CN113705230B

CN113705230B - Method, device, equipment and medium for evaluating policy specifications based on artificial intelligence

Info

Publication number: CN113705230B
Application number: CN202111014577.4A
Authority: CN
Inventors: 叶向荣
Original assignee: Ping An Property and Casualty Insurance Company of China Ltd
Current assignee: Ping An Property and Casualty Insurance Company of China Ltd
Priority date: 2021-08-31
Filing date: 2021-08-31
Publication date: 2023-08-25
Anticipated expiration: 2041-08-31
Also published as: CN113705230A

Abstract

The application relates to the technical field of artificial intelligence, and discloses an evaluation method, device, equipment and medium of a policy holder based on artificial intelligence, wherein the method comprises the following steps: word segmentation is carried out on the to-be-evaluated policy holder to obtain a target policy holder phrase set; obtaining a target grammar error phrase set from the target policy special offer phrase set; carrying out semantic relevance assessment according to the target policy special about phrase set to obtain a target semantic relevance assessment result; according to the acquired policy special appointment semantic library, performing risk assessment on the target policy special appointment phrase set to obtain a target risk assessment result; and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result. The automatic grammar error evaluation, semantic relevance evaluation and risk evaluation of the order taker are realized, the requirements for knowledge of the insurance professional level and the special contract specification of the order taker are reduced, and the accuracy of the insurance contract is improved.

Description

Method, device, equipment and medium for evaluating policy specifications based on artificial intelligence

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to an artificial intelligence based policy holder about evaluation method, apparatus, device, and medium.

Background

In the insurance industry, since the format terms of the policy do not fully satisfy the needs of the customer, special contract terms (abbreviated as special contracts) need to be customized in the policy according to the individual needs of the customer. The special offers in the policy have a significant impact on the performance and risk management of the insurer and insurer with respect to the performance and/or reimbursement of subsequent policies. The traditional special appointment is a term manually added or edited by a policy issuing individual after communicating with a client, and because of complex and changeable client demands, the requirement on knowledge of insurance professional level and special appointment specifications is higher, and phenomena such as ambiguity, conflict with format description in a policy template and the like are easy to generate when the policy issuing individual appointment, so that the policy is invalid or potential risks exist.

Disclosure of Invention

The application mainly aims to provide an artificial intelligence-based method, device, equipment and medium for evaluating a policy special offer, which aim to solve the technical problems that the policy is invalid or has potential risks due to the fact that requirements of customers are complex and changeable, the requirements on knowledge of the insurance professional level and the special offer standard are relatively high, ambiguity is easy to generate when a single person makes a special offer, the phenomena of conflict with format description in a policy template and the like exist.

In order to achieve the above object, the present application provides an artificial intelligence based method for evaluating a policy holder, the method comprising:

acquiring a special offer evaluation request corresponding to a target policy, wherein the special offer evaluation request carries a policy offer to be evaluated;

word segmentation is carried out on the to-be-evaluated policy holder to obtain a target policy holder phrase set;

finding out the phrase with grammar error from the target policy special about phrase set to obtain a target grammar error phrase set;

carrying out semantic relevance assessment according to the target policy special about phrase set to obtain a target semantic relevance assessment result;

performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result;

and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

Further, the step of obtaining the offer evaluation request of the target policy includes:

acquiring an incremental special examination request corresponding to the target policy;

responding to the increment special examination request, and acquiring a special segment with editing from the target policy as the policy special to be evaluated;

And generating the offer evaluation request according to the to-be-evaluated policy offer and the policy identifier corresponding to the target policy.

Further, before the step of performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result, the method further includes:

acquiring a semantic library update request, wherein the semantic library update request carries a policy identification set to be analyzed;

matching a policy information field set from an obtained policy database according to each policy identifier in the policy identifier set to be analyzed to obtain a policy information field set to be analyzed corresponding to each policy identifier;

respectively performing word segmentation and stop word deletion on the special policy fields in the to-be-analyzed policy information field set to obtain a to-be-analyzed policy about phrase set;

carrying out text weight calculation of each phrase according to the policy about phrase set to be analyzed to obtain a text weight set;

according to the complaint information field and the claim information field corresponding to the to-be-analyzed policy about phrase set, calculating the risk weight score of each phrase in the to-be-analyzed policy about phrase set to obtain a risk weight diversity;

And updating the policy about semantic library according to the policy information field set to be analyzed, the policy about phrase set to be analyzed, the text weight set and the risk weight set.

Further, the step of calculating a risk weight score for each phrase in the policy about phrase set to be analyzed according to the complaint information field and the claim information field corresponding to the policy about phrase set to be analyzed, to obtain a risk weight diversity includes:

performing risk score evaluation of the policy according to the complaint information fields and the claim information fields in the policy special offer phrase set to be analyzed to obtain a policy risk score;

performing word frequency calculation on each phrase in the to-be-analyzed policy about phrase set to obtain a word frequency set;

and according to the policy risk score and the word frequency set, calculating the risk weight score of each phrase in the policy about phrase set to be analyzed to obtain the risk weight diversity.

Further, the step of performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result includes:

performing similarity calculation on each special term in the special term semantic library and each term in the target special term phrase set to obtain a similarity set to be analyzed corresponding to each term in the target special term phrase set;

Obtaining a preset similarity threshold, and finding out the similarity larger than the preset similarity threshold from the similarity set to be analyzed as a target similarity set;

acquiring each text weight corresponding to each similarity in the target similarity set from the policy special about semantic library to obtain a text weight set to be analyzed;

acquiring a preset text weight threshold, and finding out text weights larger than the preset text weight threshold from the text weight set to be analyzed as a target text weight set;

and acquiring complaint information fields and claim information fields corresponding to each target text weight set from the policy special offer semantic library to obtain the target risk assessment result.

Further, after the step of generating the offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result, the method further includes:

acquiring a special appointment auditing request of the target policy;

responding to the offer auditing request, and acquiring an insurance policy from the target insurance policy as the insurance policy to be audited;

word segmentation is carried out on the policy special offers to be audited, and a policy special offer phrase set to be audited is obtained;

Finding out the phrase with grammar error from the to-be-checked policy special about phrase set to obtain a to-be-processed grammar error phrase set;

according to the grammar error phrase set to be processed, carrying out problem special section proportion statistics on the policy special to be audited to obtain target problem special section proportion;

according to the policy special offer semantic library, performing risk auditing on each phrase in the policy special offer phrase set to be audited to obtain a risk auditing result;

and generating a special examination result according to the grammar error phrase set to be processed, the target problem special examination fragment proportion and the risk examination result.

Further, the step of performing risk auditing on each phrase in the to-be-audited policy appointment phrase set according to the policy appointment semantic library to obtain a risk auditing result includes:

matching each risk weight score of each phrase in the to-be-checked policy about phrase set from the policy about semantic library to serve as a to-be-analyzed risk weight diversity corresponding to each phrase in the to-be-checked policy about phrase set;

acquiring a preset risk weight sub-threshold, and finding out a risk weight sub-threshold larger than the preset risk weight sub-threshold from the risk weight diversity to be analyzed to obtain a target risk weight diversity;

Acquiring a policy information field set corresponding to each risk weight in the target risk weight scores from the policy special about semantic library to obtain a policy information field set to be displayed;

performing target risk weight diversity calculation which is not null according to each target risk weight diversity as the number of risk phrases;

and taking the number of the risk phrases and each to-be-displayed policy information field set as the risk auditing result.

The application also provides an artificial intelligence-based device for evaluating the policy specifications, which comprises:

the request acquisition module is used for acquiring an offer evaluation request corresponding to the target policy, wherein the offer evaluation request carries the policy offer to be evaluated;

the word segmentation module is used for segmenting the to-be-evaluated policy holder to obtain a target policy holder phrase set;

the target grammar error phrase set determining module is used for finding out the phrases with grammar errors from the target policy special offer phrase set to obtain a target grammar error phrase set;

the target semantic relevance evaluation result determining module is used for evaluating semantic relevance according to the target policy special about phrase set to obtain a target semantic relevance evaluation result;

The target risk assessment result determining module is used for carrying out risk assessment on the target policy holder phrase set according to the acquired policy holder semantic library to obtain a target risk assessment result;

and the evaluation result display module is used for generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relativity evaluation result and the target risk evaluation result.

The application also proposes a computer device comprising a memory storing a computer program and a processor implementing the steps of any of the methods described above when the processor executes the computer program.

The application also proposes a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the method of any of the above.

According to the artificial intelligence-based method, device, equipment and medium for evaluating the policy holder, firstly, a target policy holder phrase set is obtained by segmenting the policy holder to be evaluated, secondly, grammar error phrases are found out from the target policy holder phrase set to obtain the target grammar error phrase set, semantic relevance evaluation is carried out according to the target policy holder phrase set to obtain a target semantic relevance evaluation result, risk evaluation is carried out on the target policy holder phrase set according to the obtained policy holder semantic library to obtain a target risk evaluation result, and finally, a feature evaluation result corresponding to the target policy is generated according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result, so that automatic grammar error evaluation, semantic relevance evaluation and risk evaluation are carried out on single operators, requirements on the insurance level and feature specification of the single operators are reduced, the accuracy of the policy holder is improved, and the policy holder is prevented from being invalid or having potential risks; and moreover, manual inspection of single person is avoided, and the efficiency of single person is improved.

Drawings

FIG. 1 is a flow chart of an artificial intelligence based policy about evaluation method according to an embodiment of the application;

FIG. 2 is a block diagram schematically illustrating an apparatus for evaluating policy specifications based on artificial intelligence according to an embodiment of the present application;

fig. 3 is a schematic block diagram of a computer device according to an embodiment of the present application.

The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

Referring to fig. 1, in an embodiment of the present application, there is provided a method for evaluating an artificial intelligence-based policy, including:

s1: acquiring a special offer evaluation request corresponding to a target policy, wherein the special offer evaluation request carries a policy offer to be evaluated;

s2: word segmentation is carried out on the to-be-evaluated policy holder to obtain a target policy holder phrase set;

s3: finding out the phrase with grammar error from the target policy special about phrase set to obtain a target grammar error phrase set;

S4: carrying out semantic relevance assessment according to the target policy special about phrase set to obtain a target semantic relevance assessment result;

s5: performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result;

s6: and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

According to the method, firstly, a target policy holder phrase set is obtained by word segmentation of the to-be-evaluated policy holder, secondly, grammar error phrases are found out from the target policy holder phrase set to obtain the target grammar error phrase set, semantic relevance evaluation is carried out according to the target policy holder phrase set to obtain a target semantic relevance evaluation result, risk evaluation is carried out on the target policy holder phrase set according to an obtained policy holder semantic library to obtain a target risk evaluation result, and finally, a feature evaluation result corresponding to the target policy holder is generated according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result, so that automatic grammar error evaluation, semantic relevance evaluation and risk evaluation are carried out on a single person, requirements for knowledge of insurance specialty level and feature specifications of the single person are reduced, the accuracy of policy holder invalidation or potential risk is avoided; and moreover, manual inspection of single person is avoided, and the efficiency of single person is improved.

For S1, an offer evaluation request corresponding to a target policy input by a user (for example, a person who goes out of a list) may be obtained, or an offer evaluation request corresponding to a target policy input by a third party application system may be obtained, or an offer evaluation request corresponding to a target policy that is actively triggered by a program of the present application may be obtained. For example, the special offer evaluation request corresponding to the target policy actively triggered by the program of the present application is implemented when a single person inputs a symbol, and the present application is not limited in detail.

Target policy, i.e., policy that requires an offer assessment. The policy generally refers to an insurance policy, which is written proof that an insurer has signed an insurance contract with an applicant.

The offer evaluation request, i.e., a request to evaluate an offer in a target policy.

The policy offer to be evaluated, i.e., the offer evaluation request, is an offer that is desired to be evaluated. It is understood that the policy to be evaluated may be all the policies in the target policy, or may be some of the policies in the target policy, which is not limited herein.

The policy specifier, also known as an insurance special agreement, is a special agreement term in the policy, which is agreed with the insurance company via the applicant and is attached to a special agreement or term on the policy.

And S2, performing word segmentation on the to-be-evaluated policy treatises by adopting a Principal Component Analysis (PCA), and taking each phrase obtained by word segmentation as a target policy treatise phrase set.

Optionally, a word segmentation dictionary is adopted to segment the to-be-evaluated policy holder, and each phrase obtained by word segmentation is used as a target policy holder phrase set.

The word segmentation dictionary includes one or more phrases.

And S3, acquiring a grammar error detection model, inputting the target policy special phrase set into the grammar error detection model to search for grammar error phrases, and taking each phrase output by the grammar error detection model as a target grammar error phrase set.

The grammar error detection model is a model obtained based on a dependency tree algorithm and machine learning training. And the grammar error detection model is used for carrying out long-distance grammar error detection.

And S4, acquiring a semantic relevance scoring model, inputting the target policy holder phrase set into a semantic relevance detection model to evaluate the semantic relevance in the target policy holder phrase set, and taking evaluation data output by the semantic relevance scoring model as a target semantic relevance evaluation result. That is, the target semantic relevance evaluation result is an evaluation result of semantic relevance inside the policy to be evaluated, and the target semantic relevance evaluation result is an evaluation result of semantic relevance between the respective terms in the target policy phrase set.

The semantic relevance scoring model is a model trained based on an initial model. The initial model includes a convolution layer, a pooling layer, a full connection layer, and the like.

For S5, the policy about semantic library can be obtained from the database, and the policy about semantic library can also be obtained from a third-party application system.

The policy special offer semantic library comprises: special about phrase, policy identification, text weight, risk weight score, policy risk score, and policy information field set, wherein the policy information field set comprises: the method comprises the steps of a policy basic information field, a policy special offer field, a complaint information field and a claim information field, wherein special phrases are phrases in the policy special offer field, text weights are text weights of special phrases in the policy special offer field, and risk weight scores are scores occupied by the special phrases in the policy risk scores. The policy identifier may be data that uniquely identifies a policy, such as a policy name, a policy ID, etc. The policy risk score is a policy risk score calculated from the complaint information field and the claim information field. The policy basic information field is basic information of the policy, such as an insurance start-stop period, an insurance standard, and is not specifically limited herein by way of example. The policy specification field is a specification of the policy. The complaint information field is complaint information of a policy, such as a complaint category, a complaint subject, a complaint number, and the like, and is not particularly limited. The claim information field is claim information of a policy, such as a type of claim, an amount of claim, and the number of claims, and is not particularly limited.

And matching each phrase in the target special appointment phrase set with special appointment phrases with similarity meeting the similarity requirement in a special appointment semantic library, enabling text weight corresponding to the matched special appointment phrases to meet the text weight requirement, and taking complaint information fields and claim information fields corresponding to the special appointment phrases meeting the similarity requirement and the text weight as target risk assessment results.

And S6, adopting a preset special offer evaluation result generation specification, and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

And displaying the offer evaluation result by adopting a preset display mode, so as to be used for carrying out grammar verification, risk retrieval and error reminding of semantic relevance on the order taker of the target policy.

Preset display modes include, but are not limited to: the bullet frame is displayed and marked in the target policy in a preset format.

Optionally, the present application further includes: acquiring an association request corresponding to the target policy, wherein the association request carries a phrase to be associated; and matching the special phrases meeting the similarity requirement and the text weight in a special meaning library of the to-be-associated phrase, and taking the special field of the to-be-associated phrase as an association result corresponding to the to-be-associated phrase. Therefore, the requirements on knowledge of the insurance professional level and special specifications of the order taker are further reduced, and the efficiency of the order taker is further improved.

In one embodiment, the step of obtaining the offer evaluation request of the target policy includes:

s11: acquiring an incremental special examination request corresponding to the target policy;

s12: responding to the increment special examination request, and acquiring a special segment with editing from the target policy as the policy special to be evaluated;

s13: and generating the offer evaluation request according to the to-be-evaluated policy offer and the policy identifier corresponding to the target policy.

The method and the device for generating the special offer assessment request according to the special offer fragment with the editing function are realized, so that the efficiency of responding to the special offer assessment request is improved, the method and the device are favorable for assisting the order taker to quickly correct the error of the special offer of the target special offer, the accuracy of the special offer is improved, and the efficiency of the order taker is improved.

For S11, the incremental special examination request corresponding to the target policy input by the user may be obtained, or the incremental special examination request corresponding to the target policy may be obtained from a third party application system, or the incremental special examination request corresponding to the target policy that is actively triggered by the program of the present application may be obtained. For example, when the order taker edits the target policy, the order taker enters punctuation marks that will actively trigger the incremental special review request, examples of which are not specifically limited herein. For another example, when the order taker edits the target policy, the order taker inputs a pause duration exceeding a preset pause duration, and actively triggers the incremental special examination request, which is not specifically limited herein.

The delta offer check request is a request for an offer assessment of the presence of an edited offer fragment.

For S12, responding to the incremental special examination request, taking the text between two punctuations in the target policy as a special segment, taking the text before the first punctuation in the target policy as a special segment, and taking the text after the last punctuation in the target policy as a special segment; and judging whether editing exists in each special section of the target policy according to the policy offer corresponding to the latest special evaluation request of the target policy, and taking each special section with editing as the policy offer to be evaluated when the judgment result is that the editing exists. That is, each offer fragment of the policy offer to be evaluated is an offer fragment in which an edit exists.

Punctuation marks are written symbols used to designate periods and moods.

And S13, when the offer evaluation request is generated according to the to-be-evaluated policy and the policy identifier corresponding to the target policy, taking the to-be-evaluated policy and the policy identifier corresponding to the target policy as parameters of the offer evaluation request.

In an embodiment, before the step of performing risk assessment on the target policy phrase set according to the obtained policy about semantic library to obtain a target risk assessment result, the method further includes:

s511: acquiring a semantic library update request, wherein the semantic library update request carries a policy identification set to be analyzed;

s512: matching a policy information field set from an obtained policy database according to each policy identifier in the policy identifier set to be analyzed to obtain a policy information field set to be analyzed corresponding to each policy identifier;

s513: respectively performing word segmentation and stop word deletion on the special policy fields in the to-be-analyzed policy information field set to obtain a to-be-analyzed policy about phrase set;

s514: carrying out text weight calculation of each phrase according to the policy about phrase set to be analyzed to obtain a text weight set;

s515: according to the complaint information field and the claim information field corresponding to the to-be-analyzed policy about phrase set, calculating the risk weight score of each phrase in the to-be-analyzed policy about phrase set to obtain a risk weight diversity;

s516: and updating the policy about semantic library according to the policy information field set to be analyzed, the policy about phrase set to be analyzed, the text weight set and the risk weight set.

The embodiment provides a basis for risk assessment based on the obtained security contract semantic library by updating the security contract semantic library before the obtained security contract semantic library.

For S511, the semantic library update request input by the user may be acquired, or the semantic library update request may be acquired from the third party application system.

Semantic library update requests, i.e., requests for updates to the policy specification semantic library.

The set of policy identifications to be analyzed, i.e., the set of policy identifications for which the semantic library update request wants to update to the policy about semantic data to the policy about semantic library. The set of policy identifications to be analyzed includes one or more policy identifications.

For S512, the policy database includes: a policy identifier and a set of policy information fields.

And searching each policy identifier in the policy identifier set to be analyzed in each policy identifier of the policy database, and taking the policy information field set corresponding to each policy identifier searched in the policy database as a policy information field set to be analyzed.

For S513, segmenting the policy about field in the policy information field set to be analyzed, and taking each phrase obtained by segmentation as a policy about phrase set of the stop word to be deleted; acquiring a stop word set, deleting each stop word in the stop word set from the to-be-deleted stop word's policy about phrase set, and taking the to-be-deleted stop word's policy about phrase set after deletion as the to-be-analyzed policy about phrase set.

The stop word set includes one or more stop words. Stop Words refer to that in information retrieval, certain Words or Words are automatically filtered before or after processing natural language data (or text) in order to save storage space and improve search efficiency, and are called Stop Words.

Optionally, a principal component analysis method is adopted to divide words of the special policy fields in the to-be-analyzed policy information field set respectively, and each phrase obtained by dividing words is used as a special policy phrase set to be de-duplicated.

Optionally, a word segmentation dictionary is adopted to segment the special policy fields in the to-be-analyzed policy information field set respectively, and each phrase obtained by word segmentation is used as a to-be-de-duplicated special policy phrase set.

For S514, a document frequency method is adopted, text weight calculation of each phrase is performed according to the to-be-analyzed policy phrase set, and each calculated text weight is used as a text weight set.

That is, the text weight is the text weight of the phrase in one of the guaranty's fields.

For S515, according to the complaint information field and the claim information field corresponding to the policy about phrase set to be analyzed, performing risk score evaluation on the policy corresponding to the policy about phrase set to be analyzed, and then performing risk weight score calculation on each phrase according to the policy about phrase set to be analyzed and the risk score, and taking each calculated risk weight score as risk weight diversity.

And for S516, taking the set of the to-be-analyzed policy information fields, the set of the to-be-analyzed policy about phrases, the set of the text weights and the diversity of the risk weights, which correspond to the same policy identifier, as associated data, and updating the associated data into the policy about semantic library.

In one embodiment, the step of calculating the risk weight score for each phrase in the to-be-analyzed policy phrase set according to the complaint information field and the claim information field corresponding to the to-be-analyzed policy phrase set to obtain the risk weight diversity includes:

s5151: performing risk score evaluation of the policy according to the complaint information fields and the claim information fields in the policy special offer phrase set to be analyzed to obtain a policy risk score;

s5152: performing word frequency calculation on each phrase in the to-be-analyzed policy about phrase set to obtain a word frequency set;

s5153: and according to the policy risk score and the word frequency set, calculating the risk weight score of each phrase in the policy about phrase set to be analyzed to obtain the risk weight diversity.

According to the embodiment, risk score evaluation is carried out on the insurance policy corresponding to the to-be-analyzed insurance policy phrase set according to the complaint information field and the claim information field corresponding to the to-be-analyzed insurance policy phrase set, and then risk weight score calculation of each phrase is carried out according to the to-be-analyzed insurance policy phrase set and the risk score, so that the risk weight score of each phrase is obtained, and a basis is provided for risk evaluation based on the insurance policy semantic library.

And S5151, adopting a preset risk score evaluation model, carrying out risk score evaluation of the policy according to the complaint information fields and the claim information fields in the to-be-analyzed policy special about phrase set, and taking the score obtained by evaluation as the policy risk score.

The preset risk score evaluation model may be a model for classification prediction, and each classification label is a risk score range. It may be understood that the start value of the risk score range may be used as the risk score corresponding to the classification label, the end value of the risk score range may be used as the risk score corresponding to the classification label, and the average value of the risk score range may be used as the risk score corresponding to the classification label.

It is to be understood that the preset risk score evaluation model may be other models, which are not limited herein.

For S5152, performing word frequency calculation in the to-be-analyzed policy phrase set on each phrase in the to-be-analyzed policy phrase set, and taking each calculated word frequency as a word frequency set.

Word frequency is a commonly used weighting technique for information retrieval and text mining to evaluate the degree of repetition of a word to a set of domain documents in a document or corpus.

For S5153, a word frequency corresponding to the target phrase in the word frequency set is multiplied by the policy risk score, and the data obtained by the multiplication is used as a risk weight score corresponding to the target phrase, where the target phrase is any phrase in the policy about phrase set to be analyzed.

In one embodiment, the step of performing risk assessment on the target policy phrase set according to the obtained policy about semantic library to obtain a target risk assessment result includes:

s521: performing similarity calculation on each special term in the special term semantic library and each term in the target special term phrase set to obtain a similarity set to be analyzed corresponding to each term in the target special term phrase set;

s522: obtaining a preset similarity threshold, and finding out the similarity larger than the preset similarity threshold from the similarity set to be analyzed as a target similarity set;

s523: acquiring each text weight corresponding to each similarity in the target similarity set from the policy special about semantic library to obtain a text weight set to be analyzed;

s524: acquiring a preset text weight threshold, and finding out text weights larger than the preset text weight threshold from the text weight set to be analyzed as a target text weight set;

S525: and acquiring complaint information fields and claim information fields corresponding to each target text weight set from the policy special offer semantic library to obtain the target risk assessment result.

According to the embodiment, the special phrases with the similarity meeting the similarity requirement are matched in the special security semantic library according to each phrase in the target special security phrase set, the text weight corresponding to the matched special phrases meets the text weight requirement, and the complaint information field and the claim information field corresponding to the special phrases meeting the similarity requirement and the text weight are taken as target risk assessment results, so that automatic risk assessment is carried out on the order taker, the requirements for knowing the insurance professional level and special specification of the order taker are reduced, the accuracy of the special security is improved, and the order taking is prevented from being invalid or having potential risks.

For S521, a KMP (Knuth-Morria-Pratt) algorithm is adopted to calculate the similarity between each special phrase in the special case semantic library and each phrase in the target special case phrase set, and each similarity corresponding to one phrase in the target special case phrase set is used as the similarity set to be analyzed. That is, the number of similarity sets to be analyzed is the same as the number of phrases of the target policy about phrase set.

It is to be understood that other algorithms may be used to calculate the similarity between each special term in the special term semantic library and each term in the target special term phrase set, which is not limited herein.

For S522, the preset similarity threshold may be obtained from the database, or may be obtained from the third party application system, or may be written into a program for implementing the present application. The preset similarity threshold is a specific value.

And finding out the similarity greater than the preset similarity threshold value from the similarity set to be analyzed, and taking each found similarity as a target similarity set. That is, the similarities in the target similarity set are all greater than the preset similarity threshold.

And for S523, finding out the special about phrases corresponding to each similarity in the target similarity set, and taking each text weight corresponding to each found special about phrase in the policy special about semantic library as a text weight set to be analyzed. And finding a text weight set to be analyzed corresponding to the special phrase with the similarity meeting the similarity requirement (namely, the similarity is larger than the preset similarity threshold value).

For S524, the preset text weight threshold may be obtained from the database, or may be obtained from a third party application system, or may be written into a program for implementing the present application. The preset text weight threshold is a specific value.

And finding out text weights larger than the preset text weight threshold from the text weight set to be analyzed, and taking each found text weight larger than the preset text weight threshold as a target text weight set. That is, the text weights in the target text weight set are each greater than the preset text weight threshold. Thereby, a target text weight set corresponding to the special phrase with the similarity meeting the similarity requirement (namely, being larger than the preset similarity threshold value) and the text weight meeting the text weight requirement (namely, being larger than the preset text weight threshold value) is found.

And for S525, acquiring complaint information fields and claim information fields corresponding to the text weights in each target text weight set from the policy special offer semantic library, and taking the acquired complaint information fields and claim information fields as the target risk assessment result.

In one embodiment, after the step of generating the offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result, the method further includes:

s71: acquiring a special appointment auditing request of the target policy;

s72: responding to the offer auditing request, and acquiring an insurance policy from the target insurance policy as the insurance policy to be audited;

s73: word segmentation is carried out on the policy special offers to be audited, and a policy special offer phrase set to be audited is obtained;

s74: finding out the phrase with grammar error from the to-be-checked policy special about phrase set to obtain a to-be-processed grammar error phrase set;

s75: according to the grammar error phrase set to be processed, carrying out problem special section proportion statistics on the policy special to be audited to obtain target problem special section proportion;

s76: according to the policy special offer semantic library, performing risk auditing on each phrase in the policy special offer phrase set to be audited to obtain a risk auditing result;

s77: and generating a special examination result according to the grammar error phrase set to be processed, the target problem special examination fragment proportion and the risk examination result.

The embodiment realizes automatic reminding of grammar error audit, semantic relevance audit and risk audit for the auditor, reduces the requirements on the knowledge of the insurance professional level and special rules of the auditor, improves the auditing accuracy of the auditor, further improves the accuracy of the policy special rules, and avoids the invalidity or the potential risk of the policy; and only the auditor is required to check the special audit result, so that the efficiency of the auditor is improved.

For S71, a request for checking the offer of the target policy sent by the individual may be obtained, or the request for checking the offer of the target policy may be obtained from a third party application system.

The offer auditing request is a request for auditing an offer in a target policy.

And S72, responding to the special examination request, acquiring a policy special from the target policy, and taking the acquired policy special as the policy special to be examined.

The policy offers to be audited are all the offers in the target policy.

And S73, segmenting the to-be-audited policy, and taking each phrase obtained by segmentation as a to-be-audited policy phrase set by adopting a principal component analysis method.

Optionally, a word segmentation dictionary is adopted to segment the policy and appointment to be audited, and the policy and appointment phrase set to be audited is obtained.

And for S74, inputting the to-be-audited policy about phrase set into the grammar error detection model to search for grammar error phrases, and taking each phrase output by the grammar error detection model as the to-be-processed grammar error phrase set.

For S75, searching problematic special section of the policy special to be audited according to the grammar error phrase set to be processed; dividing the number of the found problematic offer fragments by the total number of the offer fragments in the to-be-checked policy offer, and taking the data obtained by the division as the target problem offer fragment proportion.

And for S76, matching each phrase in the to-be-audited special phrase set with special phrases with similarity meeting the similarity requirement in a special semantic library of the special phrases, enabling the text weight corresponding to the matched special phrases to meet the text weight requirement, and taking the special information field set and the special phrase quantity corresponding to the special phrases meeting the similarity requirement and the text weight as risk auditing results.

And S77, adopting a preset special contract auditing result generation specification to generate a special contract auditing result according to the grammar error phrase set to be processed, the target problem special segment proportion and the risk auditing result.

In one embodiment, the step of performing risk auditing on each phrase in the to-be-audited policy appointment phrase set according to the policy appointment semantic library to obtain a risk auditing result includes:

s761: matching each risk weight score of each phrase in the to-be-checked policy about phrase set from the policy about semantic library to serve as a to-be-analyzed risk weight diversity corresponding to each phrase in the to-be-checked policy about phrase set;

s762: acquiring a preset risk weight sub-threshold, and finding out a risk weight sub-threshold larger than the preset risk weight sub-threshold from the risk weight diversity to be analyzed to obtain a target risk weight diversity;

s763: acquiring a policy information field set corresponding to each risk weight in the target risk weight scores from the policy special about semantic library to obtain a policy information field set to be displayed;

s764: performing target risk weight diversity calculation which is not null according to each target risk weight diversity as the number of risk phrases;

S765: and taking the number of the risk phrases and each to-be-displayed policy information field set as the risk auditing result.

In the embodiment, each phrase in the to-be-checked special security phrase set is matched with the special phrase with the similarity meeting the similarity requirement in the special security semantic library, the text weight corresponding to the matched special phrase meets the text weight requirement, and the special information field set and the special phrase number corresponding to the special phrase meeting the similarity requirement and the text weight are used as risk checking results, so that automatic reminding of risk checking for an auditor is realized, the requirements for knowing the insurance specialty level and special specification of the auditor are reduced, the auditing accuracy of the auditor is improved, the accuracy of the special security is further improved, and the invalidity or potential risk of the security is avoided.

And for S761, matching each risk weight score of each phrase in the to-be-checked policy and appointment phrase set from the policy and appointment semantic library, and taking each risk weight score matched by each phrase as a to-be-analyzed risk weight diversity. That is, the number of risk weight diversity to be analyzed is the same as the number of phrases in the set of security appointment phrases to be reviewed.

For S762, the preset risk weight sub-threshold may be obtained from the database, or may be obtained from the third party application system, or may be written into a program for implementing the present application. The preset risk weight sub-threshold is a specific value.

And finding out risk weight scores larger than the preset risk weight score threshold value from the risk weight diversity to be analyzed, and taking each found risk weight score larger than the preset risk weight score threshold value as a target risk weight diversity.

For S763, acquiring, from the policy about semantic library, a policy information field set corresponding to each of the target risk weight scores, and taking each acquired policy information field set as a policy information field set to be displayed.

And for S764, performing target risk weight diversity calculation which is not null according to each target risk weight diversity, thereby determining the number of phrases with risk weights greater than the preset risk weight threshold value in the to-be-checked policy and special about phrase set.

And for S765, directly taking the number of the risk phrases and each to-be-displayed policy information field set as the risk auditing result.

Referring to fig. 2, the application further provides an artificial intelligence based device for evaluating a policy offer, which comprises:

the request acquisition module 100 is configured to acquire an offer evaluation request corresponding to a target policy, where the offer evaluation request carries a policy offer to be evaluated;

the word segmentation module 200 is configured to segment the to-be-evaluated policy holder to obtain a target policy holder phrase set;

a target grammar error phrase set determining module 300, configured to find a phrase with grammar errors from the target policy specification phrase set, and obtain a target grammar error phrase set;

the target semantic relevance evaluation result determining module 400 is configured to perform semantic relevance evaluation according to the target policy about phrase set to obtain a target semantic relevance evaluation result;

the target risk assessment result determining module 500 is configured to perform risk assessment on the target policy about phrase set according to the obtained policy about semantic library, so as to obtain a target risk assessment result;

the evaluation result display module 600 is configured to generate a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

Referring to fig. 3, in an embodiment of the present application, there is further provided a computer device, which may be a server, and an internal structure thereof may be as shown in fig. 3. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the computer is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing data such as an evaluation method of the policy specifications based on artificial intelligence. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of evaluating an artificial intelligence based policy agreement. The artificial intelligence-based policy appointment assessment method comprises the following steps: acquiring a special offer evaluation request corresponding to a target policy, wherein the special offer evaluation request carries a policy offer to be evaluated; word segmentation is carried out on the to-be-evaluated policy holder to obtain a target policy holder phrase set; finding out the phrase with grammar error from the target policy special about phrase set to obtain a target grammar error phrase set; carrying out semantic relevance assessment according to the target policy special about phrase set to obtain a target semantic relevance assessment result; performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result; and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

An embodiment of the present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for evaluating policy specifications based on artificial intelligence, including the steps of: acquiring a special offer evaluation request corresponding to a target policy, wherein the special offer evaluation request carries a policy offer to be evaluated; word segmentation is carried out on the to-be-evaluated policy holder to obtain a target policy holder phrase set; finding out the phrase with grammar error from the target policy special about phrase set to obtain a target grammar error phrase set; carrying out semantic relevance assessment according to the target policy special about phrase set to obtain a target semantic relevance assessment result; performing risk assessment on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk assessment result; and generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result.

According to the method for evaluating the security appointment based on the artificial intelligence, firstly, the target security appointment phrase set is obtained by word segmentation of the security appointment to be evaluated, secondly, the grammar error phrases are found out from the target security appointment phrase set to obtain the target grammar error phrase set, the semantic relevance evaluation is carried out according to the target security appointment phrase set to obtain the target semantic relevance evaluation result, the risk evaluation is carried out on the target security appointment phrase set according to the obtained security appointment semantic library to obtain the target risk evaluation result, and finally, the special appointment evaluation result corresponding to the target security appointment is generated according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result, so that automatic grammar error evaluation, semantic relevance evaluation and risk evaluation are carried out on a single person, requirements on the security specialty level and knowledge of the special appointment specification of the single person are reduced, the accuracy of the security appointment is improved, and the security appointment is prevented from being invalid or potential risk is avoided; and moreover, manual inspection of single person is avoided, and the efficiency of single person is improved.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium provided by the present application and used in embodiments may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, apparatus, article or method that comprises the element.

The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the scope of the application, and all equivalent structures or equivalent processes using the descriptions and drawings of the present application or directly or indirectly applied to other related technical fields are included in the scope of the application.

Claims

1. An artificial intelligence based method for evaluating a policy offer, the method comprising:

acquiring a policy and appointment semantic library from a database or acquiring the policy and appointment semantic library from a third-party application system;

generating a special offer evaluation result corresponding to the target policy according to the target grammar error phrase set, the target semantic relevance evaluation result and the target risk evaluation result;

the step of performing risk assessment on the target policy holder phrase set according to the obtained policy holder semantic library to obtain a target risk assessment result comprises the following steps:

2. The method for evaluating an artificial intelligence based policy offer according to claim 1, wherein the step of obtaining an offer evaluation request of a target policy comprises:

3. The method for evaluating an artificial intelligence based policy about according to claim 1, wherein before the step of performing risk evaluation on the target policy about phrase set according to the obtained policy about semantic library to obtain a target risk evaluation result, the method further comprises:

acquiring a policy special offer semantic library update request, wherein the policy special offer semantic library update request carries a policy identification set to be analyzed;

4. The method for evaluating an artificial intelligence-based policy appointment according to claim 3, wherein the step of calculating a risk weight score for each phrase in the policy appointment phrase set to be analyzed according to the complaint information field and the claim information field corresponding to the policy appointment phrase set to be analyzed to obtain a risk weight diversity comprises the following steps:

5. The method for evaluating an artificial intelligence based policy offer according to claim 1, further comprising, after the step of generating an offer evaluation result corresponding to the target policy from the target grammar error phrase set, the target semantic relevance evaluation result, and the target risk evaluation result:

Acquiring a special appointment auditing request of the target policy;

6. The method for evaluating an artificial intelligence-based policy appointment according to claim 5, wherein the step of performing risk auditing on each phrase in the set of to-be-audited policy appointment phrases according to the policy appointment semantic library to obtain a risk auditing result comprises the steps of:

7. An artificial intelligence based policy offer assessment apparatus for implementing the method of any one of claims 1 to 6, the apparatus comprising:

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.

9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.