Search

CN-121996791-A - Event list classification method and device, program product and electronic equipment

CN121996791ACN 121996791 ACN121996791 ACN 121996791ACN-121996791-A

Abstract

The invention discloses a classification method of event sheets, a device, a program product and electronic equipment thereof, and relates to the field of financial science and technology or other related fields. The method and the device solve the technical problem of low efficiency of classifying the event list in the related technology.

Inventors

  • CHEN XIANG

Assignees

  • 中国工商银行股份有限公司

Dates

Publication Date
20260508
Application Date
20260130

Claims (10)

  1. 1. A method for classifying event sheets, comprising: acquiring an event list, wherein the event list comprises a question description fed back by a user when using a financial application, and the question description is recorded as an event description text; Inputting the event description text into a preset classification model and outputting a target class label, wherein the preset classification model at least comprises a text representation structure, a semantic extraction structure and a classification structure, wherein the text representation structure is used for extracting the characteristics of the event description text to obtain text vector representation, the semantic extraction structure is used for processing the text vector representation to obtain time sequence fusion characteristics, and the classification structure is used for processing the time sequence fusion characteristics to obtain class labels; and determining the target category to which the event ticket belongs based on the target category label.
  2. 2. The classification method according to claim 1, further comprising, before inputting the event description text to a preset classification model and outputting a target class label: constructing the text representation structure based on a language representation model; Constructing the semantic extraction structure based on a two-way long-short-term memory network; And constructing an initial classification model based on the text representation structure, the semantic extraction structure and the classification structure, wherein the classification structure at least comprises a linear layer and an output layer, the linear layer is used for carrying out feature dimension transformation on the time sequence fusion features, and the output layer is used for outputting class labels corresponding to the time sequence fusion features after feature dimension transformation.
  3. 3. The classification method according to claim 2, further comprising, after constructing the initial classification model: Collecting historical event sheets, and marking the historical event sheets based on event description texts contained in each historical event sheet to obtain marking labels corresponding to the historical event sheets; Training the initial classification model by adopting the historical event sheets and the labeling labels corresponding to each historical event sheet; Under the condition that each training is completed, obtaining a category label of each historical event list output by the initial classification model, and determining the fitting rate of the current training based on the category label and the labeling label; Under the condition that the fitting rate reaches the maximum, determining that training of the initial classification model is completed, and taking the trained model parameters indicated by the maximum fitting rate as target model parameters; and adjusting the initial classification model by adopting target model parameters to obtain the preset classification model.
  4. 4. The classification method according to claim 1, wherein the step of inputting the event description text into a preset classification model and outputting a target class label comprises: extracting features of the event description text by adopting the text representation structure to obtain the text vector representation; Processing the text vector representation by adopting the semantic extraction structure to obtain the time sequence fusion characteristic; And processing the time sequence fusion characteristics by adopting the classification structure to obtain the target class label.
  5. 5. The method of classifying as set forth in claim 4, wherein the step of extracting features of the event description text using the text representation structure to obtain the text vector representation includes: performing word segmentation processing on the event description text to obtain a plurality of segmented words; extracting features of each word segment by adopting the text representation structure to obtain a word vector corresponding to each word segment; the text vector representation is generated based on all of the word vectors.
  6. 6. The method of classifying as set forth in claim 4, wherein the semantic extraction structure includes at least a forward long-short-term memory unit and a backward long-short-term memory unit, and the step of processing the text vector representation using the semantic extraction structure to obtain the temporal fusion feature includes: inputting the text vector representation to the semantic extraction structure; Performing forward propagation processing on the text vector representation by adopting the forward long-short-term memory unit to obtain a forward time sequence characteristic of the last moment, wherein the input of the forward long-short-term memory unit at the initial moment is the text vector representation, and the output at the initial moment is the input of the next moment; the backward long-short-term memory unit is adopted to carry out backward propagation processing on the text vector representation, and backward time sequence characteristics of the last moment are obtained; and splicing the forward time sequence feature and the backward time sequence feature to obtain the time sequence fusion feature.
  7. 7. The method of classifying as set forth in claim 4, wherein the step of processing the time-series fusion feature using the classification structure to obtain the target class label includes: Performing feature dimension transformation by adopting the time sequence fusion features of the linear layer in the classification structure to obtain the time sequence fusion features after feature dimension transformation; and normalizing the time sequence fusion characteristics after feature dimension transformation by adopting a preset activation function to obtain the target class label.
  8. 8. A device for sorting event sheets, comprising: An acquisition unit, configured to acquire an event ticket, where the event ticket includes a question description that is fed back by a user when using a financial application, and the question description is recorded as an event description text; The input unit is used for inputting the event description text into a preset classification model and outputting a target class label, wherein the preset classification model at least comprises a text representation structure, a semantic extraction structure and a classification structure, the text representation structure is used for extracting the characteristics of the event description text to obtain text vector representation, the semantic extraction structure is used for processing the text vector representation to obtain time sequence fusion characteristics, and the classification structure is used for processing the time sequence fusion characteristics to obtain class labels; And the determining unit is used for determining the target category to which the event list belongs based on the target category label.
  9. 9. A computer program product comprising a non-volatile computer readable storage medium storing a computer program which, when executed by a processor, implements the method of classifying event tickets according to any of claims 1 to 7.
  10. 10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of classifying event tickets of any of claims 1 to 7.

Description

Event list classification method and device, program product and electronic equipment Technical Field The invention relates to the field of financial science and technology, in particular to a method and a device for classifying event sheets, a program product and electronic equipment. Background The event ticket refers to an event ticket which is manually generated by a website business person according to the description of a client after a banking client feeds back the problems encountered in the process of using a banking application to a banking website. The event list is usually circulated among various institutions of the bank in the form of mails, and mail processors forward the mails by taking event descriptions in the event list as references, and finally send the mails to a designated functional responsible group. During the circulation process, the event list needs to be classified according to the application function to which the event belongs. Currently, the event list is classified by adopting a manual mode, and after the current processor of the mail reads the event description, the mail is sent to the next processor according to subjective judgment, and the like, and finally the mail is distributed to the developer. With the rapid development of financial science and technology, banks are accelerating science and technology and digitalization. Thus, more and more banking applications will be launched into the market, and the iteration of the various applications will be faster and faster. As the number of bank clients increases, the number of event sheets related to various applications will also increase, and if the event sheets are classified by means of manual mode all the time, human resources will be greatly consumed. Meanwhile, due to the limitation of manpower and the limited energy, the event list classification is inaccurate and slow, so that the event list processing efficiency is reduced. For the above problems, no effective solution has been proposed at present. Disclosure of Invention The embodiment of the invention provides a method and a device for classifying event sheets, a program product and electronic equipment, which at least solve the technical problem of low efficiency of classifying event sheets in related technologies. According to one aspect of the embodiment of the invention, a classification method of an event list is provided, which comprises the steps of obtaining the event list, wherein the event list comprises problem descriptions fed back by a user when the user uses financial application, the problem descriptions are recorded as event description texts, inputting the event description texts into a preset classification model and outputting target class labels, the preset classification model at least comprises a text representation structure, a semantic extraction structure and a classification structure, the text representation structure is used for extracting characteristics of the event description texts to obtain text vector representations, the semantic extraction structure is used for processing the text vector representations to obtain time sequence fusion characteristics, the classification structure is used for processing the time sequence fusion characteristics to obtain class labels, and determining the target class to which the event list belongs based on the target class labels. The method comprises the steps of inputting event description texts into a preset classification model and outputting target class labels, constructing a text representation structure based on a language representation model, constructing a semantic extraction structure based on a two-way long-short-term memory network, and constructing an initial classification model based on the text representation structure, the semantic extraction structure and a classification structure, wherein the classification structure at least comprises a linear layer and an output layer, the linear layer is used for carrying out feature dimension transformation on time sequence fusion features, and the output layer is used for outputting class labels corresponding to the time sequence fusion features after feature dimension transformation. The method comprises the steps of establishing an initial classification model, acquiring historical event sheets, marking the historical event sheets based on event description texts contained in each historical event sheet to obtain marking labels corresponding to the historical event sheets, training the initial classification model by adopting the historical event sheets and the marking labels corresponding to each historical event sheet, acquiring a class label of each historical event sheet output by the initial classification model under the condition that each training is completed, determining the fitting rate of current training based on the class labels and the marking labels, determining that training the initial classification model is comple