[HU, Yifei 胡艺飞, XU, Guoqiang 徐国强, QIU, Han 邱寒] Qianhai Complex A201, Qianwan Road 1 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518000Qianhai Shenzhen-Hong Kong Cooperation ZoneShenzhen, Guangdong 518000;1 Raffles Place, #36-01, 新加坡新加坡市1号莱佛士坊36楼01室莱佛士坊第一号, Singapore 048616One Raffles Place, Singapore 048616 Provided in the present application is a risk prediction method and device, a storage medium, and a server, comprising: acquiring a video image of an applicant during an interview; clustering the video frames in the video image, and determining a key frame video image; on the basis of the determined key frame video image and a micro-expression fraud probability model, determining the fraud probability corresponding to the micro-expressions in the key frame video image; on the basis of said fraud probability, performing risk prediction regarding the repayment capacity of the applicant. By means of analyzing micro-expressions of an applicant during an interview, the present invention estimates a probability of fraud by the applicant, and according to said probability, performs risk prediction regarding repayment capacity of the applicant, providing an objective additional judgment to estimate in advance the repayment capacity of the applicant, thus improving the accuracy and efficiency of risk prediction. Provided in the present application is a risk prediction method and device, a storage medium, and a server, comprising: acquiring a video image of an applicant during an interview; clustering the video frames in the video image, and determining a key frame video image; on the basis of the determined key frame video image and a micro-expression fraud probability model, determining the fraud probability corresponding to the micro-expressions in the key frame video image; on the basis of said fraud probability, performing risk prediction regarding the repayment capacity of the applicant. By means of analyzing micro-expressions of an applicant during an interview, the present invention estimates a probability of fraud by the applicant, and according to said probability, performs risk prediction regarding repayment capacity of the applicant, providing an objective additional judgment to estimate in advance the repayment capacity of the applicant, thus improving the accuracy and efficiency of

[MANDERS, Corey Mason 满康瑞] Qianhai Complex A201, NO. 1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation Zone 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052Shenzhen, Guangdong 518052;80 Pasir Panjang Road, Mapletree Business City #26-81/82 117372 新加坡枫树商业园区26楼81/82单元巴西班让路80号, Singapore 117372 A pupil radius detection method, comprising: obtaining an image to be measured of the pupil radius, and extracting a region-of-interest image in said image of the pupil radius (S202); performing pixel processing on the region-of-interest image to obtain a pupil region image (S204); extracting a plurality of pupil contours from the pupil region image by means of a boundary tracking algorithm (S206); obtaining the center of the region-of-interest image, a minimum bounding circle set of the plurality of pupil contours and the circle center of each minimum bounding circle in the minimum bounding circle set (S208); and obtaining the distance between the center and each circle center, screening the circle center closest to the center, taking the minimum surrounding circle corresponding to the closest circle center as a target circle, and taking the radius of the target circle as the pupil radius (S210). A pupil radius detection method, comprising: obtaining an image to be measured of the pupil radius, and extracting a region-of-interest image in said image of the pupil radius (S202); performing pixel processing on the region-of-interest image to obtain a pupil region image (S204); extracting a plurality of pupil contours from the pupil region image by means of a boundary tracking algorithm (S206); obtaining the center of the region-of-interest image, a minimum bounding circle set of the plurality of pupil contours and the circle center of each minimum bounding circle in the minimum bounding circle set (S208); and obtaining the distance between the center and each circle center, screening the circle center closest to the center, taking the minimum surrounding circle corresponding to the closest circle center as a target circle, and taking the radius of the target circle as the pupil radius (S210).

[HUANG, Haijie 黄海杰] Qianhai Complex A201, No.1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation Zone 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052Shenzhen, Guangdong 518052;80 Pasir Panjang Road, Mapletree Business City #26-81/82 新加坡枫树商业园区26楼81/82单元巴西班让路80号, Singapore 117372117372 A method for configuring a chat robot, comprising: acquiring a scanned image of a service table; extracting table feature information in the scanned image of the service table; determining field types of text blocks in the table feature information, and identifying target text blocks that need to perform field filling according to the field types; according to said target text blocks, establishing an association relationship between the text blocks; querying a preset data type configuration table according to the target text blocks; determining a data type of a field to be filled corresponding to each target text block; according to the data type of the field corresponding to each target text block, the association relationship between the text blocks, and a preset sentence template, generating a service segment sentence of each target text block; and according to position information of each target text block in the table feature information and the service segment sentence of each target text block, configuring a chat robot. A method for configuring a chat robot, comprising: acquiring a scanned image of a service table; extracting table feature information in the scanned image of the service table; determining field types of text blocks in the table feature information, and identifying target text blocks that need to perform field filling according to the field types; according to said target text blocks, establishing an association relationship between the text blocks; querying a preset data type configuration table according to the target text blocks; determining a data type of a field to be filled corresponding to each target text block; according to the data type of the field corresponding to each target text block, the association relationship between the text blocks, and a preset sentence template, generating a service segment sentence of each target text block; and according to position information of each target text block in the table feature information and the service segment sentence of each target text block, configuring a chat robot.

[JI, Bo 计波, ZOU, Mengfan 邹檬帆, TAO, Yue 陶玥, TANG, Lijun 唐利军] Qianhai Complex A201, Qianwan Road 1 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518000Qianhai Shenzhen-Hong Kong Cooperation ZoneShenzhen, Guangdong 518000;Mapletree Business City #26-81/82 新加坡巴西班让路80号枫树商业园区26楼81/82单元, Singapore 11737280 Pasir Panjang Road, 117372 Disclosed are a login token generation method and apparatus, a login token verification method and apparatus, and a server. The login token generation method comprises: receiving a login request submitted by a user by means of a browser, and verifying a user ID and a password included in the login request; after the verification is successful, obtaining browser information and user information; splicing the user ID, the user information, and the browser information into a key plaintext according to pre-defined splicing mode and splicing sequence; generating an original data plaintext on the basis of the user ID and a preset token expiration time, and encrypting the original data plaintext using a preset first encryption algorithm to obtain original data; encrypting pre-generated header, the original data, and the key plaintext using a preset second encryption algorithm to obtain a dynamic key; and combining the header, the original data, and the dynamic key into a login token. Thus, the security of a login token and a system is improved.

[HUANG, Haijie 黄海杰] Qianhai Complex A201, NO.1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation Zone 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052Shenzhen, Guangdong 518052;80 Pasir Panjang Road, Mapletree Business City #26-81/82 117372 新加坡枫树商业园区26楼81/82单元巴西班让路80号, Singapore 117372 A data processing method comprises: obtaining first speech emotion data according to a subtle speech feature; converting audio data of an interviewee into text data, dividing the text data into multiple sentences, performing, for respective words in the respective sentences, look-up and matching operations with respect to a preset dictionary corresponding to a trained emotion categorization network, determining confidence values of the text data belonging to respective preset emotion categories according to look-up and matching results, and obtaining second speech emotion data; inputting the text data into a trained syntactic analysis network, and obtaining a syntactic score of the text data; obtaining a confidence value of video data according to a subtle expression feature; and determining an interview result of the interviewee according to the first speech emotion data, the second speech emotion data, the syntactic score and the confidence value of video data. A data processing method comprises: obtaining first speech emotion data according to a subtle speech feature; converting audio data of an interviewee into text data, dividing the text data into multiple sentences, performing, for respective words in the respective sentences, look-up and matching operations with respect to a preset dictionary corresponding to a trained emotion categorization network, determining confidence values of the text data belonging to respective preset emotion categories according to look-up and matching results, and obtaining second speech emotion data; inputting the text data into a trained syntactic analysis network, and obtaining a syntactic score of the text data; obtaining a confidence value of video data according to a subtle expression feature; and determining an interview result of the interviewee according to the first speech emotion data, the second speech emotion data, the syntactic score and the confidence value of video data.

[GUO, Yiming 郭奕明] Qianhai Complex A201, No. 1 Qianwan 1st Road 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052Qianhai Shenzhen-Hong Kong Cooperation ZoneShenzhen, Guangdong 518052 The present application relates to a data decision-making-based test data generation method and apparatus, a computer device and a storage medium for intelligent decision-making in artificial intelligence technology. The method comprises: acquiring item configuration data sent by a terminal, the item configuration data comprising an item type; analyzing the item configuration data to obtain index data to be tested of an item to be tested; performing decision-making processing according to the item type and the index data to obtain a test strategy of the item; calling a trained data integration model according to the item type, inputting item requirement data and the index data into the data integration model, and outputting test data generated according to the test strategy; and encoding the test data to obtain an encoded test data set, storing the test data into a test database, and testing the item by using the test data set. The use of the present method is able to efficiently generate more comprehensive test data that is more highly available. The present application relates to a data decision-making-based test data generation method and apparatus, a computer device and a storage medium for intelligent decision-making in artificial intelligence technology. The method comprises: acquiring item configuration data sent by a terminal, the item configuration data comprising an item type; analyzing the item configuration data to obtain index data to be tested of an item to be tested; performing decision-making processing according to the item type and the index data to obtain a test strategy of the item; calling a trained data integration model according to the item type, inputting item requirement data and the index data into the data integration model, and outputting test data generated according to the test strategy; and encoding the test data to obtain an encoded test data set, storing the test data into a test database, and testing the item by using the test data set. The use of the present method is able to efficiently generate more comprehensive test data that is more highly available.

[WANG, Xiao 王晓] Qianhai Complex A201, No.1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation Zone 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052Shenzhen, Guangdong 518052 A deduction information processing method, comprising: acquiring a data transfer request, and acquiring data transfer object information corresponding to the data transfer request; acquiring, according to the data transfer object information, a data transfer mode corresponding to the data transfer object information and a deduction model corresponding to each data transfer mode; acquiring an input transfer numerical value, and using the transfer numerical value as an input of each reduction model to obtain, by means of calculation, a deduction numerical value corresponding to each data transfer mode; and sorting and displaying corresponding data transfer modes according to the size of the deduction numerical values. The deduction information processing method facilitates improving the efficiency of data transfer and saves resources to the greatest extent. A deduction information processing method, comprising: acquiring a data transfer request, and acquiring data transfer object information corresponding to the data transfer request; acquiring, according to the data transfer object information, a data transfer mode corresponding to the data transfer object information and a deduction model corresponding to each data transfer mode; acquiring an input transfer numerical value, and using the transfer numerical value as an input of each reduction model to obtain, by means of calculation, a deduction numerical value corresponding to each data transfer mode; and sorting and displaying corresponding data transfer modes according to the size of the deduction numerical values. The deduction information processing method facilitates improving the efficiency of data transfer and saves resources to the greatest extent.

[DING, Jingjing 丁晶晶] Qianhai Complex A201 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518000No.1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation ZoneShenzhen, Guangdong 518000 Disclosed are a license plate verification method and apparatus, and a computer device and a storage medium. Different deviation comparison modes are determined by means of the result of comparing the length of a target segmentation character string with the length of a segmentation character string to be verified; a first target comparison position and a second target comparison position are updated according to the deviation comparison mode by comparing the character of the target segmentation character string with the character of the segmentation character string to be verified; when the first target comparison position is the last position, the number of character verification errors is adjusted according to the second comparison result of the last position. By comparing the preset number of allowable character verification errors with the number of character verification errors, the influence of errors occurring in the license plate recognition process on license plate verification is well avoided, so that the pass rate and accuracy of license plate recognition are improved, and the misjudgment rate of license plate verification is reduced. Disclosed are a license plate verification method and apparatus, and a computer device and a storage medium. Different deviation comparison modes are determined by means of the result of comparing the length of a target segmentation character string with the length of a segmentation character string to be verified; a first target comparison position and a second target comparison position are updated according to the deviation comparison mode by comparing the character of the target segmentation character string with the character of the segmentation character string to be verified; when the first target comparison position is the last position, the number of character verification errors is adjusted according to the second comparison result of the last position. By comparing the preset number of allowable character verification errors with the number of character verification errors, the influence of errors occurring in the license plate recognition process on license plate verification is well avoided, so that the pass rate and accuracy of license plate recognition are improved, and the misjudgment rate of license plate verification is reduced.

[LU, Ning 卢宁, XU, Guoqiang 徐国强] Qianhai Complex A201, No.1 Qianwan 1st Road, Qianhai Shenzhen-Hong Kong Cooperation Zone 中国广东省深圳市前海深港合作区前湾一路1号A栋201室（入驻深圳市前海商务秘书有限公司）, Guangdong 518000Shenzhen, Guangdong 518000 A video-based eyeball turning determination method and system. The method comprises: acquiring a target video, wherein the target video is a video of a target user viewing a target product (S100); inputting the target video into an eyeball turning feature recognition model to obtain an eyeball turning feature queue matrix; and on the basis of the eyeball turning feature queue matrix, determining a target turning angle of the target user. According to the method, the eyeball turning feature queue matrix is acquired by means of inputting the target video into the eyeball turning feature recognition model, and the target turning angle and target turning time of a corresponding target product are then acquired by means of the eyeball turning feature queue matrix, thereby improving the precision of eyeball tracking. A video-based eyeball turning determination method and system. The method comprises: acquiring a target video, wherein the target video is a video of a target user viewing a target product (S100); inputting the target video into an eyeball turning feature recognition model to obtain an eyeball turning feature queue matrix; and on the basis of the eyeball turning feature queue matrix, determining a target turning angle of the target user. According to the method, the eyeball turning feature queue matrix is acquired by means of inputting the target video into the eyeball turning feature recognition model, and the target turning angle and target turning time of a corresponding target product are then acquired by means of the eyeball turning feature queue matrix, thereby improving the precision of eyeball tracking.

[XU, Guoping 徐国平] Qianhai Complex A201, 中国广东省深圳市前海深港合作区前湾一路1号A栋201室, Guangdong 518052No. 1 Qianwan 1st RoadQianhai Shenzhen-Hong Kong Cooperation Zone,Shenzhen, Guangdong 518052 The present application relates to the field of image recognition, and provides a face recognition-based vehicle charging method, a related device, and a storage medium. The method comprises: receiving a binding request of a user, acquiring facial information of the user using the 3D structured light technology; if the facial information is checked to match a driver's license photo in driver's license information, checking whether real-name authentication information is consistent with the identity information in the driver's license information; if yes, checking whether the license plate number is associated with the facial information, and if so, determining that the authentication is passed; establishing a binding relationship between the license plate number and a payment account; calling a camera to acquire and recognize the license plate number of a vehicle to be charged, querying the payment account bound to the license plate number according to the binding relationship, and calculating the charge amount for said vehicle; generating a first prompt information and displaying same and the charge amount; after the facial identity verification on the facial information acquired by the camera is passed, sending a first payment request to a payment service provider. The present solution can improve payment efficiency. The present application relates to the field of image recognition, and provides a face recognition-based vehicle charging method, a related device, and a storage medium. The method comprises: receiving a binding request of a user, acquiring facial information of the user using the 3D structured light technology; if the facial information is checked to match a driver's license photo in driver's license information, checking whether real-name authentication information is consistent with the identity information in the driver's license information; if yes, checking whether the license plate number is associated with the facial information, and if so, determining that the authentication is passed; establishing a binding relationship between the license plate number and a payment account; calling a camera to acquire and recognize the license plate number of a vehicle to be charged, querying the payment account bound to the license plate number according to the binding relationship, and calculating the charge amount for said vehicle; generating a first prompt information and displaying same and the charge amount; after the facial identity verification on the facial information acquired by the camera is passed, sending a first payment request to a payment service provider. The present solution can improve payment efficiency.