CN115119017B - Medical health insurance recommendation and management method and system based on artificial intelligence - Google Patents
Medical health insurance recommendation and management method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN115119017B CN115119017B CN202210533377.8A CN202210533377A CN115119017B CN 115119017 B CN115119017 B CN 115119017B CN 202210533377 A CN202210533377 A CN 202210533377A CN 115119017 B CN115119017 B CN 115119017B
- Authority
- CN
- China
- Prior art keywords
- barrage
- time
- user
- history
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000007726 management method Methods 0.000 title claims abstract description 18
- 230000036541 health Effects 0.000 title claims abstract description 16
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 43
- 238000000034 method Methods 0.000 claims description 24
- 230000000875 corresponding effect Effects 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 5
- 230000002596 correlated effect Effects 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010295 mobile communication Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/235—Processing of additional data, e.g. scrambling of additional data or processing content descriptors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/21—Server components or server architectures
- H04N21/218—Source of audio or video content, e.g. local disk arrays
- H04N21/2187—Live feed
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/431—Generation of visual interfaces for content selection or interaction; Content or additional data rendering
- H04N21/4312—Generation of visual interfaces for content selection or interaction; Content or additional data rendering involving specific graphical features, e.g. screen layout, special fonts or colors, blinking icons, highlights or animations
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/47—End-user applications
- H04N21/478—Supplemental services, e.g. displaying phone caller identification, shopping application
- H04N21/47815—Electronic shopping
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Acoustics & Sound (AREA)
- Technology Law (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Human Computer Interaction (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The embodiment of the application provides a medical health insurance recommendation and management method and system based on artificial intelligence, aiming at improving the recommendation efficiency of insurance products in a live broadcast room. The recommendation and management method is applied to the live broadcast server and comprises the following steps: acquiring real-time voice data of a host in an insurance recommendation live broadcast room, and performing voice recognition on the real-time voice data to acquire a voice recognition result; matching the voice recognition result with a plurality of real-time barrages, and matching a target real-time barrages corresponding to the voice recognition result; comparing the target real-time barrage with a plurality of historical barrages carrying offline identification, and determining a target historical barrage similar to the target real-time barrage; the offline identification is used for representing that a user who issues the historical barrage currently leaves the insurance recommendation live broadcast room; and pushing the target real-time barrage and the voice recognition result to a user who publishes the target history barrage.
Description
Technical Field
The application relates to the field of insurance recommendation, in particular to a medical health insurance recommendation and management method and system based on artificial intelligence.
Background
With the development of internet technology and mobile communication technology, commodity sales through a live broadcasting room becomes an important marketing mode. Taking insurance selling as an example, a live player views a live barrage sent by a user at a live broadcast room and recommends insurance products for the user in a personalized manner according to the user conditions reflected in the barrage. However, in many cases, the online users and the backdrop of the live broadcasting room are large, and the number of the anchor is only one or two, so that the anchor is difficult to reply to each backdrop, and therefore, many users are difficult to obtain the recommendation information of the insurance product by sending the backdrop.
Disclosure of Invention
The embodiment of the application aims to provide a medical health insurance recommending and managing method and system based on artificial intelligence, aiming at improving the recommending efficiency of insurance products in a live broadcasting room, and the specific technical scheme is as follows:
in a first aspect of an embodiment of the present application, there is provided an artificial intelligence based medical health insurance recommendation and management method, the method being applied to a live server, the method comprising:
acquiring real-time voice data of a host in an insurance recommendation live broadcast room, and performing voice recognition on the real-time voice data to acquire a voice recognition result;
matching the voice recognition result with a plurality of real-time barrages, and matching a target real-time barrages corresponding to the voice recognition result;
comparing the target real-time barrage with a plurality of historical barrages carrying offline identification, and determining a target historical barrage similar to the target real-time barrage; the offline identification is used for representing that a user who issues the historical barrage currently leaves the insurance recommendation live broadcast room;
and pushing the target real-time barrage and the voice recognition result to a user who publishes the target history barrage.
In a second aspect of the embodiment of the present application, there is provided an artificial intelligence-based medical health insurance recommendation and management system, the system being applied to a live server, the system comprising:
the voice recognition module is used for acquiring real-time voice data of a host player in the insurance recommendation live broadcast room, and carrying out voice recognition on the real-time voice data to acquire a voice recognition result;
the bullet screen matching module is used for matching the voice recognition result with a plurality of real-time bullet screens and matching a target real-time bullet screen corresponding to the voice recognition result;
the bullet screen comparison module is used for comparing the target real-time bullet screen with a plurality of historical bullet screens carrying offline marks and determining target historical bullet screens similar to the target real-time bullet screen; the offline identification is used for representing that a user who issues the historical barrage currently leaves the insurance recommendation live broadcast room;
and the result pushing module is used for pushing the target real-time barrage and the voice recognition result to a user who publishes the target history barrage.
According to the application, the live broadcast server records the historical barrage released by the user who leaves the live broadcast room, after the voice recognition result of the real-time voice data of the host player is obtained, the target real-time barrage corresponding to the voice recognition result is matched, and the historical barrage similar to the target real-time barrage is determined, so that the voice recognition result can be used as the reply of the historical barrage, and finally the voice recognition result is pushed to the user who releases the historical barrage, so that the user can obtain the reply of the live player aiming at the similar barrage after leaving the insurance recommendation live broadcast room, and the insurance product recommendation efficiency of the live broadcast room can be effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It is evident that the drawings in the following description are only some embodiments of the present application and that other drawings may be obtained from these drawings by those of ordinary skill in the art without inventive effort.
FIG. 1 is a flow chart of a medical health insurance recommendation and management method according to an embodiment of the application;
fig. 2 is a schematic structural diagram of a medical health insurance recommendation and management system according to an embodiment of the application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It will be apparent that the described embodiments are some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
With the development of internet technology and mobile communication technology, commodity sales through a live broadcasting room becomes an important marketing mode. Taking insurance selling as an example, a live player views a live barrage sent by a user at a live broadcast room and recommends insurance products for the user in a personalized manner according to the user conditions reflected in the barrage. However, in many cases, the online users and the backdrop of the live broadcasting room are large, and the number of the anchor is only one or two, so that the anchor is difficult to reply to each backdrop, and therefore, many users are difficult to obtain the recommendation information of the insurance product by sending the backdrop.
The application provides a medical health insurance recommending and managing method and system based on artificial intelligence, which improves the recommending efficiency of insurance products in a living broadcast room. Referring to fig. 1, fig. 1 is a flowchart of a medical health insurance recommendation and management method according to an embodiment of the application, and the method is applied to a live server. As shown in fig. 1, the method comprises the steps of:
s110: and acquiring real-time voice data of a host in the insurance recommendation live broadcasting room, and performing voice recognition on the real-time voice data to acquire a voice recognition result.
Wherein, the real-time voice data refers to the current voice data of the anchor. After the current voice data is obtained, voice recognition can be carried out on the voice data through the existing voice recognition technology, so that a semantic recognition result is obtained. In the present application, the speech recognition result may be in the form of text.
S120: and matching the voice recognition result with a plurality of real-time barrages, and matching a target real-time barrages corresponding to the voice recognition result.
The real-time barrage refers to the barrage displayed on a screen of a user terminal (namely a live APP page of a spectator). When the voice recognition result is matched with the real-time barrage, whether the voice recognition result contains the user name of the issuing user of the real-time barrage or not or whether the voice recognition result contains the real-time barrage or not can be detected, and if the voice recognition result contains the user name of the issuing user of the real-time barrage, the voice recognition result is matched with the real-time barrage. For ease of understanding, it is assumed by way of example that there are 10 tiles currently displayed on the user terminal screen, with the 6 th tile being "light cloud: please ask what insurance a 6 year old child is fit to buy? ". In the bullet screen, the 'light and cloudy wind' is the user name of the issuing user, and the 'asking what insurance the 6-year-old child is suitable for buying' is the real-time bullet screen. If the voice recognition result is' Miss of wind you good, children aged 6 can buy the accident risk of children, and also can buy the medical insurance of children. And the 6 th barrage is determined to be the target real-time barrage by matching with the 10 real-time barrages one by one, and the user name of the issuing user of the 6 th barrage is contained in the voice recognition result.
S130: comparing the target real-time barrage with a plurality of historical barrages carrying offline identification, and determining a target historical barrage similar to the target real-time barrage; wherein the offline identification is used to characterize that the user who issued the historical barrage has currently left the insurance recommendation direct broadcast room.
In some embodiments, when comparing the target real-time bullet screen with a plurality of history bullet screens carrying offline identification, the following substeps may be specifically performed:
s130-1: based on the word vector generation model, generating the word vector of the target real-time barrage.
S130-2: and comparing the word vector of the target real-time barrage with the word vectors of the plurality of history barrages carrying the offline identification, and determining the similarity of the word vector of the target real-time barrage and the word vector of each history barrage. Wherein the word vector of each history barrage is generated based on a word vector generation model.
S130-3: and determining the historical barrages with the similarity larger than the first preset threshold value as target real-time barrages.
In some embodiments, the target real-time barrage may be input into an existing word vector generation model (e.g., word2 vec) to obtain a word vector output by the word vector generation model.
In some embodiments, when the word vector of the target real-time barrage is compared with the word vector of each history barrage, the euclidean distance between the word vector of the target real-time barrage and the word vector of the history barrage can be calculated, then normalization processing is performed on the euclidean distance to obtain a normalization result which is greater than or equal to 0 and less than or equal to 1, and finally the normalization result is subtracted by 1 to obtain the similarity of the word vector of the target real-time barrage and the word vector of the target real-time barrage.
According to the method, the similarity between the target real-time barrage and the target real-time barrage is calculated, and the historical barrage with the similarity larger than the first preset threshold value is used as the target historical barrage. Thus, a history barrage very similar to the target real-time barrage, namely the target history barrage, can be determined from a plurality of history barrages. Since the target history barrage is very similar to the target real-time barrage, the semantics of the target history barrage is very similar to those of the target real-time barrage, and therefore the problems concerned by the publisher of the target history barrage are very similar to those of the target real-time barrage.
S140: and pushing the target real-time barrage and the voice recognition result to a user who issues the target historical barrage.
In some embodiments, the target real-time barrage and voice recognition results may be pushed to users who have left the insurance recommendation live room by way of client system messages, or by way of stranger messages. It should be noted that, the specific sending mode of the target real-time barrage and the voice recognition result is not limited by the present application.
As described above, the problem of the target history barrage publisher is likely to be similar to the problem of the target real-time barrage publisher, so that the user can obtain the content of interest from the target real-time barrage and the voice recognition result by pushing the target real-time barrage and the voice recognition result to the user who publishes the target history barrage, thereby being beneficial to improving the insurance product recommendation efficiency of the live broadcasting room.
In some embodiments, each user corresponds to a different first preset threshold. The method can update the first preset threshold according to the leaving time of each user leaving the insurance recommended live broadcasting room and the number of times that each user is pushed with the voice recognition result. The size of the first preset threshold is positively correlated with the leaving time and the pushed times. In other words, the longer the user leaves the insurance recommendation live room, and the more times the user is pushed with speech recognition results, the greater the first preset threshold.
In the application, as the user leaves the insurance recommendation live room longer and longer, the first preset threshold value is gradually increased. Thus, only when the target real-time barrage is particularly similar to the user's history barrage so that the similarity is greater than a first preset threshold value, the history barrage is determined to be the target history barrage, and the voice recognition result is pushed to the user.
For ease of understanding, assuming that the first preset threshold is equal to 0.75 at 15 minutes after the user leaves the insurance recommended live room, if the similarity between a target real-time bullet screen and the user's history bullet screen is equal to 0.79, the history bullet screen is determined to be the target history bullet screen and the voice recognition result is pushed to the user because the similarity is greater than the first preset threshold of 0.75. Further, assuming that the first preset threshold value is equal to 0.85 at 24 minutes after the user leaves the insurance recommendation live broadcast room, if the similarity between a target real-time barrage and the user's history barrage is equal to 0.81, the history barrage is not determined to be the target history barrage and the voice recognition result is not pushed to the user because the similarity is smaller than the first preset threshold value of 0.85 at the moment.
It should be noted that, in the present application, as the user leaves the insurance recommendation live broadcast room, the user is likely to be more and more inconvenienced by the question asked in the live broadcast room, so as to gradually increase the first preset threshold over time, so that the difficulty of pushing the voice recognition result to the user is more and more increased, except for the situation that the target real-time barrage is particularly similar to the history barrage.
In the application, as the number of times that the user is pushed with the voice recognition result is increased, the first preset threshold value is gradually increased. Thus, only when the target real-time barrage is particularly similar to the user's history barrage so that the similarity is greater than a first preset threshold value, the history barrage is determined to be the target history barrage, and the voice recognition result is pushed to the user.
For ease of understanding, it is assumed that the first preset threshold is equal to 0.70 when the user has not been pushed the voice recognition result yet, and if the similarity between a target real-time bullet screen and the user's history bullet screen is equal to 0.72, the history bullet screen is determined to be the target history bullet screen and the voice recognition result is pushed to the user because the similarity is greater than the first preset threshold of 0.70. Further, if the similarity between a target real-time bullet screen and a user's history bullet screen is equal to 0.76, the similarity is smaller than the first preset threshold value 0.80, so that the history bullet screen is not determined to be the target history bullet screen, and the voice recognition result is not pushed to the user.
It should be noted that, in the present application, as the number of times that the user is pushed with the voice recognition result increases, the user only wants to receive the anchor reply (i.e. the voice recognition result) that is more relevant to the history barrage later. Therefore, as the number of times of being pushed increases, the first preset threshold value is gradually increased, so that the difficulty of pushing the voice recognition result to the user is increased, except for the situation that the target real-time barrage is particularly similar to the history barrage.
In some embodiments, the first preset threshold is calculated as follows:
wherein T represents a first preset threshold; sigmoid () is a normalization function; t represents the time of the user leaving the insurance recommendation live broadcast room, and the unit is hours; x represents the number of times the user has been pushed with speech recognition results.
In some embodiments, the live server records the time each user leaves the live room, updates the number of times each user is pushed, and updates the first preset threshold value for each user at fixed intervals (e.g., every 5 minutes).
In some embodiments, the live server may also receive a close request sent by the anchor to close the insurance recommendation live room; and the live broadcast server responds to the closing request, closes the insurance recommendation live broadcast room, judges whether each history barrage carries a subscription identifier according to each history barrage, deletes the history barrages which do not carry the subscription identifier, and stores the history barrages carrying the subscription identifier. In addition, the live broadcast server can also receive an opening request of an opening insurance recommendation live broadcast room sent by a main broadcaster; and responding to the starting request, starting the insurance recommendation live broadcasting room, and calling the history barrage carrying the subscription identification.
In the application, when the anchor closes the insurance recommendation live broadcasting room, the history barrage can be deleted by the live broadcasting server. And the history barrages carrying the subscription identifiers are stored by the live broadcast server, and are called out when the host broadcast starts the insurance recommendation live broadcast room next time, so that the user can continue to be pushed with the voice recognition result during the next live broadcast.
In some embodiments, the live broadcast server may further receive a barrage published by the user, and determine whether the barrage carries a subscription identifier, where the subscription identifier is given to the transmitted barrage by the terminal after the terminal detects that a subscription control of the barrage transmitting page is selected; and under the condition that the barrage carries the subscription identification, the live broadcast server stores the barrage as a history barrage and stores the subscription identification of the barrage.
In specific implementation, the barrage sending page includes: the system comprises a barrage composition window, a subscription control positioned beside the barrage composition window, and a sending control positioned beside the barrage composition window. After the user inputs the barrage text in the barrage writing window, if the user wishes to subscribe to the barrage, the user can click (i.e. select) the subscription control first and then click the sending control, so as to send the barrage carrying the subscription identification. If the user does not wish to subscribe to the bullet screen, the user can click on the send control directly, so that the bullet screen which does not carry the subscription identification is sent.
In some embodiments, the live broadcast server may further receive a subscription revocation request sent by the user, where the subscription revocation request is sent by the terminal after the terminal detects that a revocation control of a corresponding barrage in the subscription management page is selected; and in response to the subscription revocation request, the live broadcast server deletes the subscription identification of the corresponding history barrage.
In particular implementations, the subscription management page displays a plurality of subscribed backlashes, and one side of each backlashes displays a revocation control. If the user wants to cancel the subscription of a barrage, clicking (i.e. selecting) a cancel control displayed on one side of the barrage, and after detecting the user operation, sending a subscription cancel request to the live broadcast server, wherein the subscription cancel request carries the barrage identifier of the corresponding barrage. And indexing the corresponding barrage by the live broadcast server according to the barrage identification carried by the subscription revocation request, thereby deleting the barrage.
In some embodiments, during live broadcast, the live broadcast server further performs the steps of: receiving a barrage issued by a user, and removing special characters in the barrage to obtain a residual character string; judging whether the character length of the residual character strings meets the preset length requirement; determining an insurance consultation attribute value of the residual character string based on a pre-trained insurance consultation attribute recognition model aiming at the residual character string meeting the preset length requirement; and if the insurance consultation attribute value of the residual character string is larger than or equal to a second preset threshold value, the residual character string is used as a history barrage and is stored, and if the insurance consultation attribute value of the residual character string is smaller than the second preset threshold value, the residual character string is discarded.
In the application, whether the length of the residual character string meets the preset length requirement is judged, if the length requirement is not met, the residual character string is discarded, and if the length requirement is met, the insurance consultation attribute value of the residual character string is continuously determined. Thus, the residual character strings with shorter character lengths can be filtered out quickly. Because the residual character strings with shorter lengths usually have no complete semantics or the insurance requirements are difficult to describe clearly, the number of invalid history barrages can be controlled by rapidly filtering the residual character strings with shorter character lengths, so that the barrage comparison amount in the step S130 is reduced, and the system operation efficiency is improved.
Similarly, by determining the insurance consultation attribute value, the remaining character strings with lower insurance consultation attribute values are discarded, and the remaining character strings with higher insurance consultation attribute values are reserved, so that the number of invalid historical barrages can be controlled, the barrage comparison amount in the step S130 is further reduced, and the system operation efficiency is improved.
In particular implementations, the insurance consultation attribute identification model may be trained by:
first, a plurality of bullet screen data are acquired as a plurality of sample data, including positive samples and negative samples. The forward sample is insurance consultation type barrages, such as "please ask 6 years old children which insurance can be purchased", "60 years old can also buy serious diseases" and the like. The negative sample is a non-insurance consultation type barrage, such as "the anchor is very bar", "the thank the anchor", "when the next live broadcast is, etc. The insurance consultation attribute value carried by each positive sample is 1, and the insurance consultation attribute value carried by each negative sample is 0.
And generating word vectors based on word2vec aiming at each sample, and inputting the word vectors into a convolutional neural network CNN to obtain the insurance consultation attribute values predicted by the CNN. And then calculating a difference value between the predicted insurance consultation attribute value and the insurance consultation attribute value carried by the sample, taking the absolute value of the difference value as a loss value loss, and finally updating CNN and word2vec based on the loss value loss.
In some embodiments, after the live server matches the target real-time bullet screen in step S120, on the one hand, the target real-time bullet screen may be compared with a plurality of historical bullet screens carrying offline identification by executing step S130. On the other hand, the target real-time barrage can be compared with a plurality of history barrages which do not carry offline identification, the similarity of each history barrage and the target real-time barrage is determined, if the similarity of the history barrage and the target real-time barrage is larger than or equal to a third preset threshold value, the history barrage is deleted, and if the similarity of the history barrage and the target real-time barrage is smaller than the third preset threshold value, the history barrage is reserved.
In the application, if the history barrage does not carry the offline identification, the user who issues the history barrage is not leaving (i.e. exiting) the insurance recommendation live broadcast room. Therefore, if the similarity between the historical barrage and the target real-time barrage is greater than or equal to the third preset threshold value, the user is informed that the user has watched the reply of the target real-time barrage on line, and the reply is very relevant to the barrage issued before the user, so that the live broadcast server can directly delete the historical barrage.
In some embodiments, the live server may further receive an exit request sent by the user to exit the insurance recommendation live room; responding to the exit request, judging whether a history barrage corresponding to the user is stored; if the historical barrage corresponding to the user is stored, an offline identification is added for the historical barrage corresponding to the user.
In the application, after receiving the exit request sent by the user, if the live broadcast server stores the history barrage of the user, the live broadcast server indicates that the history barrage of the user is not replied by the host, and the host does not reply other barrages similar to the history barrage. In order to enable a user to still have an opportunity to receive a host response after exiting an insurance recommendation live broadcast room, the live broadcast server adds an offline identifier for a history barrage corresponding to the user. Thus, through steps S110 to S140 described above, the user has an opportunity to receive the anchor reply.
Above, the application discloses a medical health insurance recommendation and management method through the embodiment. In the following, the present application discloses a medical health insurance recommendation and management system by way of embodiments. In view of the above method and the following system based on the same inventive concept, a brief description of system embodiments is provided below in order to avoid repetition.
Referring to fig. 2, fig. 2 is a schematic structural diagram of a medical health insurance recommendation and management system according to an embodiment of the application, and the system is applied to a live server. As shown in fig. 2, the system comprises the steps of:
the voice recognition module 210 is configured to obtain real-time voice data of a presenter in the insurance recommendation live broadcast room, and perform voice recognition on the real-time voice data to obtain a voice recognition result.
The barrage matching module 220 is configured to match the voice recognition result with a plurality of real-time barrages, and match a target real-time barrage corresponding to the voice recognition result.
The barrage comparison module 230 is configured to compare the target real-time barrage with a plurality of historical barrages carrying offline identification, and determine a target historical barrage similar to the target real-time barrage; wherein the offline identification is used to characterize that the user who issued the historical barrage has currently left the insurance recommendation direct broadcast room.
The result pushing module 240 is configured to push the target real-time barrage and the voice recognition result to the user who publishes the target history barrage.
Optionally, in some embodiments, the barrage comparison module 230 is specifically configured to: generating a word vector of the target real-time barrage based on the word vector generation model; comparing the word vector of the target real-time barrage with the word vectors of a plurality of history barrages carrying offline identification, and determining the similarity of the word vector of the target real-time barrage and the word vector of each history barrage, wherein the word vector of each history barrage is generated based on a word vector generation model; and determining the historical barrages with the similarity larger than the first preset threshold value as target real-time barrages.
In the system shown in fig. 2, each user corresponds to a different first preset threshold. As shown in fig. 2, the system further comprises:
the threshold updating module 250 is configured to update the first preset threshold according to a departure time of each user from the insurance recommendation live broadcast room and a number of times each user is pushed with a voice recognition result, where a magnitude of the first preset threshold is positively related to a departure time and a number of times each user is pushed.
Optionally, in some embodiments, as shown in fig. 2, the system further comprises:
the character rejecting module 260 is configured to receive the barrage issued by the user, and reject special characters in the barrage to obtain the remaining character strings.
The character length determining module 270 is configured to determine whether the character length of the remaining character strings meets a preset length requirement.
The attribute value determining module 280 is configured to determine, for a remaining string that meets a preset length requirement, an insurance consultation attribute value of the remaining string based on a pre-trained insurance consultation attribute recognition model.
The barrage storage module 290 is configured to take the remaining character strings as a history barrage and store the remaining character strings if the insurance consultation attribute values of the remaining character strings are greater than or equal to a second preset threshold, and discard the remaining character strings if the insurance consultation attribute values of the remaining character strings are less than the second preset threshold.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus 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, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.
Claims (3)
1. A medical health insurance recommendation and management method based on artificial intelligence, the method being applied to a live server, the method comprising:
acquiring real-time voice data of a host in an insurance recommendation live broadcast room, and performing voice recognition on the real-time voice data to acquire a voice recognition result;
matching the voice recognition result with a plurality of real-time barrages, and matching a target real-time barrages corresponding to the voice recognition result;
comparing the target real-time barrage with a plurality of historical barrages carrying offline identification, and determining a target historical barrage similar to the target real-time barrage; the offline identification is used for representing that a user who issues the historical barrage currently leaves the insurance recommendation live broadcast room;
pushing the target real-time barrage and the voice recognition result to a user who publishes the target history barrage;
comparing the target real-time barrage with a plurality of historical barrages carrying offline identification to determine a target historical barrage similar to the target real-time barrage, comprising:
generating a word vector of the target real-time barrage based on a word vector generation model;
comparing the word vector of the target real-time barrage with the word vectors of the plurality of history barrages carrying the offline identification, and determining the similarity of the word vector of the target real-time barrage and the word vector of each history barrage; wherein the word vector of each history barrage is generated based on the word vector generation model;
determining a historical barrage with similarity larger than a first preset threshold value as a target real-time barrage;
the method further comprises the steps of:
each user corresponds to different first preset thresholds respectively, and the first preset thresholds are updated according to the leaving time of each user leaving the insurance recommendation live broadcast room and the times of pushing voice recognition results of each user; wherein the size of the first preset threshold is positively correlated with the leaving time and the pushed times;
the calculation formula of the first preset threshold is as follows:
wherein T represents a first preset threshold; />Is a normalization function; t represents the time of the user leaving the insurance recommendation live broadcast room, and the unit is hours; x represents the number of times the user has been pushed with speech recognition results;
the live broadcast server records the time when each user leaves a live broadcast room, updates the pushed times of each user, and updates a first preset threshold value of each user every fixed time;
receiving a closing request for closing the insurance recommendation live broadcasting room, which is sent by the anchor;
responding to the closing request, closing the insurance recommendation live broadcasting room, judging whether each history barrage carries a subscription identifier or not according to each history barrage, deleting the history barrages which do not carry the subscription identifier, and storing the history barrages carrying the subscription identifier;
the method further comprises the steps of:
receiving an opening request for opening the insurance recommendation living broadcast room, which is sent by the anchor;
responding to the starting request, starting the insurance recommendation live broadcasting room, and calling a history barrage carrying a subscription identifier;
receiving a barrage published by the user, and judging whether the barrage carries a subscription identifier, wherein the subscription identifier is given to the transmitted barrage by a terminal after the terminal detects that a subscription control of a barrage transmitting page is selected;
storing the barrage as a history barrage and storing the subscription identification of the barrage under the condition that the barrage carries the subscription identification;
the method further comprises the steps of:
receiving a subscription revocation request sent by the user, wherein the subscription revocation request is sent by the terminal after the terminal detects that a revocation control of a corresponding barrage in a subscription management page is selected;
deleting subscription identifiers of corresponding historical barrages in response to the subscription revocation requests;
receiving a barrage issued by the user, and removing special characters in the barrage to obtain a residual character string;
judging whether the character length of the residual character strings meets the preset length requirement or not;
determining an insurance consultation attribute value of the residual character string based on a pre-trained insurance consultation attribute recognition model aiming at the residual character string meeting the preset length requirement;
if the insurance consultation attribute value of the residual character string is larger than or equal to a second preset threshold value, the residual character string is used as a history barrage and is stored, and if the insurance consultation attribute value of the residual character string is smaller than the second preset threshold value, the residual character string is discarded;
comparing the target real-time barrage with a plurality of historical barrages which do not carry offline identification, and determining the similarity of each historical barrage and the target real-time barrage;
and deleting the historical barrage if the similarity between the historical barrage and the target real-time barrage is greater than or equal to a third preset threshold value, and retaining the historical barrage if the similarity between the historical barrage and the target real-time barrage is less than the third preset threshold value.
2. The method according to claim 1, wherein the method further comprises:
receiving an exit request which is sent by a user and exits the insurance recommendation live broadcasting room;
responding to the exit request, judging whether a history barrage corresponding to the user is stored or not;
and if the history barrage corresponding to the user is stored, adding the offline identification for the history barrage corresponding to the user.
3. An artificial intelligence based medical health insurance recommendation and management system, the system being applied to a live server, the system comprising:
the voice recognition module is used for acquiring real-time voice data of a host player in the insurance recommendation live broadcast room, and carrying out voice recognition on the real-time voice data to acquire a voice recognition result;
the bullet screen matching module is used for matching the voice recognition result with a plurality of real-time bullet screens and matching a target real-time bullet screen corresponding to the voice recognition result;
the bullet screen comparison module is used for comparing the target real-time bullet screen with a plurality of historical bullet screens carrying offline marks and determining target historical bullet screens similar to the target real-time bullet screen; the offline identification is used for representing that a user who issues the historical barrage currently leaves the insurance recommendation live broadcast room;
the result pushing module is used for pushing the target real-time barrage and the voice recognition result to a user who publishes the target history barrage;
the barrage comparison module is specifically used for: generating a word vector of the target real-time barrage based on a word vector generation model; comparing the word vector of the target real-time barrage with the word vectors of a plurality of history barrages carrying offline identification, and determining the similarity of the word vector of the target real-time barrage and the word vector of each history barrage, wherein the word vector of each history barrage is generated based on the word vector generation model; determining a historical barrage with similarity larger than a first preset threshold value as a target real-time barrage;
in the system, each user corresponds to a different first preset threshold value, and the system further comprises:
the threshold updating module is used for updating the first preset threshold according to the leaving time of each user leaving the insurance recommendation live broadcast room and the number of times that each user is pushed with a voice recognition result, wherein the size of the first preset threshold is positively correlated with the leaving time and the number of times that each user is pushed;
the system further comprises:
the character rejecting module is used for receiving the barrage issued by the user and rejecting special characters in the barrage to obtain residual character strings;
the character length judging module is used for judging whether the character length of the residual character strings meets the preset length requirement;
the attribute value determining module is used for determining the insurance consultation attribute value of the residual character string based on a pre-trained insurance consultation attribute identification model aiming at the residual character string meeting the preset length requirement;
and the barrage storage module is used for taking the residual character strings as historical barrages and storing the residual character strings if the insurance consultation attribute values of the residual character strings are larger than or equal to a second preset threshold value, and discarding the residual character strings if the insurance consultation attribute values of the residual character strings are smaller than the second preset threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210533377.8A CN115119017B (en) | 2022-05-17 | 2022-05-17 | Medical health insurance recommendation and management method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210533377.8A CN115119017B (en) | 2022-05-17 | 2022-05-17 | Medical health insurance recommendation and management method and system based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115119017A CN115119017A (en) | 2022-09-27 |
| CN115119017B true CN115119017B (en) | 2023-12-08 |
Family
ID=83326877
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210533377.8A Active CN115119017B (en) | 2022-05-17 | 2022-05-17 | Medical health insurance recommendation and management method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115119017B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016095576A1 (en) * | 2014-12-19 | 2016-06-23 | 百度在线网络技术（北京）有限公司 | Method, device and system for pushing information to user on barrage carrier |
| WO2016095525A1 (en) * | 2014-12-19 | 2016-06-23 | 百度在线网络技术（北京）有限公司 | Method, device and system for pushing information to user on bullet screen carrier |
| CN106572399A (en) * | 2016-11-18 | 2017-04-19 | 广州爱九游信息技术有限公司 | Information recommendation method and device, server and user terminal |
| CN113596587A (en) * | 2021-07-28 | 2021-11-02 | 腾讯科技（深圳）有限公司 | Bullet screen implementation method and device, electronic equipment and readable medium |
| CN114071174A (en) * | 2021-11-15 | 2022-02-18 | 咪咕文化科技有限公司 | Processing method of live content, electronic equipment and readable storage medium |
| CN114390306A (en) * | 2021-12-27 | 2022-04-22 | 中国电信股份有限公司 | Live broadcast interactive abstract generation method and device |
-
2022
- 2022-05-17 CN CN202210533377.8A patent/CN115119017B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016095576A1 (en) * | 2014-12-19 | 2016-06-23 | 百度在线网络技术（北京）有限公司 | Method, device and system for pushing information to user on barrage carrier |
| WO2016095525A1 (en) * | 2014-12-19 | 2016-06-23 | 百度在线网络技术（北京）有限公司 | Method, device and system for pushing information to user on bullet screen carrier |
| CN106572399A (en) * | 2016-11-18 | 2017-04-19 | 广州爱九游信息技术有限公司 | Information recommendation method and device, server and user terminal |
| CN113596587A (en) * | 2021-07-28 | 2021-11-02 | 腾讯科技（深圳）有限公司 | Bullet screen implementation method and device, electronic equipment and readable medium |
| CN114071174A (en) * | 2021-11-15 | 2022-02-18 | 咪咕文化科技有限公司 | Processing method of live content, electronic equipment and readable storage medium |
| CN114390306A (en) * | 2021-12-27 | 2022-04-22 | 中国电信股份有限公司 | Live broadcast interactive abstract generation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115119017A (en) | 2022-09-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110430471B (en) | Television recommendation method and system based on instantaneous calculation | |
| US11606323B2 (en) | Prioritizing messages within a message network | |
| CN106484858B (en) | hot content pushing method and device | |
| CN111163359B (en) | Bullet screen generation method and device and computer readable storage medium | |
| CN110543598B (en) | Information recommendation method and device and terminal | |
| CN111835760B (en) | Alarm information processing method and device, computer storage medium and electronic equipment | |
| CN104111925B (en) | Item recommendation method and device | |
| JP2017533531A (en) | Focused sentiment classification | |
| CN111279332A (en) | Generating request-agnostic interaction scores for electronic communications using machine-learning models and utilizing request-agnostic interaction scores | |
| US20150112814A1 (en) | System and method for an integrated content publishing system | |
| CN112347787A (en) | Method, device and equipment for classifying aspect level emotion and readable storage medium | |
| CN116405453B (en) | Information pushing method based on multiple features, storage medium and electronic equipment | |
| CN113177160A (en) | Pushed document generation method and device, electronic equipment and storage medium | |
| CN114428910A (en) | Resource recommendation method and device, electronic equipment, product and medium | |
| US20160275193A1 (en) | Socially driven feed of aggregated content in substantially real time | |
| CN111242110A (en) | Training method of self-adaptive conditional random field algorithm for automatically breaking news items | |
| CN115119017B (en) | Medical health insurance recommendation and management method and system based on artificial intelligence | |
| CN117057430B (en) | Model reasoning method and device based on rule accumulation and electronic equipment | |
| CN110781341A (en) | Audio album recommendation method and system fusing multi-strategy recall data sets | |
| US20180114279A1 (en) | Automatic Newsletter Creation | |
| CN116610872B (en) | Training method and device for news recommendation model | |
| CN112163415A (en) | User intention identification method and device for feedback content and electronic equipment | |
| CN114913012A (en) | Stock public opinion monitoring method and system based on artificial intelligence | |
| US20210103636A1 (en) | Summarization of group chat threads | |
| CN111859124A (en) | Intelligent target object recommendation method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20231114 Address after: Room 3419, No. 58 Fumin Branch Road, Hengsha Township, Chongming District, Shanghai 200085 (Hengtai Economic Development Zone, Shanghai) Applicant after: Shanghai Nanyan Information Technology Co.,Ltd. Address before: No. 11-1, Jiangsu Road, Chengguan District, Lhasa City, Tibet Autonomous Region 850000 Applicant before: Zhang Chungui |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |