CN112085594A

CN112085594A - Identity verification method, equipment and readable storage medium

Info

Publication number: CN112085594A
Application number: CN202010964817.6A
Authority: CN
Inventors: 周楠楠; 于夕畔; 汤耀华; 杨海军; 徐倩
Original assignee: WeBank Co Ltd
Current assignee: WeBank Co Ltd
Priority date: 2020-09-14
Filing date: 2020-09-14
Publication date: 2020-12-15
Anticipated expiration: 2040-09-14
Also published as: CN112085594B

Abstract

The application discloses an identity verification method, equipment and a readable storage medium, wherein the method comprises the following steps: when an identity verification instruction for verifying the identity of the applicant is detected, determining the conversation state information of the current turn according to the conversation content of the current turn; and dynamically determining a next turn of dialog execution text according to the current turn of dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text. According to the method and the device, the identity verification result of the applicant is obtained only through a few conversation processes, so that the identity verification process is saved, and the technical problem that the identity verification efficiency of the applicant is low in the prior art is solved.

Description

Identity verification method, equipment and readable storage medium

Technical Field

The present application relates to the field of artificial intelligence technology for financial technology (Fintech), and in particular, to an identity verification method, device and readable storage medium.

Background

With the continuous development of financial technologies, especially internet technology and finance, more and more technologies (such as distributed, Blockchain, artificial intelligence, etc.) are applied to the financial field, but the financial industry also puts higher requirements on the technologies, such as the financial industry also has higher requirements on identity verification.

In the process of applying for loan by an applicant, a financial institution such as a bank generally needs to verify personal information of the applicant by identity to avoid security problems of the property, and at present, an intelligent self-verification robot generally queries the applicant once by using a fixed technology according to preset problems in order to verify the identity information of the applicant, for example, even if two or three rounds of conversations are needed to judge the authenticity of the information of the applicant, if the preset problems are 10, at least 10 rounds of conversations, namely 10 rounds of full flows, are needed to complete the verification of the identity information of the applicant, that is, the existing identity verification information has a problem of low efficiency.

Disclosure of Invention

The present application mainly aims to provide an identity verification method, an identity verification apparatus and a readable storage medium, and aims to solve the technical problem of low applicant identity verification efficiency in the prior art.

In order to achieve the above object, the present application provides an identity verification method applied to an intelligent robot, where the identity verification method includes:

when an identity verification instruction for verifying the identity of the applicant is detected, determining the conversation state information of the current turn according to the conversation content of the current turn;

and dynamically determining a next turn of dialog execution text according to the current turn of dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text.

Optionally, the step of determining the dialog state information of the current turn according to the dialog content of the current turn when the identity verification instruction for verifying the identity of the applicant is detected includes:

when an identity verification instruction for verifying the identity of the applicant is detected, outputting the inquiry question of the current turn of conversation;

receiving the response content input by the applicant aiming at the question to obtain the slot value pair information;

determining the dialogue state information of the intelligent core robot and the applicant in the previous round and the behavior information of the intelligent core robot in the previous round;

and determining the conversation state information of the applicant in the current round identity verification process according to the slot value pair information, the conversation state information of the previous round and the behavior information of the previous round, so as to determine the conversation state information of the current round according to the conversation content of the current round.

Optionally, the step of determining dialog state information with the applicant in the current round of identity verification process according to the slot value pair information, the previous round of dialog state information, and the last round of behavior information includes:

inputting the slot value pair information, the previous turn of conversation state information and the previous turn of behavior information into a preset conversation state tracking model as input information;

processing the input information based on the preset conversation state tracking model to obtain conversation state information of the applicant in the current round identity verification process;

the preset dialogue state tracking model is a first target model obtained after iterative training is carried out on a first preset basic model based on dialogue data with preset dialogue labels.

Optionally, the step of receiving the response content input by the applicant to the question and obtaining the slot value pair information includes:

receiving response content input by the applicant for the question;

performing dependency syntax analysis on the reply content based on a preset dependency syntax model to obtain a dependency syntax analysis result corresponding to the reply content;

and based on the dependency syntax analysis result, carrying out word slot filling on a preset word slot to be filled corresponding to the reply content to obtain slot value pair information.

Optionally, the preset dependency syntax model includes a dependency discrimination model and a dependency type prediction model, and the step of performing dependency syntax analysis on the reply content based on the preset dependency syntax model to obtain a dependency syntax analysis result corresponding to the reply content includes:

vectorizing the reply content to obtain a vectorized statement;

based on the dependency relationship judging model, judging the dependency relationship of the vectorized statement to obtain a dependency relationship judging result;

and performing dependency relationship type prediction on the vectorized statement based on the dependency relationship type prediction model and the dependency relationship judgment result to obtain the dependency syntax analysis result.

Optionally, the step of performing word slot filling on a preset word slot to be filled corresponding to the reply content based on the dependency parsing result to obtain slot value pair information includes:

extracting words to be filled in the reply content based on the dependency syntax analysis result;

judging whether the sentence pattern of the reply content meets the requirement of a preset sentence pattern, and if the sentence pattern of the reply content meets the requirement of the preset sentence pattern, judging whether the position of a slot value bit appearing in the reply content is a preset position;

if the position of the slot value bit appearing in the reply content is not a preset position, inputting the reply content into a preset sequence labeling model to predict the reply content to obtain a content prediction result;

and filling the current preset word slot according to the content prediction result to obtain slot value pair information.

Optionally, after the step of determining whether the sentence pattern of the reply content meets the preset sentence pattern requirement, and if the sentence pattern of the reply content meets the preset sentence pattern requirement, determining whether the position of the slot value bit appearing in the reply content is a preset position, the method includes:

if the position of the slot value position appearing in the reply content is a preset position, extracting a target word in the preset position;

judging whether the part of speech of the target word meets the requirement of a preset part of speech;

and if the part of speech of the target word meets the requirement of the preset part of speech, filling the current word slot with the target word to obtain slot value pair information.

Optionally, the step of dynamically determining a next round of dialog execution text according to the current round of dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text, includes:

dynamically inputting the current turn conversation state information into a preset conversation strategy model;

predicting the current turn of dialog state information based on the preset dialog strategy model to obtain a prediction result of a dialog execution text corresponding to the execution;

obtaining the identity verification result of the applicant according to the prediction result of the dialog execution text corresponding to the execution;

the preset dialogue strategy model is a second target model obtained after iterative training is carried out on a second preset basic model based on dialogue state data with preset dialogue execution text labels.

acquiring a preset configuration file, wherein the preset configuration file stores a mapping relation between session state information and a corresponding session execution text;

and dynamically determining a next turn of dialogue execution text according to the preset configuration file and the current turn of dialogue state information, and determining the identity verification result of the applicant based on the dynamically determined dialogue execution text.

The application still provides an identity verification device, is applied to intelligent nuclear robot, the identity verification device includes:

the acquisition module is used for determining the conversation state information of the current turn according to the conversation content of the current turn when an identity verification instruction for verifying the identity of the applicant is detected;

and the determining module is used for dynamically determining the next turn of the dialog execution text according to the current turn of the dialog state information and determining the identity verification result of the applicant based on the dynamically determined dialog execution text.

Optionally, the obtaining module includes:

the output unit is used for outputting the inquiry question of the current turn of conversation when an identity verification instruction for verifying the identity of the applicant is detected;

a receiving unit, configured to receive response content input by the applicant for the question to obtain slot value pair information;

the first determining unit is used for determining the conversation state information of the intelligent core robot and the applicant in the previous round and the behavior information of the intelligent core robot in the previous round;

and the second determining unit is used for determining the conversation state information of the applicant in the identity verification process of the current round according to the slot value pair information, the conversation state information of the previous round and the behavior information of the previous round so as to determine the conversation state information of the current round according to the conversation content of the current round.

Optionally, the second determining unit is configured to implement:

Optionally, the receiving unit is configured to implement:

receiving response content input by the applicant for the question;

Optionally, the preset dependency syntax model includes a dependency discrimination model and a dependency type prediction model, and the receiving unit is further configured to implement:

vectorizing the reply content to obtain a vectorized statement;

Optionally, the receiving unit is further configured to implement:

Optionally, the determining module includes:

the input unit is used for dynamically inputting the current round of conversation state information into a preset conversation strategy model;

the prediction unit is used for performing prediction processing on the current turn of conversation state information based on the preset conversation strategy model to obtain a prediction result of a corresponding conversation execution text;

the first obtaining unit is used for obtaining the identity verification result of the applicant according to the prediction result of the dialog execution text corresponding to the execution;

Optionally, the determining module includes:

the second obtaining unit is used for obtaining a preset configuration file, wherein the preset configuration file stores a mapping relation between the conversation state information and the corresponding conversation execution text;

and the third determining unit is used for dynamically determining a next turn of conversation execution text according to the preset configuration file and the current turn of conversation state information, and determining the identity verification result of the applicant based on the dynamically determined conversation execution text.

The present application further provides an identity verification apparatus, which is an entity apparatus, the identity verification apparatus including: a memory, a processor and a program of the identity verification method stored on the memory and executable on the processor, which program, when executed by the processor, is operable to carry out the steps of the identity verification method as described above.

The present application also provides a readable storage medium having stored thereon a program for implementing the above-described identity verification method, which when executed by a processor implements the steps of the above-described identity verification method.

Compared with the prior art, the method, the equipment and the readable storage medium for verifying the identity information of the applicant are low in efficiency by inquiring the applicant once through a fixed conversation by using a fixed technology according to the preset problems in sequence, and the conversation state information of the current turn is determined according to the conversation content of the current turn when an identity verification instruction for verifying the identity of the applicant is detected; and dynamically determining a next turn of dialog execution text according to the current turn of dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text. In the application, when an identity verification instruction for verifying the identity of an applicant is detected, the identity verification is not performed according to a fixed full flow every time, but current round conversation state information, namely a dynamic conversation round is determined according to conversation contents of a current round, after the conversation round is dynamically determined, a next round conversation execution text is dynamically determined according to the current round conversation state information, namely a conversation strategy corresponding to different rounds of conversations is dynamically determined, and the conversation strategy corresponding to different rounds of conversations is dynamically determined due to the dynamic conversation round determination, so that after the identity verification result can be determined through fewer conversation flows, the identity verification result of the applicant is obtained through fewer conversation flows, therefore, the flow is saved, and the technical problem of low identity verification efficiency of the applicant in the prior art is solved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.

FIG. 1 is a schematic flow chart illustrating a first embodiment of an identity verification method according to the present application;

fig. 2 is a detailed flowchart illustrating steps of determining dialog state information of a current turn according to dialog contents of the current turn when an identity verification instruction for verifying an identity of an applicant is detected in the first embodiment of the identity verification method of the present application;

fig. 3 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present application.

The objectives, features, and advantages of the present application will be further described with reference to the accompanying drawings.

Detailed Description

It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

An embodiment of the present application provides an identity verification method, and in a first embodiment of the identity verification method of the present application, referring to fig. 1, the identity verification method is applied to an intelligent robot core, and the identity verification method includes:

step S10, when an identity verification instruction for verifying the identity of the applicant is detected, determining the conversation state information of the current turn according to the conversation content of the current turn;

step S20, dynamically determining a dialog execution text of the next round according to the dialog state information of the current round, and determining an identity verification result of the applicant based on the dynamically determined dialog execution text.

The method comprises the following specific steps:

in this embodiment, it should be noted that the identity verification method is applied to an identity verification system belonging to an intelligent self-verification robot, and for the intelligent self-verification robot, besides identifying the identity of an applicant (the applicant refers to a user who performs identity verification through the intelligent self-verification robot to achieve the purpose of loan at a financial institution, etc.) through hardware such as a camera, an OCR identifier, and the like (specifically, identifying a document image of the applicant through the OCR identifier, and collecting the image of the applicant through the camera to determine whether the collected image is consistent with the document image), in order to accurately perform identity verification on the applicant, also performing intelligent question-answer confirmation on the applicant, or also performing dialog confirmation on the applicant to accurately obtain the identity verification result of the applicant, it should be noted that, in this embodiment, in order to perform identity verification or intelligent question and answer, an intelligent question and answer is pre-stored locally in the intelligent core robot, where the intelligent question and answer includes information such as an identification number, a name, an age, a marital status, and a calendar status of an applicant to be asked, and the intelligent question and answer may be pulled from a preset server corresponding to the intelligent core robot in a remote manner.

At present, in the financial industry, especially in the banking industry, after an applicant applies for a loan, an intelligent self-checking robot generally queries the applicant for preset questions in a fixed manner, so that not only is the conversation efficiency low, but also the user experience is low, for example, if there are 10 preset questions, a traditional intelligent self-checking robot needs at least 10 conversations to complete information checking of one user, for some applicant, only two or three conversations are needed to judge the authenticity of information, specifically, for example, only an identity card number needs to be obtained, and the identity of the applicant can be checked in a marital state, because at least 10 conversations, i.e., 10 complete flows, are needed to complete the identity information checking of one applicant, waste of time and resources is caused, and the identity checking efficiency is reduced.

In order to solve the technical problem, in this embodiment, when an identity verification instruction for verifying the identity of the applicant is detected, current round session state information is determined according to the session content of the current round, further, a next round session execution text is dynamically determined according to the current round session state information, the identity verification result of the applicant is determined based on the dynamically determined session execution text, and since the session execution text, i.e., a session policy, is dynamically determined to verify the identity of the applicant, the fact that the identity of the applicant is accurately verified is avoided and then the verification process is continued, so that verification efficiency is improved.

Wherein the identity verification instruction comprises a verification instruction for performing identity verification in a dialogue mode, the identity verification instruction for performing identity verification by the applicant comprises an identity verification instruction for performing identity verification on the applicant for the first time or an identity verification instruction for performing identity verification on the applicant for a non-first time, wherein, when an identity verification instruction for verifying the identity of the applicant is detected for the first time, the first round of dialogue state information of the applicant is determined, if the identity verification instruction for verifying the identity of the applicant is not detected for the first time, the dialogue state information is associated with the reply content of the applicant for answering the intelligent question-answer question, wherein the dialog state information includes the number of rounds of inquiry, the number of times that the applicant answers the current intelligent question (current question inquiry), the verification difficulty level of the current intelligent question and answer questions, the times of the unmatched contents and the actual contents returned by the applicant and other information.

Referring to fig. 2, the step of determining the dialog state information of the current turn according to the dialog content of the current turn when the identity verification instruction for verifying the identity of the applicant is detected includes:

step S11, when an identity verification instruction for verifying the identity of the applicant is detected, outputting the question of the current turn of conversation;

in this embodiment, when an identity verification instruction for verifying the identity of an applicant is detected, outputting the question of the current round of session, where when an identity verification instruction for verifying the identity of an applicant is detected for the first time, the question of the current round of session is randomly selected to be output, or only a certain question, such as a question for outputting a challenge identification number, is selected to be output, and if the current round is not the first round, the manner of outputting the question of the current round of session includes:

the first method is as follows: an inquiry question having a preset verification relationship with the last polling inquiry question can be selected;

the second method comprises the following steps: determining a conversation strategy (conversation execution text) according to the reply content of the applicant in the previous round, and further outputting a question of the current round of conversation according to the conversation strategy, wherein the determined conversation strategy comprises the following steps: repeat the current question to ask or select the next round of questions and how to select the next round of questions to ask for the next round of questions.

Step S12, receiving the response content input by the applicant aiming at the question to obtain the slot value pair information;

in this embodiment, the response content input by the applicant for the question is received, and slot value pair information is obtained, where the slot value pair information refers to a slot value bit and a slot value, for example, the slot value bit is an age, and if the response content of the applicant is: if I is full of 28 years old this year, then 28 years old is the slot value, then the slot value pair information is: age, 28 years old.

Wherein the step of receiving the response content input by the applicant for the question to obtain the slot value pair information comprises:

step S121, receiving the reply content input by the applicant for the question;

in this embodiment, a specific process of how to obtain slot value pair information is specifically, first, receiving the response content input by the applicant for the question, where if the question is: please provide your name, the reply content or dialog content may be: i call XXX.

Step S122, based on a preset dependency syntax model, performing dependency syntax analysis on the reply content to obtain a dependency syntax analysis result corresponding to the reply content;

the method includes performing dependency syntax analysis on the reply content based on a preset dependency syntax model to obtain a dependency syntax analysis result corresponding to the reply content, specifically, the preset dependency syntax model is a pre-trained dependency syntax model and is used for performing dependency syntax analysis on a sentence, wherein the dependency syntax analysis process is a process for analyzing syntax information of the sentence, the syntax information includes sentence pattern information and word component information, for example, if the sentence is called "XXX", after the dependency syntax analysis, the sentence pattern information indicates that the sentence is a leading object sentence, the word component information indicates that "me" is a leading language, "called" is a predicate, and "XXX" is an object.

Obtaining reply content, performing dependency syntax analysis on the reply content based on a preset dependency syntax model to obtain a dependency syntax analysis result corresponding to the reply content, specifically, obtaining reply content, inputting the reply content into the preset dependency syntax model, vectorizing the reply content to obtain a vectorized statement, and further performing dependency relationship judgment and dependency relationship type prediction on the vectorized statement respectively, wherein it should be noted that the dependency relationship judgment is performed for judging the dependency relationship between words, and the dependency relationship type prediction is performed for predicting the type of a pre-stored relationship, for example, assuming that the reply content is a statement "ABC", where A, B and C are both words in the reply content, after the dependency relationship judgment, it can be determined that B depends on a, C depends on B, and after the dependency relationship type prediction is performed, determining that the dependency relationship between A and B is a dominance relationship, the dependency relationship between B and C is a predicate relationship, further obtaining a dependency relationship discrimination vector and a dependency relationship type prediction probability matrix corresponding to the vectorized sentence, further fusing the dependency relationship vector and the dependency relationship type prediction probability matrix, and obtaining a dependency relationship type label between words in the reply content, wherein the dependency relationship type label is an identifier of a dependency relationship type, and further determining the sentence pattern information and the word component information of the reply content based on the dependency relationship type label, and obtaining the dependency syntax analysis result, wherein a value on each bit in the dependency relationship type prediction probability matrix is a dependency relationship type label probability prediction vector between one word and another word in the reply content, the value of each bit in the dependency type prediction vector is a probability value of a preset dependency corresponding to the bit of a dependency between one word and another word in the reply content, where the preset dependency includes a dominance relation, a motile relation, and the like, for example, if the dependency type label probability prediction vector between the word a and the word B is (0.1, 0.9), 0.1 indicates that the probability between the word a and the word B is a dominance relation of 10%, and 0.9 indicates that the probability between the word a and the word B is a motile relation of 90%.

In an implementable approach, the step of fusing the dependency vector with the dependency type prediction probability matrix to obtain the word-to-word dependency type labels in the reply content includes:

and aggregating the dependency relationship vector and each dependency relationship type label probability prediction vector in the dependency relationship type prediction probability matrix to obtain an aggregated vector corresponding to each dependency relationship type label probability prediction vector, wherein the aggregation comprises weighted summation, splicing, summation and the like, then selecting a maximum bit value from bit values in the aggregated vector for each aggregated vector, and using a preset dependency relationship label corresponding to a bit corresponding to the maximum bit value as a dependency relationship type label corresponding to the dependency relationship type label probability prediction vector.

And step S123, based on the dependency syntax analysis result, performing word slot filling on a preset word slot to be filled corresponding to the reply content to obtain slot value pair information.

In this embodiment, it should be noted that the dependency syntax analysis result includes sentence information corresponding to the reply content and dependency relationship information between words in the reply content, where the dependency relationship information includes dependency relationship distinguishing information between words in the reply content and dependency relationship types between words in the reply content, where the dependency relationship distinguishing information is a probability that there is a dependency relationship between words in the reply content.

Performing word slot filling on a preset word slot to be filled corresponding to the reply content based on the dependency syntax analysis result to obtain slot value pair information, specifically, determining whether the reply content conforms to a preset target sentence pattern based on sentence pattern information corresponding to the reply content, if the reply content conforms to the preset target sentence pattern, extracting a word to be filled from the reply content based on a slot value bit position of a slot value bit in the preset target sentence pattern, if the extraction is successful, obtaining the word to be filled, further determining part of speech of the word to be filled based on dependency relationship information between words and words in the reply content, and determining whether part of speech of the word to be filled conforms to the slot value bit part of speech, if the part of speech of the word to be filled conforms to the slot value bit of speech, filling the word to the preset word slot to be filled corresponding to the reply content to obtain the slot value pair information, if the reply content does not accord with a preset target sentence pattern or the extraction of the words to be filled fails, or the part of speech of the words to be filled does not accord with the part of speech of the slot value bit, carrying out sequence labeling on the reply content based on a preset sequence labeling model to obtain a sequence labeling result, and carrying out word slot filling on the preset word slots to be filled based on the sequence labeling result to obtain the slot value pair information.

Step S13, determining the dialogue state information of the intelligent core robot and the applicant in the previous round and the behavior information of the intelligent core robot in the previous round;

in this embodiment, the dialog state information between the intelligent core robot and the applicant in the previous round is also determined, specifically, it is only required to directly acquire or obtain the dialog state information between the intelligent core robot and the applicant in the previous round, that is, after each round of dialog is completed, the intelligent core robot obtains the dialog state information, and the manner for the intelligent core robot to obtain the dialog state information includes:

the first method is as follows: obtaining dialogue state information through a preset dialogue state tracking model;

the second method comprises the following steps: and obtaining the dialog state information through statistics.

In this embodiment, the behavior information of the intelligent body-building robot in the previous round is further determined, and the behavior information of the intelligent body-building robot in the previous round includes: and repeatedly outputting the last inquiry question or selecting a new inquiry question.

Step S14, determining the dialog state information of the applicant in the current round identity verification process according to the slot value pair information, the dialog state information of the previous round and the behavior information of the previous round, so as to determine the dialog state information of the current round according to the dialog content of the current round.

And determining conversation state information of the applicant in the current round identity verification process according to the slot value pair information, the conversation state information of the previous round and the behavior information of the previous round, specifically, converting the slot value pair information, the conversation state information of the previous round and the behavior information of the previous round into input vectors, and inputting the input vectors into a preset conversation state tracking model to determine the conversation state information of the current round according to the conversation content of the current round, such as several rounds of conversations and the like.

In this embodiment, the dialog execution text is a dialog policy, the dialog execution text of the next round is dynamically determined according to the dialog state information of the current round, and the identity verification result of the applicant is determined based on the dynamically determined dialog execution text, it should be noted that the dialog content of the current round depends on the dialog execution text of the previous round, that is, the dialog execution text of the previous round corresponds to the dialog execution text of the previous round, and the dialog content of the next round is determined. If the dialog execution text of the current turn of dialog is determined to be the problem that the selection is associated with the previous turn and the difficulty level is higher than the previous turn level according to the first turn of dialog state information, or the dialog execution text of the current turn of dialog is determined to be the problem that the selection is not associated with the previous turn and the difficulty level is consistent according to the first turn of dialog state information.

The step of dynamically determining a dialog execution text of a next turn according to the dialog state information of the current turn, and determining an identity verification result of the applicant based on the dynamically determined dialog execution text, includes:

step S21, dynamically inputting the current turn of dialogue state information into a preset dialogue strategy model;

in this embodiment, the preset dialog strategy model is a trained model capable of accurately predicting dialog state information, specifically, the preset dialog strategy model is a second target model obtained by iteratively training a second preset basic model based on dialog state data with a preset dialog execution text label, where the dialog state data includes a dialog state training sentence set and a preset dialog execution text label corresponding to the training sentence set, and specifically, the dialog state training sentence set may be subjected to sample expansion in a manual labeling manner to obtain an expanded dialog state training sentence set.

Inputting the extended dialogue state training sentence set into the second preset basic model to analyze the extended dialogue state training sentence set to obtain a dialogue execution text label, specifically, vectorizing the extended dialogue state training sentence set based on a vectorization network in the second preset basic model to obtain a vectorization training sentence, further predicting a dialogue execution text of the vectorization training sentence based on the second preset basic model to obtain a predicted dialogue execution text label, comparing the predicted dialogue execution text label with the preset dialogue execution text label to obtain a comparison result, and continuously adjusting the model parameters of the second preset basic model in a reverse direction according to the comparison result until a preset training completion condition is reached to obtain a preset dialogue strategy model.

Step S22, based on the preset dialogue strategy model, carrying out prediction processing on the current turn dialogue state information to obtain a prediction result of the execution of the corresponding dialogue execution text;

step S23, obtaining the identity verification result of the applicant according to the prediction result of the dialog execution text corresponding to the execution;

And predicting the current turn of dialog state information based on the preset dialog strategy model to obtain a prediction result of a dialog execution text corresponding to execution, namely a dynamically determined dialog execution text, and obtaining an identity verification result of the applicant according to the prediction result of the dialog execution text corresponding to execution (the dynamically determined dialog execution text), wherein if the prediction result is a dialog ending, the dialog content is obtained to obtain the identity verification result of the applicant.

Wherein, the step of dynamically determining a dialog execution text of the next turn according to the dialog state information of the current turn, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text comprises:

step A1, acquiring a preset configuration file, wherein the preset configuration file stores a mapping relation between dialog state information and a corresponding dialog execution text;

in this embodiment, another way of determining a dialog execution text is provided, and specifically, the dialog execution text may be determined through a preset configuration file, where a mapping relationship between dialog state information and a corresponding dialog execution text is stored in the preset configuration file.

Step A2, dynamically determining the dialog execution text of the next turn according to the preset configuration file and the dialog state information of the current turn, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text.

And dynamically determining a next turn of dialog execution text according to the preset configuration file and the current turn of dialog state information, and determining an identity verification result of the applicant based on the dynamically determined dialog execution text, specifically, for example, if the dialog state information is first dialog state sub-information, determining the dialog execution text as the first dialog execution text according to a mapping relationship between the dialog state information and the corresponding dialog execution text, and if the dialog state information is second dialog state sub-information, determining the dialog execution text as the second dialog execution text according to a mapping relationship between the dialog state information and the corresponding dialog execution text.

In another embodiment of the identity verification method, the step of determining dialog state information with the applicant in the current round of identity verification process according to the slot value pair information, the dialog state information of the previous round and the behavior information of the previous round includes:

step B1, inputting the slot value pair information, the previous round of dialog state information and the previous round of behavior information as input information into a preset dialog state tracking model;

in this embodiment, a preset dialogue state tracking model is trained in advance to accurately determine dialogue state information, where the preset dialogue state tracking model is a first target model obtained by performing iterative training on a first preset basic model based on dialogue data with preset dialogue tags, where the dialogue data includes an input sentence set (formed by slot value pair information, previous turn dialogue state information, and previous turn behavior information) and the preset dialogue tags corresponding to the input sentence set, and specifically, the input sentence set may be subjected to sample expansion in a manual labeling manner to obtain an expanded input sentence set. Inputting the extended input sentence set into the first preset basic model to analyze the extended input sentence set to obtain a preset dialogue tag, specifically, vectorizing the extended input sentence set based on a vectorization network in the first preset basic model to obtain a vectorized training sentence, further predicting the dialogue tag of the vectorized training sentence based on the first preset basic model to obtain a predicted dialogue tag, comparing the predicted dialogue tag with the preset dialogue tag to obtain a comparison result, and continuously adjusting the model parameters of the first preset basic model in a reverse direction according to the comparison result until corresponding training completion conditions are reached to obtain a preset dialogue state tracking model.

Step B2, processing the input information based on the preset dialogue state tracking model to obtain dialogue state information of the applicant in the current round identity verification process;

And after a preset conversation state tracking model is obtained, predicting the input information based on the preset conversation state tracking model to obtain the conversation state information of the applicant in the current round identity verification process.

In this embodiment, the slot value pair information, the previous round of dialog state information, and the previous round of behavior information are input into a preset dialog state tracking model as input information; processing the input information based on the preset conversation state tracking model to obtain conversation state information of the applicant in the current round identity verification process; the preset dialogue state tracking model is a first target model obtained after iterative training is carried out on a first preset basic model based on dialogue data with preset dialogue labels. In the embodiment, a foundation is laid for accurately obtaining the conversation state information of the applicant in the current round identity verification process.

In another embodiment of the identity verification method, the preset dependency syntax model includes a dependency discrimination model and a dependency type prediction model, and the step of performing dependency syntax analysis on the reply content based on the preset dependency syntax model to obtain a dependency syntax analysis result corresponding to the reply content includes:

step C1, vectorizing the reply content to obtain vectorized sentences;

in this embodiment, the reply content is vectorized to obtain a vectorized statement, specifically, a word vector to be filled, a part-of-speech vector to be filled, and a word position vector to be filled corresponding to each word to be filled in the reply content are generated, wherein the word vector to be filled is a coding vector representing a word to be filled and is used for uniquely representing the word to be filled, the part-of-speech vector to be filled is a coding vector representing the part of speech of the word to be filled, the position vector of the word to be filled is a coding vector representing the position of the word to be filled in the reply content, and generating a vectorization word corresponding to each word to be filled based on the word vector to be filled corresponding to each word to be filled, the corresponding part-of-speech vector to be filled and the corresponding word position vector to be filled, and taking a matrix formed by each vectorization word as the vectorization statement.

Wherein the reply content at least comprises a word to be filled in, and the vectorization sentence at least comprises a vectorization word.

Step C2, based on the dependency relationship discrimination model, performing dependency relationship discrimination on the vectorized statement to obtain a dependency relationship discrimination result;

in this embodiment, it should be noted that the dependency relationship determination model is a machine learning model for determining whether there is a dependency relationship between words in the reply content.

And performing dependency relationship determination on the vectorized statement based on the dependency relationship determination model to obtain a dependency relationship determination result, specifically, inputting the vectorized statement into the dependency relationship determination model, performing dependency relationship determination on the vectorized statement to determine whether a dependency relationship exists between words in the reply content, and obtaining the dependency relationship determination result.

The dependency relationship judging model is a pre-trained model capable of accurately judging the dependency relationship of the vectorized sentences.

Step C3, based on the dependency relationship type prediction model and the dependency relationship determination result, performing dependency relationship type prediction on the vectorized statement to obtain the dependency syntax analysis result.

The step of performing word slot filling on a preset word slot to be filled corresponding to the reply content based on the dependency syntax analysis result to obtain slot value pair information includes:

a step D1 of extracting words to be filled in the reply content based on the dependency parsing result;

in this embodiment, after obtaining the dependency parsing result, extracting the word to be filled in the reply content, for example, assuming that the sentence pattern of the reply content is a predicate-object sentence pattern and the slot bit in the slot pair information is an age, the word to be filled is extracted at the object position of the reply content because the age usually appears at the object position.

Step D2, determining whether the sentence pattern of the reply content meets the requirement of the preset sentence pattern, and if the sentence pattern of the reply content meets the requirement of the preset sentence pattern, determining whether the position of the slot value bit appearing in the reply content is the preset position;

step D3, if the position of the slot value position in the reply content is not a preset position, inputting the reply content into a preset sequence labeling model to predict the reply content to obtain a content prediction result;

and D4, filling the current preset word slot according to the content prediction result to obtain slot value pair information.

And judging whether the sentence pattern of the reply content meets the requirement of a preset sentence pattern, specifically, comparing the target sentence pattern information representation of the reply content with each preset sentence pattern information representation to judge whether the reply content meets the preset target sentence pattern, namely, if each preset sentence pattern information representation has the sentence pattern information representation consistent with the target sentence pattern information representation, proving that the reply content meets the preset target sentence pattern, and if each preset sentence pattern information representation does not have the sentence pattern information representation consistent with the target sentence pattern information representation, proving that the reply content does not meet the preset target sentence pattern.

If the sentence pattern of the reply content does not meet the preset sentence pattern requirement, performing sequence labeling or prediction on the reply content based on a preset sequence labeling model to obtain a sequence labeling result (content prediction result), and based on the sequence labeling result, obtaining the slot value pair information, if the sentence pattern of the reply content meets the preset sentence pattern requirement, judging whether the position where the slot value bit appears in the reply content is a preset position, if the position where the slot value bit appears in the reply content is not the preset position, inputting the reply content into the preset sequence labeling model to predict the reply content to obtain a content prediction result, and if the position where the slot value bit appears in the reply content is the preset position, extracting the word to be filled in the sentence to be filled based on the position of the slot value bit.

After the step of determining whether the sentence pattern of the reply content meets the preset sentence pattern requirement and determining whether the position of the slot value bit appearing in the reply content is the preset position if the sentence pattern of the reply content meets the preset sentence pattern requirement, the method includes:

step E1, if the position of the slot value position appearing in the reply content is a preset position, extracting the target words in the preset position;

step E2, judging whether the part of speech of the target word meets the requirement of a preset part of speech;

and E3, if the part of speech of the target word meets the requirement of the preset part of speech, filling the current word slot with the target word to obtain slot value pair information.

In this embodiment, it should be noted that the part-of-speech of the bin bit is a part-of-speech that the bin bit set in advance may have, for example, assuming that the bin bit is an age, the bin bit may be a quantifier.

If the position of the slot value position appearing in the reply content is a preset position, extracting a target word in the preset position, judging whether the part of speech of the target word (the word to be filled) accords with the part of speech of the slot value position, if so, filling the word to be filled into the preset word slot to be filled, and obtaining the slot value pair information, and if not, obtaining the slot value pair information based on a preset sequence labeling model.

In the embodiment, the accurate acquisition of the slot value pair information is realized.

Referring to fig. 3, fig. 3 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present application.

As shown in fig. 3, the identity verification apparatus may include: a processor 1001, such as a CPU, a memory 1005, and a communication bus 1002. The communication bus 1002 is used for realizing connection communication between the processor 1001 and the memory 1005. The memory 1005 may be a high-speed RAM memory or a non-volatile memory (e.g., a magnetic disk memory). The memory 1005 may alternatively be a memory device separate from the processor 1001 described above.

Optionally, the identity verification device may further comprise a rectangular user interface, a network interface, a camera, RF (Radio Frequency) circuitry, sensors, audio circuitry, a WiFi module, and the like. The rectangular user interface may comprise a Display screen (Display), an input sub-module such as a Keyboard (Keyboard), and the optional rectangular user interface may also comprise a standard wired interface, a wireless interface. The network interface may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface).

It will be appreciated by those skilled in the art that the configuration of the identity verification device shown in figure 3 does not constitute a limitation of the identity verification device and may include more or less components than those shown, or some components in combination, or a different arrangement of components.

As shown in fig. 3, a memory 1005, which is a kind of computer-readable storage medium, may include therein an operating system, a network communication module, and an identity verification program. The operating system is a program that manages and controls the hardware and software resources of the identity verification device, supporting the operation of the identity verification program as well as other software and/or programs. The network communication module is used to enable communication between the various components within the memory 1005, as well as with other hardware and software in the identity verification system.

In the identity verification apparatus shown in fig. 3, the processor 1001 is configured to execute an identity verification program stored in the memory 1005 to implement the steps of the identity verification method described in any one of the above.

The specific implementation of the identity verification apparatus of the present application is substantially the same as that of each embodiment of the identity verification method, and is not described herein again.

Optionally, the obtaining module includes:

Optionally, the second determining unit is configured to implement:

Optionally, the receiving unit is configured to implement:

receiving response content input by the applicant for the question;

vectorizing the reply content to obtain a vectorized statement;

Optionally, the receiving unit is further configured to implement:

Optionally, the determining module includes:

The specific implementation of the identity verification apparatus of the present application is substantially the same as that of the above-mentioned embodiments of the identity verification method, and is not described herein again.

The present application provides a readable storage medium, and the readable storage medium stores one or more programs, which can be further executed by one or more processors for implementing the steps of the identity verification method described in any one of the above.

The specific implementation of the readable storage medium of the present application is substantially the same as that of each embodiment of the identity verification method, and is not described herein again.

The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings, or which are directly or indirectly applied to other related technical fields, are included in the scope of the present application.

Claims

1. An identity verification method is applied to an intelligent body-building robot, and comprises the following steps:

2. The identity verification method of claim 1, wherein the step of determining the dialog state information of the current turn based on the dialog contents of the current turn upon detection of an identity verification instruction for identity verification of the applicant comprises:

and determining the dialogue state information of the applicant in the identity verification process of the current round according to the slot value pair information, the dialogue state information of the previous round and the behavior information of the previous round.

3. The identity verification method of claim 2, wherein the step of determining dialog state information with the applicant in the current round of identity verification process based on the slot value pair information, the previous round of dialog state information, and the last round of behavior information comprises:

4. An identity verification method as claimed in claim 2, wherein the step of receiving the reply content input by the applicant to the challenge question to obtain slot value pair information comprises:

receiving response content input by the applicant for the question;

5. The identity verification method of claim 4, wherein the predetermined dependency syntax model includes a dependency discrimination model and a dependency type prediction model, and the step of performing dependency syntax analysis on the reply content based on the predetermined dependency syntax model to obtain the dependency syntax analysis result corresponding to the reply content comprises:

vectorizing the reply content to obtain a vectorized statement;

6. The identity verification method according to claim 4, wherein the step of performing word slot filling on a preset word slot to be filled corresponding to the reply content based on the dependency parsing result to obtain slot value pair information comprises:

7. The identity verification method of claim 6, wherein the step of determining whether the sentence pattern of the reply content meets a predetermined sentence pattern requirement, and if the sentence pattern of the reply content meets the predetermined sentence pattern requirement, determining whether the position of the slot value bit appearing in the reply content is a predetermined position is followed by the method comprising:

8. The identity verification method of claim 1, wherein the step of dynamically determining a next turn of the dialog execution text based on the current turn of the dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text, comprises:

9. The identity verification method of claim 1, wherein the step of dynamically determining a next turn of the dialog execution text based on the current turn of the dialog state information, and determining the identity verification result of the applicant based on the dynamically determined dialog execution text, comprises:

10. An identity verification apparatus, characterized in that the identity verification apparatus comprises: a memory, a processor, and a program stored on the memory for implementing the identity verification method,

the memory is used for storing a program for realizing the identity verification method;

the processor is adapted to execute a program implementing the identity verification method to implement the steps of the identity verification method as claimed in any one of claims 1 to 9.

11. A readable storage medium having stored thereon a program for implementing an identity verification method, the program being executable by a processor for implementing the steps of the identity verification method as claimed in any one of claims 1 to 9.