifically, data correspondence is performed for discrete feature data. For example, the after-sale dimension audit result is subjected to data correspondence, and if the audit is passed, the data correspondence is 0; if the verification is in progress, corresponding to 1; if the audit is not passed, the result corresponds to 2.

The correspondence rule for the multivalued discrete data is the same as that for the discrete data. For example, a user of an order dimension has recently purchased three items, item a, corresponding to 0; item B corresponds to 1, item C corresponds to 2.

And (4) storing the average value of all values under the characteristic data aiming at the continuous data. For example, the item price of the order dimension, averages the multiple consecutive item prices to 115.6.

In addition, the feature schema can also configure the vector dimension size of the text feature data. For example, the vector dimension size of the event summary belonging to the text feature data in the event dimension is configured to be 300 dimensions.

After the multidimensional feature data set in this embodiment is subjected to feature outline processing, the discrete feature data stores the number of different values taken by each feature data and the one-to-one correspondence relationship between the feature data and the feature vectors; the multi-value discrete data and the text characteristic data have the same structure, and the size of a word list, the maximum text length and the vector dimension are required to be stored; the continuous feature data stores the mean value of each continuous feature data, and the corresponding relationship.

In step S240, the first feature vector and the second feature vector are vector-spliced to obtain a feature vector set corresponding to the multi-dimensional feature data set.

In an exemplary embodiment of the present disclosure, vector splicing is performed on a plurality of first feature vectors and second feature vectors corresponding to a plurality of feature data in a multi-dimensional feature data set to obtain a feature vector set corresponding to the multi-dimensional feature data set.

In an exemplary embodiment of the present disclosure, the category prediction models include a department category prediction model and a business category prediction model. The department type prediction model and the service type prediction model can be ultra-Deep factorization model (xDeepFM), and the xDeepFM model combines CIN, a linear regression unit and a fully-connected neural network unit by taking the design of Wide & Deep and DeepFM and other models as reference. The xDeepFM model inherits the memory capability of the existing line model, inherits the display crossing of the multi-dimensional characteristic data set, simultaneously considers the implicit characteristic crossing and generalization capability of the DNN network, and adopts pooling calculation to reduce the dimension in the CIN network, thereby effectively avoiding the condition of dimension explosion.

In an exemplary embodiment of the present disclosure, the prediction category includes a prediction department category and a prediction business category, and fig. 3 is a schematic flowchart of a method for obtaining the prediction category, where the flowchart at least includes steps S310 to S320, and the following is described in detail:

in step S310, a feature vector set of the multi-dimensional feature data set is input into the department category prediction model to obtain a predicted department category corresponding to the multi-dimensional feature data set and a first prediction probability corresponding to the predicted department category.

In an exemplary embodiment of the present disclosure, the first prediction probability represents a probability that a business department category corresponding to the multi-dimensional feature data set is a real department category, and the greater the first prediction probability is, the greater the probability that the prediction department category is the real department category is.

In an exemplary embodiment of the present disclosure, a department category prediction model is used to obtain a plurality of predicted department categories corresponding to the multi-dimensional feature dataset. When the multi-dimensional feature data set and the prediction department categories are input into the business category prediction model, all the prediction department categories can be input into the business category prediction model, and a plurality of prediction business categories corresponding to the prediction department categories can be obtained.

In addition, the types of the prediction departments can be screened according to the first prediction probabilities corresponding to the types of the prediction departments, the type of the prediction department with the higher first prediction probability is used as the type of the target prediction department, and the type of the target prediction department is input into the business type prediction model.

Specifically, firstly, a plurality of prediction department categories are ranked according to a first prediction probability corresponding to the prediction department categories, and a preset number of prediction department categories are acquired as target prediction department categories.

The multiple prediction department categories are ranked according to the first prediction probability, the first prediction probability corresponding to the prediction department category ranked in the front is high, and the first prediction probability corresponding to the prediction department category ranked in the back is low. In addition, the preset number can be set according to the actual situation, for example, the preset number can be set to 1, that is, the prediction department category with the highest first prediction probability is selected as the target prediction department category; the preset number may also be set to 3, that is, the prediction department category with the first prediction probability at TOP (3) is selected as the target prediction department category, and the larger the value of the preset number is, the more the obtained target prediction department categories are, the larger the subsequent calculation amount is, and the disclosure does not specifically limit this.

And then, acquiring a data vector of the category of the target prediction department, fusing the data vector of the category of the target prediction department and a feature vector set of the multi-dimensional feature data set to obtain a fused feature vector set, and inputting the fused feature vector set into a service category prediction model. And vectorizing the target prediction department category to obtain a data vector of the target prediction department category.

In step S320, the feature vector set and the prediction department category of the multi-dimensional feature data set are input into the service category prediction model to obtain a predicted service category corresponding to the multi-dimensional feature data set and a second prediction probability corresponding to the predicted service category.

In an exemplary embodiment of the present disclosure, a feature vector set of a multi-dimensional feature data set is fused with data vectors of categories of a prediction department to obtain a fused feature vector set, and the fused feature vector set is input into a service category prediction model.

In an exemplary embodiment of the present disclosure, a data vector of a target prediction department category and a feature vector set of a multi-dimensional feature data set are fused to obtain a fused feature vector set, and the fused feature vector set is input into a service category prediction model. Specifically, fusion is performed according to a preset number of target prediction department categories and multi-dimensional feature data sets to obtain a preset number of fusion feature vector sets, the preset number of fusion feature vector sets are input into the service category prediction model, and a plurality of prediction service categories corresponding to the fusion feature vector sets and a second prediction probability corresponding to each prediction service category are obtained.

In an exemplary embodiment of the disclosure, the second prediction probability represents a probability that a predicted service class corresponding to the multidimensional feature data set is a real service class, and the greater the second prediction probability is, the greater the probability that the predicted service class is the real service class is.

In step S130, target category information corresponding to the service request is obtained according to the prediction category and the prediction probability corresponding to the prediction category.

In an exemplary embodiment of the present disclosure, the prediction category is a prediction department category or a prediction business category. And judging whether the prediction probability corresponding to the prediction type is greater than a prediction probability threshold value or not, and if the prediction probability is greater than the prediction probability threshold value, taking the prediction type as target type information corresponding to the service request.

Specifically, if the prediction probabilities corresponding to the prediction categories are greater than the prediction probability threshold, the service request corresponds to the target category information. It indicates that the service request is associated with multiple department categories or multiple business categories.

In an exemplary embodiment of the present disclosure, the prediction categories include a prediction department category and a prediction business category. Fig. 4 is a schematic flowchart of a method for obtaining object category information, where the flowchart at least includes step S410 to step S430, and the following details are introduced:

in step S410, a target predicted service class is obtained according to the second prediction probability corresponding to each predicted service class.

In an exemplary embodiment of the present disclosure, the plurality of predicted traffic classes are sorted according to the second prediction probability corresponding to each predicted traffic class, and the predicted traffic class with the second prediction probability at top (n) is obtained as the target predicted traffic class. The value of n can be set according to actual conditions, and the larger the value of n is, the more the obtained target predicted service types are, the larger the subsequent calculation amount is, and the disclosure does not specifically limit this.

In step S420, it is determined whether the target predicted service category and the target prediction department category corresponding to the target predicted service category have an affiliation relationship according to a preset affiliation relationship.

In the exemplary embodiment of the present disclosure, since each department category includes a plurality of business categories, the department category and the business category have an affiliation, and the affiliation between the department category and the business category may be preset and stored in the database in a distributed format. For example, the department category and the business category are stored in the form of a key-value key value pair, the department category is used as a key value, and the business category is used as a value.

Specifically, the determining whether the target prediction business category and the target prediction department category corresponding to the target prediction business category have an affiliation relationship may include: acquiring all value values of the target prediction department category, and respectively matching the target prediction service category with all value values of the target prediction department category; if the target prediction business category is matched with the value of the target prediction department category, judging that the target prediction business category and the target prediction department category have an affiliated relationship; and if the value values of the target prediction business category and the target prediction department category are not matched, judging that the target prediction business category and the target prediction department category have no relationship.

In step S430, if the target predicted business category and the target predicted department category have an affiliated relationship, the target predicted business category and the target predicted department category are configured as target category information.

In an exemplary embodiment of the present disclosure, the target category information corresponding to the service request includes a target prediction business category and a target prediction department category, and the target prediction department category and the target prediction business category may include one or more than one. The number of the target prediction department categories and the number of the target prediction business categories can be set according to actual conditions, for example, for a more complex service request, the service request may not be solved by one department or business, and the service request may correspond to a plurality of departments or businesses, so that the number of the target prediction department categories and the number of the target prediction business categories can be set to be a plurality. And storing the service request and the target category information corresponding to the service request into a database.

In an exemplary embodiment of the disclosure, fig. 5 is a schematic flowchart of a method for obtaining a category prediction model, where the flowchart at least includes step S510 to step S530, and the following is described in detail:

in step S510, a historical multidimensional feature data set corresponding to the historical service request is obtained, and a historical feature vector set corresponding to the historical multidimensional feature data set is obtained according to the vector extraction model.

In an exemplary embodiment of the disclosure, a historical multidimensional feature data set corresponding to a work order number is obtained in a database according to the work order number corresponding to a historical service request.

In step S520, the historical feature vector set is configured as a feature vector set sample, and the target category information corresponding to the historical service request is configured as a category label of the feature vector set sample.

In an exemplary embodiment of the present disclosure, target category information corresponding to a history multi-dimensional feature data set is obtained in a database, and the target category information is configured as a category label corresponding to a feature vector set of the multi-dimensional feature data set.

In step S530, the to-be-trained class prediction model is trained through the feature vector set sample and the class label of the feature vector set sample, so as to obtain a class prediction model.

In an exemplary embodiment of the disclosure, fig. 6 is a schematic flowchart illustrating a method for training a class prediction model to be trained, and as shown in fig. 6, the flow at least includes steps S610 to S620, which are described in detail as follows:

in step S610, the feature vector samples are input into the class prediction model to be trained to obtain the sample class corresponding to the feature vector samples.

In an exemplary embodiment of the disclosure, the class prediction model to be trained comprises a department class prediction model to be trained and a business class prediction model to be trained, and the feature vector sample is input into the department class prediction model to be trained to obtain a sample department class corresponding to the feature vector sample; and inputting the department class of the sample and the characteristic vector sample into a service class prediction model to be trained to obtain a sample service class corresponding to the characteristic vector sample.

In an exemplary embodiment of the disclosure, fig. 7 is a schematic flowchart illustrating a method for obtaining a sample category, and as shown in fig. 7, the flowchart at least includes steps S710 to S730, which are described in detail as follows:

in step S710, the feature vector set sample is input into the department category prediction model to be trained, so as to obtain a plurality of sample department categories corresponding to the feature vector set sample and a first probability corresponding to each sample department category.

In step S720, a target sample department category is determined among the plurality of sample department categories according to the respective first probabilities.

In an exemplary embodiment of the present disclosure, the sample department categories are ranked according to the first probabilities corresponding to the plurality of sample department categories, and the top m sample department categories are taken as target sample department categories. The value of m may be set according to actual conditions, which is not specifically limited by the present disclosure.

In step S730, the feature vector set sample and the target sample department category are input into the service category prediction model to be trained, so as to obtain a plurality of sample service categories corresponding to the feature vector set sample and a second probability corresponding to each sample service category.

In the exemplary embodiment of the disclosure, a feature vector sample of a target sample department category is obtained, feature fusion is performed on the feature vector set sample and the feature vector sample of the target sample department category to obtain a fused feature vector set sample, and the fused feature vector set sample is input into a service category prediction model to be trained.

In step S620, a loss function is determined according to the sample class and the class label, and the class prediction model is obtained by adjusting parameters of the class prediction model to be trained until the loss function reaches the minimum.

In an exemplary embodiment of the present disclosure, the category labels include department category labels and business category labels, the department category prediction model is determined by the sample department category and department category labels, respectively, and the business category prediction model is determined by the sample business category and business category labels.

Specifically, a first loss function is determined according to the class of the department and the class label of the department of the sample, and the class prediction model of the department is obtained by adjusting the parameters of the class prediction model of the department to be trained until the first loss function reaches the minimum.

And determining a second loss function according to the sample service class and the service class label, and adjusting the parameters of the service class prediction model to be trained until the second loss function reaches the minimum value to obtain the service class prediction model.

The first loss function and the second loss function are both distillation loss functions, and the formula of the distillation loss function is shown in formula (1):

L＝(1-α)H+αD (1)

wherein, L is a distillation loss function, alpha is a parameter, H is a cross entropy loss function, and D is KL divergence, and the specific formula is as follows:

wherein, P is a class label of the sample, q is a prediction class of the sample, x is a sample identifier, n represents the nth sample, and T is a parameter.

In the past training process, only the difference between a prediction class with high probability and a real class is concerned by using a cross entropy loss function, and the similarity is completely constructed by a sufficient number of samples. However, small probability results contribute very little to the loss function. Therefore, the loss function of the present embodiment uses a distillation loss function instead of a common cross entropy loss function, and amplifies the loss values corresponding to the probability values of other classes when calculating the loss function, and the distillation loss function introduces a parameter T, so that the prediction classes of different classes are considered equally by adjusting the size of T.

In an exemplary embodiment of the present disclosure, the class prediction model is deployed to the server in an HTTP manner by tensrflow Serving. And in order to ensure that the input data format of the category prediction model is correct, after the multi-dimensional feature data set is obtained, a feature vector set corresponding to the multi-dimensional feature data set is obtained according to the vector extraction model. Namely, the feature outline and the word vector model are stored in the redis cache, so that a feature vector set of the multi-dimensional feature data set can be rapidly acquired, and timeliness of data request and return is guaranteed.

In an exemplary embodiment of the present disclosure, since there are more department categories and business categories, there is a possibility that there will be more or less department categories and business categories while running online. Therefore, the offline training data and the online real-time data are not aligned, and it can be specifically shown that the online training accuracy is higher than the online accuracy, and in order to ensure the accuracy of online class prediction, the accuracy of the class prediction model needs to be monitored in real time after the class prediction model is online, and the class prediction model needs to be updated.

Fig. 8 is a schematic flow chart of a method for obtaining an updated category prediction model, which includes at least steps S810 to S830, and is described in detail as follows:

in step S810, target category information corresponding to the service request is sent to a service window of the terminal, and a correction operation for the target category information is received, where the correction operation includes real category information.

In an exemplary embodiment of the present disclosure, the terminal refers to a terminal device where the customer service staff is located, and displays the predicted target category information in a service window, so that the customer service staff can check the target category information. If the verification is passed, the customer service personnel can click the confirmation operation on the business window, and the server configures the target class information into real class information after receiving the confirmation operation; if the verification is not passed, the customer service personnel can perform correction operation on the service window, the correction operation can be to input real category information or to modify target category information, and the server receives the real category information corresponding to the correction operation. In addition, the embedded points can be added, the real category information and the target category information can be stored, and the multidimensional characteristic data set acquired by the real-time data interface can be acquired, so that the calculation efficiency of the prediction accuracy can be improved.

In step S820, a prediction accuracy is determined according to the target category information and the real category information, and whether the prediction accuracy is smaller than a first accuracy threshold is determined.

In the exemplary embodiment of the disclosure, the total number of service requests sent within a preset time period is counted, the number of target category information corresponding to the service requests, which is the same as the number of real category information, is counted, and the total number is divided by the same number to obtain the prediction accuracy.

The preset time period may be set according to actual conditions, for example, may be set to one minute or five minutes, which is not specifically limited in this disclosure. In addition, the first accuracy threshold may also be set according to actual situations, for example, may be set to 0.9, and may also be set to 0.95, which is not specifically limited by the present disclosure.

In step S830, if the prediction accuracy is smaller than the first accuracy threshold, the multi-dimensional feature data set is used as a training sample, the real category information is used as a label corresponding to the training sample, and the category prediction model is trained through the training sample and the label corresponding to the training sample to obtain an updated category prediction model.

In an exemplary embodiment of the present disclosure, if the prediction accuracy is smaller than the first accuracy threshold, the multidimensional feature data set acquired by the real-time data interface is used as a training sample, and the category prediction model is trained online. The process of performing online training on the class prediction model is the same as the training process of the class prediction model to be trained in the above embodiments, and is not described herein again.

In the exemplary embodiment of the present disclosure, after the updated category prediction model is obtained, the category prediction model needs to be replaced, but in order to ensure the stability of the updated category prediction model, the original category prediction model needs to be replaced all after online verification.

Specifically, a flow control switch is configured, and a plurality of multi-dimensional feature data sets are distributed to the category prediction model and the updated category prediction model through the flow control switch; and monitoring the prediction accuracy of the updated category prediction model in real time, and replacing the category prediction model with the updated category prediction model when the prediction accuracy of the updated category prediction model is greater than a second accuracy threshold.

The flow control switch can be a random function, the random function can generate a random number between 0 and 1, the random number is used as the traffic ratio of the updating type prediction model, the on-line traffic is distributed to the updating type prediction model according to the traffic ratio, and the prediction accuracy of the updating type prediction model is monitored in real time. The second accuracy threshold may be set according to practical situations, for example, may be set to 0.9, or may be set to 0.95, which is not specifically limited by the present disclosure.

In addition, the first accuracy threshold and the second accuracy threshold both correspond to the prediction accuracy of the category prediction model, but the first accuracy threshold and the second accuracy threshold are applied to different scenarios in this embodiment. And if the prediction accuracy of the category prediction model is smaller than the first accuracy threshold, performing online training on the category prediction model to obtain an updated category prediction model. And controlling the flow of the category prediction model and the updated category prediction model by using the flow control switch, wherein the second accuracy threshold is used for judging whether all the flow needs to be distributed to the updated category prediction model, and if the prediction accuracy of the updated category prediction model is greater than the second accuracy threshold, the updated category prediction model is used for replacing the category prediction model, namely, all the flow is distributed to the updated category prediction model. Of course, the first accuracy threshold and the second accuracy threshold may be the same or different. For example, the first accuracy threshold is 90% and the second accuracy threshold is 98%. And when the accuracy of the category prediction model is less than 90%, performing online training, performing category prediction by using the category prediction model and an updated category prediction model obtained by online training, and replacing the category prediction model with a new updated category prediction model when the accuracy of the updated category prediction model is more than 98%.

The method of the embodiment avoids the hysteresis of the prediction accuracy of the statistical type prediction model, greatly reduces the unstable factors of the online model, enables the online model to be switched more smoothly, better guarantees the stable operation of the online model, and enables a user to have no perception of the online model.

In an exemplary embodiment of the present disclosure, the information acquisition method is applied to an information acquisition system, and fig. 9 shows an architecture diagram of the information acquisition system, and as shown in fig. 9, the information acquisition system includes a model side, a data side, and a service side.

On the model side, firstly, acquiring a historical multi-dimensional feature data set from a distributed file system, preprocessing the historical multi-dimensional feature data set, and deleting useless data;

then, model training is performed, including: acquiring text data on an e-commerce platform, and training a word vector model to be trained by using the text data to obtain a word vector model; obtaining a feature vector set sample corresponding to the historical multi-dimensional feature data set by using the feature outline and the word vector model, and training the to-be-trained category prediction model by using the feature vector set sample and the target category information of the historical multi-dimensional feature data set to obtain a category prediction model;

and finally, deploying the category prediction model in a server, and storing the feature outline and the word vector model in a redis cache.

On the data side, firstly, calling a real-time data interface, acquiring a multidimensional characteristic data set corresponding to a service request, and storing the multidimensional characteristic data set in a distributed file system;

then, calling a vector extraction model in the cache, wherein the vector extraction model comprises a feature outline and a word vector model, and performing data conversion on the multi-dimensional feature data set by using the vector extraction model to obtain a feature vector set of the multi-dimensional feature data set;

and finally, inputting the feature vector set into a category prediction model to obtain a prediction category and prediction probability corresponding to the prediction category, and obtaining target category information of the service request according to the prediction category and the prediction probability.

And at the service side, sending the target category information to a service window corresponding to the service request, receiving real category information corresponding to the service request, adding a buried point to the service side, and storing the target category information, the real category information and the multi-dimensional feature data set.

In addition, the whole information acquisition system is monitored in real time, the prediction accuracy of the category prediction model is calculated, when the prediction accuracy is lower than the prediction accuracy threshold, the category prediction model is trained on line to obtain an updated category prediction model, and the category prediction model is updated by the updated category prediction model through the flow control switch.

Those skilled in the art will appreciate that all or part of the steps implementing the above embodiments are implemented as computer programs executed by a CPU. The computer program, when executed by the CPU, performs the functions defined by the method provided by the present invention. The program may be stored in a computer readable storage medium, which may be a read-only memory, a magnetic or optical disk, or the like.

Furthermore, it should be noted that the above-mentioned figures are only schematic illustrations of the processes involved in the method according to exemplary embodiments of the invention, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.

The following describes embodiments of the apparatus of the present disclosure, which may be used to perform the above-mentioned information acquisition method of the present disclosure. For details that are not disclosed in the embodiments of the apparatus of the present disclosure, refer to the embodiments of the information obtaining method described above in the present disclosure.

Fig. 10 schematically shows a block diagram of an information acquisition apparatus according to one embodiment of the present disclosure.

Referring to fig. 10, an information acquisition apparatus 1000 according to an embodiment of the present disclosure, the information acquisition apparatus 1000 includes: a data acquisition module 1001, a category prediction module 1002, and an information acquisition module 1003. Specifically, the method comprises the following steps:

a data obtaining module 1001, configured to obtain a multidimensional feature data set corresponding to a service request, where the multidimensional feature data set includes multiple feature data of multiple dimensions;

the category prediction module 1002 is configured to obtain a feature vector set of a multidimensional feature data set, and input the feature vector set of the multidimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multidimensional feature data set and a prediction probability corresponding to the prediction category;

the information obtaining module 1003 is configured to obtain target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

In an exemplary embodiment of the present disclosure, the data obtaining module 1001 may further be configured to invoke real-time data interfaces of multiple dimensions, and obtain, through each real-time data interface, multiple feature data corresponding to the service request.

In an exemplary embodiment of the present disclosure, the category prediction models include a department category prediction model and a business category prediction model, and the prediction categories include a prediction department category and a prediction business category. The category prediction module 1002 may further be configured to input a feature vector set of the multidimensional feature data set into a department category prediction model, so as to obtain a prediction department category corresponding to the multidimensional feature data set and a first prediction probability corresponding to the prediction department category; and inputting the feature vector set of the multi-dimensional feature data set and the prediction department category into a service category prediction model to obtain a prediction service category corresponding to the multi-dimensional feature data set and a second prediction probability corresponding to the prediction service category.

In an exemplary embodiment of the present disclosure, the prediction section category includes a plurality. The category prediction module 1002 may be further configured to sort the plurality of prediction department categories according to the first prediction probabilities corresponding to the prediction department categories, and acquire a preset number of prediction department categories as target prediction department categories; acquiring a data vector of a target prediction department category, fusing the data vector of the target prediction department category and a feature vector set of a multi-dimensional feature data set to obtain a fused feature vector set, and inputting the fused feature vector set into a service category prediction model.

In an exemplary embodiment of the present disclosure, the predicted traffic class includes a plurality. The category prediction module 1002 may be further configured to obtain a target predicted service category according to a second prediction probability corresponding to each predicted service category; judging whether the target prediction service class and a target prediction department class corresponding to the target prediction service class have an affiliated relationship according to a preset affiliated relationship; and if so, configuring the target prediction business category and the target prediction department category as target category information.

In an exemplary embodiment of the present disclosure, the category prediction module 1002 may further be configured to input the multidimensional feature data set into a vector extraction model to obtain a feature vector set corresponding to the multidimensional feature data set.

In an exemplary embodiment of the present disclosure, the feature data in the multi-dimensional feature data set includes discrete feature data, continuous feature data, multi-valued discrete feature data, text feature data. The category prediction module 1002 may be further configured to configure the vector extraction model as a feature outline if the feature data is discrete feature data, continuous feature data, and multi-valued discrete feature data; and if the feature data are text feature data, configuring the vector extraction model into a word vector model.

In an exemplary embodiment of the present disclosure, the category prediction module 1002 may be further configured to determine whether feature data in the multi-dimensional feature data set is text feature data; if so, inputting the feature data into the word vector model to obtain a first feature vector corresponding to the feature data; if not, inputting the feature data into the feature outline to obtain a second feature vector corresponding to the feature data; and carrying out vector splicing on the first feature vector and the second feature vector to obtain a feature vector set corresponding to the multi-dimensional feature data set.

In an exemplary embodiment of the present disclosure, the category prediction module 1002 may be further configured to use the word vector model to include three convolutional neural network layers, three maximum pooling layers, and one fully-connected layer, respectively.

In an exemplary embodiment of the present disclosure, the information obtaining apparatus 1000 further includes a word vector model training module (not shown in the figure), which is configured to obtain all text data in the resource conversion platform, and configure the text data into text data samples; and training the word vector model to be trained through the text data sample to obtain the word vector model.

In an exemplary embodiment of the present disclosure, the information obtaining apparatus 1000 further includes a category prediction model training module (not shown in the figure), which is configured to obtain a historical multidimensional feature data set corresponding to the historical service request, and obtain a historical feature vector set corresponding to the historical multidimensional feature data set according to the vector extraction model; configuring a historical feature vector set as a feature vector set sample, and configuring target category information corresponding to the historical service request as a category label of the feature vector set sample; and training the to-be-trained class prediction model through the feature vector set sample and the class label of the feature vector set sample to obtain the class prediction model.

In an exemplary embodiment of the present disclosure, the class prediction model training module may be further configured to input the feature vector sample into a class prediction model to be trained, so as to obtain a sample class corresponding to the feature vector sample; and determining a loss function according to the sample class and the class label, and adjusting the parameters of the class prediction model to be trained until the loss function reaches the minimum value to obtain the class prediction model.

In an exemplary embodiment of the present disclosure, the class prediction model to be trained includes a department class prediction model to be trained and a business class prediction model to be trained. The category prediction model training module can also be used for inputting the feature vector set samples into the category prediction model of the department to be trained so as to obtain a plurality of sample department categories corresponding to the feature vector set samples and first probabilities corresponding to the sample department categories; determining a target sample department category from the plurality of sample department categories according to each first probability; and inputting the feature vector set sample and the department class of the target sample into the service class prediction model to be trained to obtain a plurality of sample service classes corresponding to the feature vector set sample and a second probability corresponding to each sample service class.

In an exemplary embodiment of the present disclosure, the category label includes a department category label and a business category label. The category prediction model training module can also be used for determining a first loss function according to the category of the department and the category label of the department of the sample, and obtaining a department category prediction model by adjusting the parameters of the category prediction model of the department to be trained until the first loss function reaches the minimum; and determining a second rectification loss function according to the sample service class and the service class label, and adjusting the parameters of the service class prediction model to be trained until the second rectification loss function reaches the minimum value to obtain the service class prediction model.

In an exemplary embodiment of the present disclosure, the information obtaining apparatus 1000 further includes a category prediction model updating module (not shown in the figure), which is configured to send target category information corresponding to the service request to a service window of the terminal, and receive a correction operation for the target category information, where the correction operation includes a real category information; determining prediction accuracy according to the target category information and the real category information, and judging whether the prediction accuracy is smaller than a first accuracy threshold value; if so, taking the multi-dimensional feature data set as a training sample, taking the real category information as a label corresponding to the training sample, and training the category prediction model through the training sample and the label corresponding to the training sample to obtain an updated category prediction model.

In an exemplary embodiment of the present disclosure, the information obtaining apparatus 1000 further includes an online model switching module (not shown in the figure) configured to configure a flow control switch, and allocate the plurality of multidimensional feature data sets to the category prediction model and the updated category prediction model through the flow control switch; and monitoring the prediction accuracy of the updated category prediction model in real time, and replacing the category prediction model with the updated category prediction model when the prediction accuracy of the updated category prediction model is greater than a second accuracy threshold.

The details of each information acquisition device are already described in detail in the corresponding information acquisition method, and therefore are not described herein again.

It should be noted that although in the above detailed description several modules or units of the apparatus for performing are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.

In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or program product. Thus, various aspects of the invention may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.

An electronic device 1100 according to this embodiment of the invention is described below with reference to fig. 11. The electronic device 1100 shown in fig. 11 is only an example and should not bring any limitations to the function and the scope of use of the embodiments of the present invention.

As shown in fig. 11, electronic device 1100 is embodied in the form of a general purpose computing device. The components of the electronic device 1100 may include, but are not limited to: the at least one processing unit 1110, the at least one memory unit 1120, a bus 1130 connecting different system components (including the memory unit 1120 and the processing unit 1110), and a display unit 1140.

Wherein the storage unit stores program code that is executable by the processing unit 1110 to cause the processing unit 1110 to perform steps according to various exemplary embodiments of the present invention as described in the above section "exemplary methods" of the present specification. For example, the processing unit 1110 may execute step S110 shown in fig. 1, and obtain a multidimensional feature data set corresponding to the service request, where the multidimensional feature data set includes a plurality of feature data of multiple dimensions; step S120, a feature vector set of the multi-dimensional feature data set is obtained, and the feature vector set of the multi-dimensional feature data set is input into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category; step S130, obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

The storage unit 1120 may include a readable medium in the form of a volatile memory unit, such as a random access memory unit (RAM)11201 and/or a cache memory unit 11202, and may further include a read only memory unit (ROM) 11203.

Storage unit 1120 may also include a program/utility 11204 having a set (at least one) of program modules 11205, such program modules 11205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

Bus 1130 may be representative of one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 1100 may also communicate with one or more external devices 1300 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a viewer to interact with the electronic device 1100, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 1100 to communicate with one or more other computing devices. Such communication may occur via an input/output (I/O) interface 1150. Also, the electronic device 1100 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the internet) via the network adapter 1160. As shown, the network adapter 1160 communicates with the other modules of the electronic device 1100 over the bus 1130. It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with the electronic device 1100, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

In an exemplary embodiment of the present disclosure, there is also provided a computer-readable storage medium having stored thereon a program product capable of implementing the above-described method of the present specification. In some possible embodiments, aspects of the invention may also be implemented in the form of a program product comprising program code means for causing a terminal device to carry out the steps according to various exemplary embodiments of the invention described in the above section "exemplary methods" of the present description, when said program product is run on the terminal device.

Referring to fig. 12, a program product 1200 for implementing the above method according to an embodiment of the present invention is described, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited in this regard and, in the present document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

A computer readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

Furthermore, the above-described figures are merely schematic illustrations of processes involved in methods according to exemplary embodiments of the invention, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.

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 to be limited only by the terms of the appended claims.

Claims

1. An information acquisition method, comprising:

acquiring a multi-dimensional feature data set corresponding to a service request, wherein the multi-dimensional feature data set comprises a plurality of feature data of a plurality of dimensions;

acquiring a feature vector set of a multi-dimensional feature data set, and inputting the feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category;

and obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

2. The information acquisition method according to claim 1, wherein the category prediction model includes a department category prediction model and a business category prediction model, and the prediction category includes a prediction department category and a prediction business category;

inputting a feature vector set of the multi-dimensional feature data set into a category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category, including:

inputting a feature vector set of the multi-dimensional feature data set into the department category prediction model to obtain a prediction department category corresponding to the multi-dimensional feature data set and a first prediction probability corresponding to the prediction department category;

and inputting the feature vector set of the multi-dimensional feature data set and the prediction department category into the service category prediction model to obtain a predicted service category corresponding to the multi-dimensional feature data set and a second prediction probability corresponding to the predicted service category.

3. The information acquisition method according to claim 2, wherein the prediction department category includes a plurality;

inputting the feature vector set of the multi-dimensional feature data set and the prediction department category into the business category prediction model, including:

sequencing the plurality of prediction department categories according to a first prediction probability corresponding to the prediction department categories, and acquiring the prediction department categories with the preset number as target prediction department categories;

and acquiring the data vector of the target prediction department category, fusing the data vector of the target prediction department category and the feature vector set of the multi-dimensional feature data set to obtain a fused feature vector set, and inputting the fused feature vector set into the service category prediction model.

4. The information acquisition method according to claim 3, wherein the predicted traffic class includes a plurality;

obtaining target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category, wherein the target category information comprises:

obtaining a target prediction service class according to a second prediction probability corresponding to each prediction service class;

judging whether the target prediction service class and a target prediction department class corresponding to the target prediction service class have an affiliated relationship according to a preset affiliated relationship;

and if so, configuring the target prediction business category and the target prediction department category as target category information.

5. The information acquisition method according to claim 1, wherein acquiring a feature vector set of a multi-dimensional feature data set includes:

and inputting the multi-dimensional feature data set into a vector extraction model to obtain a feature vector set corresponding to the multi-dimensional feature data set.

6. The information acquisition method according to claim 5, wherein the feature data in the multidimensional feature data set includes discrete feature data, continuous feature data, multivalued discrete feature data, text feature data;

the method further comprises the following steps:

if the feature data are the discrete feature data, the continuous feature data and the multi-valued discrete feature data, configuring the vector extraction model as a feature outline;

and if the feature data are the text feature data, configuring the vector extraction model into a word vector model.

7. The information acquisition method according to claim 6, wherein inputting the multidimensional feature data set into a vector extraction model to obtain a feature vector set corresponding to the multidimensional feature data set comprises:

judging whether the feature data in the multi-dimensional feature data set is the text feature data or not;

if so, inputting the feature data into the word vector model to obtain a first feature vector corresponding to the feature data;

if not, inputting the feature data into the feature outline to obtain a second feature vector corresponding to the feature data;

and carrying out vector splicing on the first characteristic vector and the second characteristic vector to obtain a characteristic vector set corresponding to the multi-dimensional characteristic data set.

8. The information acquisition method according to claim 6, wherein the word vector model includes three convolutional neural network layers, three maximum pooling layers, and one full-link layer, respectively.

9. The information acquisition method according to claim 6, characterized in that the method further comprises:

acquiring all text data in a resource conversion platform, and configuring the text data into text data samples;

and training a word vector model to be trained through the text data sample to obtain the word vector model.

10. The information acquisition method according to claim 1, characterized in that the method further comprises:

acquiring a historical multi-dimensional feature data set corresponding to a historical service request, and acquiring a historical feature vector set corresponding to the historical multi-dimensional feature data set according to a vector extraction model;

configuring the historical feature vector set as a feature vector set sample, and configuring target category information corresponding to the historical service request as a category label of the feature vector set sample;

and training the class prediction model to be trained through the feature vector set sample and the class label of the feature vector set sample to obtain the class prediction model.

11. The information acquisition method according to claim 10, wherein training the to-be-trained class prediction model by using the feature vector set samples and the class labels of the feature vector set samples to obtain the class prediction model comprises:

inputting the feature vector sample into the class prediction model to be trained to obtain a sample class corresponding to the feature vector sample;

and determining a loss function according to the sample class and the class label, and adjusting the parameters of the class prediction model to be trained until the loss function reaches the minimum value to obtain the class prediction model.

12. The information acquisition method according to claim 11, wherein the to-be-trained category prediction model includes a to-be-trained department category prediction model and a to-be-trained business category prediction model;

inputting the feature vector sample into the class prediction model to be trained to obtain a sample class corresponding to the feature vector sample, including:

inputting the feature vector set sample into the department category prediction model to be trained to obtain a plurality of sample department categories corresponding to the feature vector set sample and first probabilities corresponding to the sample department categories;

determining a target sample department category from the plurality of sample department categories according to each of the first probabilities;

and inputting the feature vector set sample and the target sample department category into the service category prediction model to be trained to obtain a plurality of sample service categories corresponding to the feature vector set sample and second probabilities corresponding to the sample service categories.

13. The information acquisition method according to claim 12, wherein the category label includes a department category label and a business category label;

determining a loss function according to the sample class and the class label, and obtaining the class prediction model by adjusting parameters of the class prediction model to be trained until the loss function reaches the minimum, wherein the method comprises the following steps:

determining a first loss function according to the sample department category and the department category label, and obtaining the department category prediction model by adjusting parameters of the to-be-trained department category prediction model until the first loss function reaches the minimum;

and determining a second rectification loss function according to the sample service class and the service class label, and obtaining the service class prediction model by adjusting the parameters of the service class prediction model to be trained until the second rectification loss function reaches the minimum.

14. The information acquisition method according to claim 1, characterized in that the method further comprises:

sending target category information corresponding to the service request to a service window of a terminal, and receiving a correction operation aiming at the target category information, wherein the correction operation comprises real category information;

determining a prediction accuracy according to the target category information and the real category information, and judging whether the prediction accuracy is smaller than a first accuracy threshold value;

if so, taking the multi-dimensional feature data set as a training sample, taking the real category information as a label corresponding to the training sample, and training the category prediction model through the training sample and the label corresponding to the training sample to obtain an updated category prediction model.

15. The information acquisition method according to claim 14, characterized by further comprising:

configuring a flow control switch through which a plurality of the multi-dimensional feature data sets are assigned to the class prediction model and the updated class prediction model;

and monitoring the prediction accuracy of the updated category prediction model in real time, and replacing the category prediction model with the updated category prediction model when the prediction accuracy of the updated category prediction model is greater than a second accuracy threshold.

16. The information acquisition method according to claim 1, wherein acquiring the multidimensional feature dataset corresponding to the service request comprises:

and calling real-time data interfaces with multiple dimensions, and acquiring multiple characteristic data corresponding to the service request through each real-time data interface.

17. An information acquisition apparatus characterized by comprising:

the data acquisition module is used for acquiring a multi-dimensional feature data set corresponding to the service request, wherein the multi-dimensional feature data set comprises a plurality of feature data of a plurality of dimensions;

the category prediction model is used for acquiring a feature vector set of a multi-dimensional feature data set and inputting the feature vector set of the multi-dimensional feature data set into the category prediction model to obtain a prediction category corresponding to the multi-dimensional feature data set and a prediction probability corresponding to the prediction category;

and the information acquisition module is used for acquiring target category information corresponding to the service request according to the prediction category and the prediction probability corresponding to the prediction category.

18. A computer-readable storage medium on which a computer program is stored, the program implementing the information acquisition method according to any one of claims 1 to 16 when executed by a processor.

19. An electronic device, comprising:

one or more processors;

storage means for storing one or more programs that, when executed by the one or more processors, cause the one or more processors to implement the information acquisition method according to any one of claims 1 to 16.