CN116956939A - Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan - Google Patents

Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan Download PDF

Info

Publication number
CN116956939A
CN116956939A CN202310955056.1A CN202310955056A CN116956939A CN 116956939 A CN116956939 A CN 116956939A CN 202310955056 A CN202310955056 A CN 202310955056A CN 116956939 A CN116956939 A CN 116956939A
Authority
CN
China
Prior art keywords
text
dialogue
label
post
debtor
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310955056.1A
Other languages
Chinese (zh)
Inventor
罗歆昱
苏江
崔健
章彦博
陈洧
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
DMAI Guangzhou Co Ltd
Original Assignee
DMAI Guangzhou Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by DMAI Guangzhou Co Ltd filed Critical DMAI Guangzhou Co Ltd
Priority to CN202310955056.1A priority Critical patent/CN116956939A/en
Publication of CN116956939A publication Critical patent/CN116956939A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/03—Credit; Loans; Processing thereof
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Finance (AREA)
  • Databases & Information Systems (AREA)
  • Accounting & Taxation (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Human Computer Interaction (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Machine Translation (AREA)

Abstract

The invention relates to the technical field of text semantic understanding, and discloses a method, a device and equipment for semantic understanding of a text of a post-credit scene, wherein the method comprises the following steps: acquiring a harvest-oriented dialogue text of a post-loan scene and processing the harvest-oriented dialogue text based on a time sequence to obtain a dialogue list; based on preset text labels facing the post-credit scene, semantic analysis is carried out on the words of the receiver and the debtor in the dialogue list respectively; constructing a tag matrix of the dialogue list based on the single sentence semantic analysis result; training the long-term and short-term memory network model based on the label matrix to obtain a time sequence multi-label classification model; and inputting a label matrix corresponding to the to-be-analyzed furcation dialogue text into a model to obtain a multi-label classification result serving as a text semantic understanding result. The invention reserves the whole semantic information of the dialogue text, and carries out semantic understanding on each role single sentence, so that the model can understand semantic information with finer granularity, thereby obtaining higher accuracy.

Description

Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan
Technical Field
The invention relates to the field of text semantic understanding, in particular to a method, a device and equipment for semantic understanding of a text of a post-loan scene oriented to the text of a user.
Background
Post-credit management is the final link of credit management, and post-credit management can manage, identify and collect for credit users with overdue symptoms or behaviors, and the degree of refinement of post-credit management will affect the quality of business management in downstream credit scenarios. The traditional post-loan wind control model is mainly manually audited, the data feature scale is small, the data association degree is low by means of expert experience, the intelligent post-loan wind control is mainly based on autonomous analysis and decision of a model strategy system, and the data feature scale is large. The semantic understanding of the post-credit text is essentially to classify the text of the dialogue text generated in the post-credit text, and aims to provide semantic features for the post-credit wind control model and finally improve the post-credit wind control model effect.
In the prior art, the field of the collected text is strong, and no specific field pre-training model self-adaptive learning is performed for the collected text; after the credit, the length of the collected text is longer, part of important information can be lost by common truncation, and corresponding processing is not carried out on the long text, so that the semantic understanding precision of the collected text is lower.
Disclosure of Invention
In view of the above, the invention provides a post-credit scene oriented method, a post-credit scene oriented device, a post-credit scene oriented computer device and a post-credit scene oriented storage medium for solving the problem of lower accuracy of results of post-credit scene semantic understanding in the prior art.
In a first aspect, the present invention provides a post-credit scene oriented text semantic understanding method, including:
acquiring a collect-urging dialogue text oriented to a post-credit scene, and processing the collect-urging dialogue text based on a time sequence to obtain a dialogue list;
semantic analysis is carried out on the words of the receiver and the words of the debtor in the dialogue list based on preset text labels facing the post-credit scene respectively, so that the words semantic analysis results of the receiver and the debtor are obtained;
constructing a tag matrix of the dialogue list based on the single sentence semantic analysis results of the receiver and the debtor, wherein the tag matrix is used for representing the tag characteristics of each sentence of the dialogue list;
training a long-term and short-term memory network model based on a label matrix, and taking the trained model as a time sequence multi-label classification model for generating a multi-classification label of the dialogue text for accelerating the understanding of the whole section of semantics;
and inputting a label matrix corresponding to the to-be-analyzed furcation dialogue text into a time sequence multi-label classification model to obtain a multi-label classification result which is used as a furcation dialogue text semantic understanding result.
The semantic understanding method of the harvest-oriented text for the scene after the credit maintains the whole semantic information of the dialogue text, and semantic understanding is carried out on each role single sentence, so that a model can understand semantic information with finer granularity, and higher accuracy is obtained.
In an alternative embodiment, the process of processing the collect dialog text based on time sequence to obtain the dialog list includes:
identifying each sentence in the collect dialog text by an identifier to distinguish it as a collect person text or a debtor text; dividing sentences in the prompting dialogue text by line-feed symbols according to the time sequence of the dialogue; the receipts dialog text is filled in order to a dialog list l= [ C, Z ] by splitting the line connector and the identifier, where C represents the receipts' text and Z represents the debtor text.
The embodiment of the invention can identify the text of the collect person and the text of the debtor after processing the collect dialogue text, and is used for the subsequent more accurate semantic identification.
In an optional implementation manner, the semantic analysis is performed on the collect person single sentence and the debtor single sentence in the dialogue list based on the preset text labels facing the post-credit scene, including:
labeling data is carried out on the text of the receiver based on preset receiver labels facing the post-credit scene, fine tuning training is carried out on the BERT model based on the labeling data of the text of the receiver to obtain a first text multi-label classification model, semantic analysis is carried out on the receiver single sentence in a dialogue list based on the first text multi-label classification model, and a receiver single sentence text multi-label classification result is obtained;
labeling data is carried out on the debtor text based on preset debtor labels facing the post-credit scene, fine tuning training is carried out on the BERT model based on the labeling data of the debtor text, a first text multi-label classification model is obtained, semantic analysis is carried out on the debtor single sentence based on a second text multi-label classification model, and a debtor single sentence text multi-label classification result is obtained.
The embodiment of the invention is oriented to a post-loan scene, different text labels are respectively set for the repayment person and the debtor, and semantic analysis is respectively carried out on the two roles, so that the repayment possibility of the debtor and the repayment intention of the repayment person can be obtained, and the repayment policy of the repayment person can be adjusted by the repayment person to improve the repayment possibility of the debtor.
In an alternative embodiment, the constructing a tag matrix of the dialogue list based on the results of the semantic analysis of the single sentences of the incentives and the debtors includes:
defining a vector with dimension of 1×n, wherein each position of N represents a label of the first text multi-label classification model or the second text multi-label classification model, if a sentence in the text belongs to a certain label, the corresponding position is 1, otherwise, the sentence is 0, and a label vector consisting of 0 and 1 is obtained and used for representing the label characteristics of the text;
a matrix is defined, the dimension being the number of sentences mxn, wherein the mth row represents the tag characteristics of the mth sentence in the dialogue list L.
In an alternative embodiment, the determining the classification label of the enrollee based on the dialogue intention of the enrollee in the post-credit oriented scenario includes: identity confirmation, pressure application, negotiation of repayment, inquiry of reasons for arrears, inquiry of willingness to repayment.
In an alternative embodiment, the determining the classification label of the debtor based on the dialogue intention of the debtor in the post-credit oriented scenario includes: promise repayment, refusal repayment and information consultation.
In an alternative embodiment, the induction dialogue text multi-category label for understanding whole-segment semantics includes: whether the person, repayment willing, customer attitude, business action and overdue reasons.
The embodiment of the invention sets the labels according to the service scene, inputs the labels as the characteristics into the subsequent post-finance credit wind control model, and sorts the repayment willingness of the debtor according to the characteristics by the wind control model so as to adjust the subsequent collection strategy.
In a second aspect, the present invention provides a post-credit scene oriented text semantic understanding device, the device comprising:
the dialogue text processing module is used for acquiring a collect-urging dialogue text facing the post-credit scene and processing the collect-urging dialogue text based on time sequence to obtain a dialogue list;
the sentence semantic analysis module is used for carrying out semantic analysis on the words of the receiver and the words of the debtor in the dialogue list based on the preset text labels facing the post-credit scene respectively, so as to obtain the sentence semantic analysis results of the receiver and the debtor;
the label matrix construction module is used for constructing a label matrix of the dialogue list based on the single sentence semantic analysis results of the collector and the debtor and used for representing the label characteristics of each sentence of the dialogue list;
the time sequence multi-label classification model building module is used for training the long-short-term memory network model based on the label matrix, taking the trained model as a time sequence multi-label classification model and generating a multi-classification label of the induced receiving dialogue text for understanding the whole section of semantics;
and the collect-forcing dialogue text voice understanding module is used for inputting a label matrix corresponding to the collect-forcing dialogue text to be analyzed into the time sequence multi-label classification model to obtain a multi-label classification result which is used as a collect-forcing dialogue text semantic understanding result.
In a third aspect, the present invention provides a computer device comprising: the memory and the processor are in communication connection, the memory stores computer instructions, and the processor executes the computer instructions to execute the method for semantic understanding of the text oriented to the post-credit scene according to the first aspect or any one of the corresponding embodiments.
In a fourth aspect, the present invention provides a computer readable storage medium having stored thereon computer instructions for causing a computer to execute the method for semantic understanding of the text semantic for post-credit oriented scenarios of the first aspect or any of its corresponding embodiments.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an example of a post-credit scene oriented method for semantic understanding of a text of interest according to an embodiment of the present invention;
FIG. 2 is a block diagram of a post-credit scene oriented text semantic understanding device according to an embodiment of the present invention;
fig. 3 is a schematic hardware structure of a computer device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In accordance with an embodiment of the present invention, there is provided an embodiment of a post-credit scene oriented method for semantic understanding of the induced text, it being noted that the steps illustrated in the flowchart of the figures may be performed in a computer system, such as a set of computer executable instructions, and that although a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
In this embodiment, a post-credit scene oriented text semantic understanding method is provided, which may be used in a computer device terminal, and fig. 1 is a flowchart of a post-credit scene oriented text semantic understanding method according to an embodiment of the present invention, and as shown in fig. 1, the flowchart includes the following steps:
step S101, acquiring a collect-urging dialogue text oriented to a post-credit scene, and processing the collect-urging dialogue text based on time sequence to obtain a dialogue list.
In the embodiment of the invention, each sentence in the collection-forcing dialogue text is identified through an identifier to distinguish whether the sentence is a collection-forcing text or a debtor text; dividing sentences in the prompting dialogue text by a line feed symbol "\n" according to the time sequence of the dialogue; the receipts dialog text is filled in order to a dialog list l= [ C, Z ] by splitting the line connector and the identifier, where C represents the receipts' text and Z represents the debtor text. After the collection dialog text is processed, the collection person text and the debtor text can be identified for subsequent more accurate semantic identification.
Step S102, semantic analysis is carried out on the words of the receiver and the words of the debtor in the dialogue list based on preset text labels facing the post-loan scene, so that the word semantic analysis results of the receiver and the debtor are obtained.
The embodiment of the invention is oriented to a post-loan scene, different text labels are respectively set for the repayment person and the debtor, and semantic analysis is respectively carried out on the two roles, so that the repayment possibility of the debtor and the repayment intention of the repayment person can be obtained, and the repayment policy of the repayment person can be adjusted by the repayment person to improve the repayment possibility of the debtor.
According to the embodiment of the invention, based on the preset post-loan scene oriented adductor labels, labeling data is carried out on adductor texts, fine tuning training is carried out on BERT models based on the labeling data of the adductor texts, a first text multi-label classification model is obtained, and semantic analysis is carried out on adductor single sentences in a dialogue list based on the first text multi-label classification model, so that an adductor single sentence text multi-label classification result is obtained. Wherein, based on the conversation intention of the enrollee facing the scene after the credit, the classification label of the enrollee is determined, comprising: identity confirmation, pressure application, negotiation of repayment, inquiry of reasons for arrears, inquiry of willingness to repayment, specific label examples are given in the table below. Each segment of the cashier's text will be classified into one or more of the above categories, and these tag features will facilitate the intended classification of the entire section of the cashier's text from a business perspective, as the pressing process is more, the likelihood of repayment for the likely liability is higher, the identity is confirmed as not itself but a claim is made, and the likelihood of repayment for the liability is also increased.
According to the embodiment of the invention, based on the preset debtor label facing the post-credit scene, labeling data is carried out on the debtor text, fine tuning training is carried out on the BERT model based on the labeling data of the debtor text, a first text multi-label classification model is obtained, semantic analysis is carried out on the debtor single sentence based on a second text multi-label classification model, and a debtor single sentence text multi-label classification result is obtained. Wherein determining a classification tag of the liability person based on the dialogue intent of the liability person in the post-lending oriented scenario comprises: promise repayment, refusal repayment and information consultation, and specific label examples are shown in the table below. Each segment of debtor text will be classified into one or more of the above categories, and these features will be advantageous from a business perspective to the intended classification of the entire segment of the cohesive text, such as the promised repayment of the debtor, the likelihood of repayment by the debtor being high, the likelihood of repayment by the debtor being refused, and the likelihood of repayment by the debtor being low.
Step S103, constructing a label matrix of the dialogue list based on the single sentence semantic analysis results of the collector and the debtor, wherein the label matrix is used for representing the label characteristics of each sentence of the dialogue list.
In one embodiment, a vector with dimension of 1×n is defined (for example, there are 30 debt human labels, 30 collect human labels, and N is taken as 60), where each position of N represents a label of the first text multi-label classification model or the second text multi-label classification model, if a sentence in the text belongs to a certain label, the corresponding position is 1, otherwise, it is 0, and a label vector consisting of 0 and 1 is obtained, so as to represent label characteristics of the text; a matrix is defined, the dimension being the number of sentences mxn, wherein the mth row represents the tag characteristics of the mth sentence in the dialogue list L.
Step S104, training the long-short-term memory network model based on the label matrix, and taking the trained model as a time sequence multi-label classification model for generating a multi-classification label of the induced-harvest dialogue text for understanding the whole section of semantics.
In the embodiment of the invention, the multi-classification label of the whole-section semantic induced-harvest dialogue text comprises the following components: whether self, willingness to repayment (high, medium, low), customer attitude (good, medium, bad), business actions (open white, transfer, repayment promise confirmation), overdue reasons (excessive loan), etc., specific label examples are shown in the following table (each serial number index represents a label, if index is independent, it is classified into a class (i.e. label takes a value of 0, 1); if index is repeated, it indicates that it is a multi-classification (i.e. label value is multiple, test_ auc is an evaluation standard of the model for classifying labels under test set, the higher the value is 0-1, the better the description effect is), the labels are set according to business scenario, these labels will be input as features to the following post-finance credit wind control model, the wind control model will sort the repayment willingness of debtors according to these features so as to adjust the following collection policy, such as adopting a milder collection method for customers with high repayment willingness, adopting a more serious collection method for customers with low repayment willingness, etc.
Step S105, inputting a label matrix corresponding to the to-be-analyzed furnacing dialogue text into a time sequence multi-label classification model to obtain a multi-label classification result which is used as a furnacing dialogue text semantic understanding result.
The semantic understanding method of the harvest-oriented text for the scene after the credit, provided by the embodiment of the invention, reserves the whole semantic information of the dialogue text, carries out semantic understanding on each role single sentence, and enables the model to understand semantic information with finer granularity, thereby obtaining higher accuracy.
The embodiment also provides a post-credit scene oriented device for semantic understanding of the text of the induced shrinkage, which is used for realizing the embodiment and the preferred implementation, and is not described again. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
The embodiment provides a post-credit scene oriented text semantic understanding device, as shown in fig. 2, including:
the dialogue text processing module 201 is configured to obtain a collect-urging dialogue text for a post-credit scene, and process the collect-urging dialogue text based on a time sequence to obtain a dialogue list;
the sentence semantic analysis module 202 is configured to perform semantic analysis on the words of the receiver and the words of the debtor in the dialogue list based on the preset text labels facing the post-credit scene, so as to obtain sentence semantic analysis results of the receiver and the debtor;
the tag matrix construction module 203 is configured to construct a tag matrix of the dialogue list based on the results of semantic analysis of the single sentences of the receiver and the debtor, and is configured to characterize tag features of each sentence of the dialogue list;
the time sequence multi-label classification model construction module 204 is used for training the long-short-term memory network model based on the label matrix, taking the trained model as a time sequence multi-label classification model and generating a multi-classification label of the induced harvest dialogue text for understanding the whole section of semantics;
the collect-promoting dialogue text speech understanding module 205 is configured to input a tag matrix corresponding to the collect-promoting dialogue text to be analyzed into the time sequence multi-tag classification model to obtain a multi-tag classification result, which is used as a collect-promoting dialogue text semantic understanding result.
The post-credit scene oriented text semantic understanding device in this embodiment is presented in the form of functional units, where the units refer to ASIC circuits, processors and memories that execute one or more software or firmware programs, and/or other devices that may provide the above-described functionality.
Further functional descriptions of the above respective modules and units are the same as those of the above corresponding embodiments, and are not repeated here.
The embodiment of the invention also provides a computer device, which is provided with the device for understanding the semantic meaning of the harvest-oriented text in the scene after the credit shown in the figure 2.
Referring to fig. 3, fig. 3 is a schematic structural diagram of a computer device according to an alternative embodiment of the present invention, and as shown in fig. 3, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components are communicatively coupled to each other using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions executing within the computer device, including instructions stored in or on memory to display graphical information of the GUI on an external input/output device, such as a display device coupled to the interface. In some alternative embodiments, multiple processors and/or multiple buses may be used, if desired, along with multiple memories and multiple memories. Also, multiple computer devices may be connected, each providing a portion of the necessary operations (e.g., as a server array, a set of blade servers, or a multiprocessor system). One processor 10 is illustrated in fig. 3.
The processor 10 may be a central processor, a network processor, or a combination thereof. The processor 10 may further include a hardware chip, among others. The hardware chip may be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general-purpose array logic, or any combination thereof.
Wherein the memory 20 stores instructions executable by the at least one processor 10 to cause the at least one processor 10 to perform the methods shown in implementing the above embodiments.
The memory 20 may include a storage program area that may store an operating system, at least one application program required for functions, and a storage data area; the storage data area may store data created from the use of the computer device of the presentation of a sort of applet landing page, and the like. In addition, the memory 20 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, memory 20 may optionally include memory located remotely from processor 10, which may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
Memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk, or solid state disk; the memory 20 may also comprise a combination of the above types of memories.
The computer device also includes a communication interface 30 for the computer device to communicate with other devices or communication networks.
The embodiments of the present invention also provide a computer readable storage medium, and the method according to the embodiments of the present invention described above may be implemented in hardware, firmware, or as a computer code which may be recorded on a storage medium, or as original stored in a remote storage medium or a non-transitory machine readable storage medium downloaded through a network and to be stored in a local storage medium, so that the method described herein may be stored on such software process on a storage medium using a general purpose computer, a special purpose processor, or programmable or special purpose hardware. The storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, a solid state disk or the like; further, the storage medium may also comprise a combination of memories of the kind described above. It will be appreciated that a computer, processor, microprocessor controller or programmable hardware includes a storage element that can store or receive software or computer code that, when accessed and executed by the computer, processor or hardware, implements the methods illustrated by the above embodiments.
Although embodiments of the present invention have been described in connection with the accompanying drawings, various modifications and variations may be made by those skilled in the art without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope of the invention as defined by the appended claims.

Claims (10)

1. A post-credit scene oriented method for semantic understanding of a text to be induced, the method comprising:
acquiring a collect-urging dialogue text oriented to a post-credit scene, and processing the collect-urging dialogue text based on a time sequence to obtain a dialogue list;
semantic analysis is carried out on the words of the receiver and the words of the debtor in the dialogue list based on preset text labels facing the post-credit scene respectively, so that the words semantic analysis results of the receiver and the debtor are obtained;
constructing a tag matrix of the dialogue list based on the single sentence semantic analysis results of the receiver and the debtor, wherein the tag matrix is used for representing the tag characteristics of each sentence of the dialogue list;
training a long-term and short-term memory network model based on a label matrix, and taking the trained model as a time sequence multi-label classification model for generating a multi-classification label of the dialogue text for accelerating the understanding of the whole section of semantics;
and inputting a label matrix corresponding to the to-be-analyzed furcation dialogue text into a time sequence multi-label classification model to obtain a multi-label classification result which is used as a furcation dialogue text semantic understanding result.
2. The method of claim 1, wherein the process of processing the collect dialog text based on the time sequence to obtain the dialog list comprises:
identifying each sentence in the collect dialog text by an identifier to distinguish it as a collect person text or a debtor text; dividing sentences in the prompting dialogue text by line-feed symbols according to the time sequence of the dialogue; the receipts dialog text is filled in order to a dialog list l= [ C, Z ] by splitting the line connector and the identifier, where C represents the receipts' text and Z represents the debtor text.
3. The method according to claim 2, wherein the semantic analysis of the collect person sentence and the debtor sentence in the dialogue list based on the preset text labels for the post-credit scenario, respectively, comprises:
labeling data is carried out on the text of the receiver based on preset receiver labels facing the post-credit scene, fine tuning training is carried out on the BERT model based on the labeling data of the text of the receiver to obtain a first text multi-label classification model, semantic analysis is carried out on the receiver single sentence in a dialogue list based on the first text multi-label classification model, and a receiver single sentence text multi-label classification result is obtained;
labeling data is carried out on the debtor text based on preset debtor labels facing the post-credit scene, fine tuning training is carried out on the BERT model based on the labeling data of the debtor text, a first text multi-label classification model is obtained, semantic analysis is carried out on the debtor single sentence based on a second text multi-label classification model, and a debtor single sentence text multi-label classification result is obtained.
4. A method according to claim 3, wherein said constructing a tag matrix of a dialogue list based on results of semantic analysis of sentences of the seeker and the debtors comprises:
defining a vector with dimension of 1×n, wherein each position of N represents a label of the first text multi-label classification model or the second text multi-label classification model, if a sentence in the text belongs to a certain label, the corresponding position is 1, otherwise, the sentence is 0, and a label vector consisting of 0 and 1 is obtained and used for representing the label characteristics of the text;
a matrix is defined, the dimension being the number of sentences mxn, wherein the mth row represents the tag characteristics of the mth sentence in the dialogue list L.
5. The method of claim 3, wherein the determining the classification tags for the enrollee based on the interaction intent of the enrollee in the post-credit oriented scenario comprises: identity confirmation, pressure application, negotiation of repayment, inquiry of reasons for arrears, inquiry of willingness to repayment.
6. The method of claim 3, wherein the determining the classification labels of the debtors based on the conversational intent of the debtors in the post-credit oriented scenario comprises: promise repayment, refusal repayment and information consultation.
7. The method of claim 6, wherein the induced dialog text multi-category label that understands whole-segment semantics comprises: whether the person, repayment willing, customer attitude, business action and overdue reasons.
8. An after-credit scene oriented text semantic understanding device, the device comprising:
the dialogue text processing module is used for acquiring a collect-urging dialogue text facing the post-credit scene and processing the collect-urging dialogue text based on time sequence to obtain a dialogue list;
the sentence semantic analysis module is used for carrying out semantic analysis on the words of the receiver and the words of the debtor in the dialogue list based on the preset text labels facing the post-credit scene respectively, so as to obtain the sentence semantic analysis results of the receiver and the debtor;
the label matrix construction module is used for constructing a label matrix of the dialogue list based on the single sentence semantic analysis results of the collector and the debtor and used for representing the label characteristics of each sentence of the dialogue list;
the time sequence multi-label classification model building module is used for training the long-short-term memory network model based on the label matrix, taking the trained model as a time sequence multi-label classification model and generating a multi-classification label of the induced receiving dialogue text for understanding the whole section of semantics;
and the collect-forcing dialogue text voice understanding module is used for inputting a label matrix corresponding to the collect-forcing dialogue text to be analyzed into the time sequence multi-label classification model to obtain a multi-label classification result which is used as a collect-forcing dialogue text semantic understanding result.
9. A computer device, comprising:
a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the post-credit scene oriented method of text semantic understanding of any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon computer instructions for causing a computer to perform the post-credit scene oriented method of text semantic understanding of the post-credit scene of any one of claims 1 to 7.
CN202310955056.1A 2023-07-31 2023-07-31 Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan Pending CN116956939A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310955056.1A CN116956939A (en) 2023-07-31 2023-07-31 Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310955056.1A CN116956939A (en) 2023-07-31 2023-07-31 Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan

Publications (1)

Publication Number Publication Date
CN116956939A true CN116956939A (en) 2023-10-27

Family

ID=88458130

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310955056.1A Pending CN116956939A (en) 2023-07-31 2023-07-31 Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan

Country Status (1)

Country Link
CN (1) CN116956939A (en)

Similar Documents

Publication Publication Date Title
CN108733722B (en) Automatic generation method and device for conversation robot
CN109872162B (en) Wind control classification and identification method and system for processing user complaint information
CN113326764A (en) Method and device for training image recognition model and image recognition
EP3441891A1 (en) Data source-based service customisation apparatus, method, system, and storage medium
CN109948160B (en) Short text classification method and device
CN111125354A (en) Text classification method and device
CN110890088A (en) Voice information feedback method and device, computer equipment and storage medium
CN109992781B (en) Text feature processing method and device and storage medium
CN112235470B (en) Incoming call client follow-up method, device and equipment based on voice recognition
CN111651600B (en) Sentence multi-intention recognition method, system, electronic equipment and storage medium
CN111831826A (en) Training method, classification method and device of cross-domain text classification model
CN111428448A (en) Text generation method and device, computer equipment and readable storage medium
CN110955770A (en) Intelligent dialogue system
CN112163081A (en) Label determination method, device, medium and electronic equipment
CN112487186A (en) Human-human conversation log analysis method, system, equipment and storage medium
CN112836053A (en) Man-machine conversation emotion analysis method and system for industrial field
CN116071150A (en) Data processing method, bank product popularization, wind control system, server and medium
CN116628163A (en) Customer service processing method, customer service processing device, customer service processing equipment and storage medium
CN113887214B (en) Willingness presumption method based on artificial intelligence and related equipment thereof
WO2020042164A1 (en) Artificial intelligence systems and methods based on hierarchical clustering
CN111666408A (en) Method and device for screening and displaying important clauses
CN111309882A (en) Method and device for realizing intelligent customer service question answering
CN116186257A (en) Method and system for classifying short texts based on mixed features
CN116956939A (en) Method, device and equipment for semantic understanding of harvest-oriented text of scene after loan
CN114741501A (en) Public opinion early warning method and device, readable storage medium and electronic equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination