A content recommendation method and apparatus, and a computer readable medium

Abstract

The application discloses a content recommendation method and device and a computer readable medium, and belongs to the technical field of data processing. The method is applied to a customer service system. The method comprises the following steps: firstly, based on the triggering of a preset time, detecting whether there is conversation content in the chat interface of a user and a customer service; secondly, if the detection result represents that there is conversation content, generating recommendation content based on the conversation content; if the detection result represents that there is no conversation content, performing feature recognition on a user who logs in the customer service system, and generating recommendation content; finally, sending the recommendation content to the user. Thus, real-time dialogue and / or commodity recommendation are performed in two scenes of having conversation content and not having conversation content, user communication of a salesperson is assisted, language expression of the customer service system is standardized, reply efficiency and accuracy of the customer service system are improved, and promotion of a commodity in a later period is facilitated.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a content recommendation method, apparatus and computer-readable medium. Background Technology

[0002] Today, in order to facilitate the simultaneous handling of various needs of a large number of users, call centers have emerged, which are composed of a group of service personnel (i.e., customer service personnel) and built using Computer Telephony Integration (CTI) technology.

[0003] Typically, users only contact customer service through call centers when they encounter problems, and their emotional state and conversational engagement often vary depending on the issue. Currently, however, customer service representatives usually use standardized scripts to respond to these situations. In other words, there is no technology to recommend scripts tailored to different users, ensuring accuracy. Furthermore, in the current market, different sales or customer service representatives communicate with users in significantly different ways, hindering standardized management. Even within the same sales or customer service representative's interactions with users, delays or inefficiencies in response times are frequent.

[0004] Therefore, there is a lack of auxiliary products for real-time customer service scripts and product recommendations. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, embodiments of the present invention provide a content recommendation method, apparatus, and computer-readable medium, which can improve the efficiency and accuracy of customer service system responses.

[0006] According to a first aspect of the present invention, a content recommendation method is provided, the method being applied to a customer service system; the method includes: detecting whether there is conversation content in the chat interface between a user and a customer service representative based on a preset time trigger; if the detection result indicates that conversation content exists, generating recommended content based on the conversation content; if the detection result indicates that there is no conversation content, performing feature recognition on the user logged into the customer service system, generating recommended content; and sending the recommended content to the user.

[0007] Optionally, if the detection result indicates the existence of conversation content, then generating recommended content based on the conversation content includes: if the detection result indicates the existence of conversation content, then retrieving the conversation content from the chat interface; selecting target conversation content from the retrieved conversation content; and selecting recommended content corresponding to the target conversation content from the database.

[0008] Optionally, selecting recommended content corresponding to the target conversation content from the database includes: performing word segmentation on the target conversation content; extracting keywords from the word segmentation results; calculating text similarity between the extracted keywords and the words in the word script library and the products in the product library; and using the recommended products and recommended words that meet the preset conditions in the similarity calculation results as the recommended content corresponding to the target conversation content.

[0009] Optionally, the step of performing feature identification on users logging into the customer service system and generating recommended content includes: obtaining user tags and user attributes of users logging into the first device; performing user feature identification based on the user tags and user attributes; performing cluster prediction on the users using a clustering model based on the identified user features to obtain the user group corresponding to the user; querying recommended scripts corresponding to the user group from the script library and querying recommended products corresponding to the user group from the product library; and using the recommended scripts and recommended products together as recommended content.

[0010] Optionally, the recommended content includes at least recommended phrases and recommended products; before sending the recommended content to the user, the method further includes: counting the total number of recommended phrases and recommended products; determining whether the total number meets a preset threshold; if the determination result indicates that the total number meets the preset threshold, then the recommended phrases and recommended products are used together as recommended content.

[0011] Optionally, the method further includes: for any of the recommended phrases or recommended products in the recommended content: obtaining all calculated indicators corresponding to the recommended phrase; obtaining the normalized indicator values ​​of the user group for each of the calculated indicators; weighting the indicator values ​​corresponding to all calculated indicators to generate a score corresponding to the recommended phrase; and updating the order of the recommended phrases corresponding to the user group based on the score of the recommended phrase.

[0012] Optionally, obtaining the normalized index value for each calculated index for the user group includes: for any calculated index: obtaining the index value corresponding to the calculated index for each user group, resulting in several index values; removing the largest index value from the several index values ​​and replacing the largest index value with the second largest index value, resulting in several new index values; and normalizing the new several index values ​​to obtain the normalized index value for the user group for the calculated index.

[0013] Optionally, the clustering model is obtained by the following method: acquiring user data of several users; wherein the user data includes user attributes and user tags; performing data cleaning processing on the user data of each user; and performing feature hashing on the cleaned user data to obtain hash features; using the hash features corresponding to each user as training samples to train the model and generate a K-means model.

[0014] According to a second aspect of the present invention, a content recommendation apparatus is provided, the apparatus being applied to a customer service system; the apparatus includes: a detection module, configured to detect whether there is conversation content in the chat interface between a user and a customer service representative based on a preset time trigger; a first generation module, configured to generate recommended content based on the conversation content if the detection result indicates that conversation content exists; a second generation module, configured to perform feature recognition on the user logged into the customer service system and generate recommended content if the detection result indicates that no conversation content exists; and a sending module, configured to send the recommended content to the user.

[0015] According to a third aspect of the present invention, a computer-readable medium is also provided, on which a computer program is stored, wherein the program, when executed by a processor, implements the method described in the first aspect.

[0016] This invention provides a content recommendation method, apparatus, and computer-readable medium. The method is applied to a customer service system. The method includes: first, based on a preset time trigger, detecting whether there is conversation content in the chat interface between the user and customer service representative; second, if the detection result indicates the existence of conversation content, generating recommended content based on the conversation content; if the detection result indicates the absence of conversation content, performing feature recognition on the user logged into the customer service system and generating recommended content; and finally, sending the recommended content to the user. Thus, real-time scripted responses and / or product recommendations are performed in both scenarios with and without conversation content, assisting sales personnel in communicating with users, standardizing the language expression of the customer service system, improving the response efficiency and accuracy of the customer service system, and facilitating subsequent product promotion. Attached Figure Description

[0017] The following sections will describe some specific embodiments of the invention in detail by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0018] Figure 1 This is a flowchart illustrating a content recommendation method provided in an embodiment of the present invention;

[0019] Figure 2 This is a schematic diagram of the process for generating recommended content based on conversation content in another embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of the process for generating recommended content based on user features in another embodiment of the present invention;

[0021] Figure 4 This is a schematic diagram of the structure of a content recommendation device provided in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] like Figure 1 The diagram shown is a flowchart illustrating a content recommendation method provided in an embodiment of the present invention.

[0024] A content recommendation method, applied to a customer service system, comprising at least the following steps:

[0025] S101, based on a preset time trigger, detects whether there is conversation content in the chat interface between the user and customer service;

[0026] S102, If the detection result indicates the existence of session content, then generate recommended content based on the session content;

[0027] S103, If the detection result indicates that there is no session content, then perform feature recognition on the user who logged into the customer service system and generate recommended content;

[0028] S104, the generated recommendation content is sent to the user.

[0029] For example, based on a preset time trigger, the customer service system checks if there is conversation content in the chat interface. If the detection result indicates the presence of conversation content, recommended content is generated in real time based on that content. If the detection result indicates the absence of conversation content, the system performs feature identification on the user logged into the customer service system and generates recommended content based on the identified user features. Then, it determines whether the recommended content meets preset conditions. If the determination result indicates that the recommended content meets the preset conditions, the generated recommended content is sent to the user based on administrator triggering, or the customer service system automatically sends the generated recommended content to the user. If the determination result indicates that the recommended content does not meet the preset conditions, a default message is displayed, and no recommended content is shown. Here, recommended content includes recommended messages and recommended products.

[0030] This embodiment is based on a preset time trigger. The customer service system detects the conversation content in the chat interface. If there is no conversation content in the chat interface, it provides real-time scripts and product recommendations based on user characteristics. If there is conversation content in the chat interface, it provides real-time scripts and product recommendations based on the conversation content. This improves the response efficiency and accuracy of the customer service system, which is beneficial for the subsequent promotion of products and enhances the user experience.

[0031] like Figure 2 The diagram shown illustrates a process for generating recommended content based on conversation content in another embodiment of the present invention. Generating recommended content based on conversation content includes at least the following steps:

[0032] S201, based on a preset time trigger, detects whether there is conversation content in the chat interface between the user and customer service;

[0033] S202, If the detection result indicates the existence of conversation content, then retrieve the conversation content from the chat interface;

[0034] S203, Select the target session content from the retrieved session content;

[0035] S204, perform word segmentation on the target session content; and extract keywords from the word segmentation results;

[0036] S205, calculate the text similarity between the extracted keywords and the texts in the script database and the products in the product database respectively;

[0037] S206, The recommended products and recommended phrases that meet the preset conditions in the similarity calculation results are used together as the recommended content corresponding to the target conversation content.

[0038] S207, Count the total number of recommended phrases and recommended products;

[0039] S208, determine whether the total quantity meets the preset threshold; if yes, proceed to S209; if no, proceed to S210.

[0040] S209, send the recommended message and recommended products together as the recommended content to the user;

[0041] S210, no recommended content, default recommended content is displayed.

[0042] Specifically, if the detection results indicate the existence of conversation content, the conversation content is retrieved from the chat interface and archived. From the retrieved conversation content, the three most recent conversations with more than 5 characters each are selected as target conversation content. For any target conversation content: the target conversation content is segmented, and keywords are extracted from the segmentation results; several conversation scripts with high similarity to the keywords are selected from the script library; the top three similarity scripts are selected as recommended scripts; several products with high similarity to the keywords are selected from the product library; the top two similarity products are selected as recommended products. The total number of recommended products and recommended scripts is counted; if the total number is not less than 1, the recommended scripts and recommended products are sent to the user as recommended content; if the total number is less than 1, the default recommended content from the customer service system is sent to the user.

[0043] This embodiment generates recommended content in real time based on the conversation between the user and customer service representative. When the recommended content meets preset conditions, it is sent to the user. As a result, suitable scripts or products can be quickly recommended based on the conversation content, which not only improves the accuracy of the customer service system's response to the user's conversation content, but also improves the user's response efficiency and thus enhances the user experience.

[0044] like Figure 3 The diagram shown is a flowchart illustrating the process of generating recommended content based on user characteristics in another embodiment of the present invention.

[0045] In scenarios where users haven't sent messages to customer service but customer service needs to push notifications to users, recommendations can be made based on user characteristics. This includes daily scheduled updates and usage:

[0046] The daily scheduled updates consist of two parts: updates to the k-means clustering model and updates to the recommended messaging for each cluster of users. The update frequency can be adjusted based on system traffic; during special events or periods of high user activity, the update frequency can be increased. The default is once a day.

[0047] Generating recommended content based on user characteristics includes at least the following steps:

[0048] S301, Obtain the user tags and user attributes of users who log in to the customer service system;

[0049] S302, perform user feature identification based on user tags and user attributes;

[0050] S303, based on the identified user characteristics, uses a clustering model to perform clustering prediction on users to obtain the user groups corresponding to the users;

[0051] S304. Retrieve recommended scripts corresponding to the user group from the script database, and retrieve recommended products corresponding to the user group from the product database;

[0052] S305, Count the total number of recommended phrases and recommended products;

[0053] S306, determine whether the total quantity meets the preset threshold; if yes, proceed to S307; if no, proceed to S308.

[0054] S307 sends the recommended message and recommended products together as the recommended content to the user;

[0055] S308, no recommended content, default recommended content is displayed.

[0056] Specifically, user attributes include gender, age, and region; user tags include high-value users, low-value users, and members.

[0057] There is a mapping relationship between the scripts in the script library and user groups, and a mapping relationship between the products in the product library and user groups. After determining the user group corresponding to a user, the recommended scripts and recommended products corresponding to that user group are retrieved from the script library based on the mapping relationship.

[0058] It should be noted that the recommended scripts for user groups in the script library and the recommended products for user groups in the product library are updated in real time.

[0059] The clustering model was obtained through the following method:

[0060] S1, Obtain user data for several users; where user data includes user attributes and user tags;

[0061] S2, perform data cleaning on the user data of each user; and perform feature hashing on the cleaned user data to obtain hash features;

[0062] S3 uses the hash features corresponding to each user as training samples to train the model and generate a K-means model.

[0063] Since the user labels for each user change, the K-means model is also updated in real time.

[0064] This embodiment determines the user group corresponding to a user based on user characteristics, and sends the recommended scripts and recommended products corresponding to the user group to the user; thus, content recommendation can be made based on user characteristics even without conversation content, improving the response efficiency and accuracy of the customer service system.

[0065] It should also be noted that the user-feature-based recommendations update the word order ranking based on the k-means model and calculated metrics. If a user then replies, the recommendations are converted to word order recommendations based on the conversation content. This allows for real-time updates to the word order or product ranking corresponding to a user group based on the replies, improving the user experience.

[0066] In a preferred embodiment of this example, updating the recommended script or recommended product for a user group includes at least the following steps:

[0067] S1, for any recommended phrase or recommended product in the recommended content: obtain all the calculated indicators corresponding to the recommended phrase; obtain the normalized indicator value of each calculated indicator for the user group; perform weighted processing on the indicator values ​​corresponding to all calculated indicators to generate the score corresponding to the recommended phrase;

[0068] S2, based on the rating of the recommended phrases, update the order of the recommended phrases corresponding to the user group.

[0069] Specifically, each sales pitch has one or more calculated metrics, the exact number of which needs to be tracked by backend system tracking. Backend tracking can be done as follows: Sales pitch: "Hi, the product you purchased before is currently on sale! You can stock up while it's on." The backend can then calculate the "number of replies" as a metric for this sales pitch. For recommended products, it can calculate the "number of clicks," "number of purchases," etc., as calculated metrics. Calculated metrics include the number of replies, the number of completed orders, the number of clicks, and the number of purchases.

[0070] In S1, for any calculated indicator: obtain the indicator value corresponding to the calculated indicator for each user group, and obtain several indicator values; remove the largest indicator value from the several indicator values, and replace the largest indicator value with the second largest indicator value to obtain several new indicator values; normalize the new several indicator values ​​to obtain the normalized indicator value for the calculated indicator for the user group.

[0071] For example: User groups 1 through 10 each have corresponding scripts 1, 2, and 3; the calculation metrics for script 1 are the number of replies, the number of completed orders, and the number of clicks; the calculation metric for script 2 is the average number of times per buyer; and the calculation metrics for script 3 are the number of conversations, the number of replies, and the number of purchases. For any calculation metric in script 1: Calculate the metric value for user groups 1 through 10, resulting in 10 metric values; remove the largest metric value from these 10 values ​​and replace it with the second largest metric value, resulting in 10 new metric values; normalize these new 10 metric values ​​to obtain the metric value for each user group for that calculation metric.

[0072] The weighted values ​​of all calculated indicators corresponding to Script 1 in User Group 1 are used to obtain the first score for Script 1; similarly, the second score for Script 2 in User Group 1 is determined; and the third score for Script 3 in User Group 2 is determined. Then, based on the scores, the scores of Script 1, Script 2, and Script 3 in User Group 1 are re-ranked.

[0073] Similarly, the ranking results of the scripts in each user group from user group 2 to user group 10 can be obtained. Therefore, in this embodiment, after each recommended script is sent, multiple calculated indicators will be tracked for the corresponding recommended script through backend tracking points to provide data for the next model update, thereby improving the user response rate.

[0074] After each clustering model update, the calculated metrics for each user group are statistically analyzed to obtain the ranking of recommended phrases for that user group. The calculation process mainly consists of steps such as removing maxima, normalization, and weighted calculation of the comprehensive value.

[0075] The content recommendation method in this embodiment will be explained in detail below with specific applications. Taking Double Eleven as an example:

[0076] During the Double Eleven shopping festival, when sales staff chat with users about Double Eleven-related topics, the recommendation system will detect the keyword "Double Eleven" in the target conversation content and recommend relevant content from the dialogue database.

[0077] Because the usage rate of specific sales scripts will increase significantly during the Double Eleven promotion, this will also affect various calculation metrics. First, increase the frequency of scheduled tasks to add promotional scripts to the recommendation list more quickly. Second, shorten the effective period of sales script metrics. For example, for the conversion rate metric, remove conversions before the Double Eleven promotion and only calculate from the start of the Double Eleven promotion. This will indirectly increase the weight of promotional scripts, ensuring they appear in the recommendation list more promptly and accurately, thus improving the customer system's response rate.

[0078] The various embodiments of the present invention are implemented through programmed processing using a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of the various embodiments of the present invention can be encapsulated into various modules.

[0079] like Figure 4 The diagram shown is a structural schematic of a content recommendation device provided in an embodiment of the present invention.

[0080] A content recommendation device, applied to a customer service system, includes: a detection module 401, used to detect whether there is conversation content in the chat interface between a user and a customer service representative based on a preset time trigger; a first generation module 402, used to generate recommended content based on the conversation content if the detection result indicates the existence of conversation content; a second generation module 403, used to perform feature recognition on the user logged into the customer service system and generate recommended content if the detection result indicates the absence of conversation content; and a sending module 404, used to send the recommended content to the user.

[0081] In a preferred embodiment, the first generation module includes: a retrieval unit, configured to retrieve conversation content from the chat interface if the detection result indicates the existence of conversation content; a first selection unit, configured to select target conversation content from the retrieved conversation content; and a second selection unit, configured to select recommended content corresponding to the target conversation content from the database.

[0082] In a preferred embodiment, the second selection unit includes: a word segmentation processing subunit, used to perform word segmentation processing on the target conversation content; and extract keywords from the word segmentation processing results; a similarity calculation subunit, used to perform text similarity calculations between the extracted keywords and the words in the word script library and the products in the product library; and a determination subunit, used to select recommended products and recommended words that meet preset conditions in the similarity calculation results as recommended content corresponding to the target conversation content.

[0083] In a preferred embodiment, the second generation module includes: an acquisition unit, configured to acquire user tags and user attributes of a user logged into the first device; a feature recognition unit, configured to perform user feature recognition based on the user tags and user attributes; a prediction unit, configured to perform cluster prediction on the user based on the identified user features using a clustering model to obtain a user group corresponding to the user; a query unit, configured to query recommended dialogues corresponding to the user group from a dialogue library and query recommended products corresponding to the user group from a product library; and a determination unit, configured to use the recommended dialogues and recommended products together as recommended content.

[0084] In a preferred embodiment, the recommended content includes at least recommended phrases and recommended products; the device further includes: a statistics module for counting the total number of recommended phrases and recommended products; a judgment module for judging whether the total number meets a preset threshold; and a determination module for using the recommended phrases and recommended products together as recommended content if the judgment result indicates that the total number meets the preset threshold.

[0085] In a preferred embodiment, the device further includes: a third generation module, configured to, for any of the recommended phrases or recommended products in the recommended content: obtain all calculated indicators corresponding to the recommended phrase; obtain the normalized indicator values ​​of the user group for each of the calculated indicators; perform weighted processing on the indicator values ​​corresponding to all calculated indicators to generate a score corresponding to the recommended phrase; and an update module, configured to update the order of the recommended phrases corresponding to the user group based on the score of the recommended phrase.

[0086] In a preferred embodiment, the generation module includes: an acquisition unit, configured to acquire, for any given calculation indicator, the indicator value corresponding to each user group for the calculation indicator, thereby obtaining several indicator values; a substitution unit, configured to remove the largest indicator value from the several indicator values ​​and substitute the largest indicator value with the second largest indicator value, thereby obtaining several new indicator values; and a normalization processing unit, configured to normalize the new several indicator values, thereby obtaining the normalized indicator values ​​for the user group for the calculation indicator.

[0087] In a preferred embodiment, the clustering model is obtained by the following method: acquiring user data of several users; wherein the user data includes user attributes and user tags; performing data cleaning processing on the user data of each user; and performing feature hashing on the cleaned user data to obtain hash features; using the hash features corresponding to each user as training samples to train the model and generate a K-means model.

[0088] The above-described apparatus can execute the content recommendation method provided in an embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the content recommendation method. Technical details not described in detail in this embodiment can be found in the content recommendation method provided in an embodiment of the present invention.

[0089] The present invention also provides an electronic device, comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the content recommendation method of the present invention.

[0090] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0091] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0092] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods according to the following embodiments of this application described in the "Exemplary Methods" section above.

[0093] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0094] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0095] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0096] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0097] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0098] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

[0099] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0100] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0101] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A content recommendation method, characterized in that, Applied to a customer service system; the method includes: Based on a preset time trigger, detect whether there is conversation content in the chat interface between the user and customer service; If the detection results indicate the presence of conversation content, then recommended content is generated based on the conversation content. If the detection result indicates that there is no session content, then the user tag and user attributes of the user who logged into the first device are obtained; user feature identification is performed based on the user tag and user attributes; based on the identified user features, a clustering model is used to perform cluster prediction on the user to obtain the user group corresponding to the user; recommended dialogues corresponding to the user group are queried from the dialogue library, and recommended products corresponding to the user group are queried from the product library; the recommended dialogues and recommended products are used together as recommended content; The recommended content will be sent to the user; For any of the recommended phrases or recommended products in the recommended content: obtain all calculated indicators corresponding to the recommended phrase; obtain the normalized indicator values ​​of the user group for each of the calculated indicators; perform weighted processing on the indicator values ​​corresponding to all calculated indicators to generate a score corresponding to the recommended phrase; Based on the rating of the recommended phrases, the order of the recommended phrases corresponding to the user group is updated.

2. The method according to claim 1, characterized in that, If the detection result indicates the presence of session content, then recommended content is generated based on the session content; including: If the detection result indicates the presence of conversation content, then the conversation content is retrieved from the chat interface; Select the target session content from the retrieved session content; Select recommended content from the database that corresponds to the target session content.

3. The method according to claim 2, characterized in that, The step of selecting recommended content from the database that corresponds to the target session content includes: The target session content is segmented into words; and keywords are extracted from the segmentation results. The extracted keywords are compared with the dialogues in the dialogue database and the products in the product database to calculate the text similarity. The recommended products and recommended phrases that meet the preset conditions in the similarity calculation results are used together as the recommended content corresponding to the target conversation content.

4. The method according to claim 1, characterized in that, Before sending the recommended content to the user, the method further includes: counting the total number of the recommended phrases and the recommended products; Determine whether the total quantity meets the preset threshold; If the judgment result indicates that the total quantity meets the preset threshold, then the recommended script and the recommended product will be used together as the recommended content.

5. The method according to claim 1, characterized in that, The step of obtaining the normalized index value for each calculated index for the user group includes: For any given calculation metric: obtain the metric value corresponding to the calculation metric for each user group, resulting in several metric values; remove the largest metric value from the several metric values ​​and replace the largest metric value with the second largest metric value to obtain several new metric values; normalize the new several metric values ​​to obtain the normalized metric values ​​for the calculation metric for the user group.

6. The method according to claim 1, characterized in that, The clustering model was obtained through the following method: Obtain user data from several users; wherein, the user data includes user attributes and user tags; The user data of each user is cleaned; and the cleaned user data is subjected to feature hashing to obtain hash features. The hash features corresponding to each user are used as training samples to train the model and generate a K-means model.

7. A content recommendation device, characterized in that, Applied to a customer service system; The device includes: The detection module is used to detect whether there is conversation content in the chat interface between the user and customer service based on a preset time trigger. The first generation module is used to generate recommended content based on the session content if the detection result indicates that session content exists. The second generation module is used to obtain the user tag and user attributes of the user who logged into the first device if the detection result indicates that there is no session content; perform user feature identification based on the user tag and user attributes; perform cluster prediction on the user using a clustering model based on the identified user features to obtain the user group corresponding to the user; query the recommended dialogue corresponding to the user group from the dialogue library and query the recommended products corresponding to the user group from the product library; and use the recommended dialogue and the recommended products together as the recommended content. The sending module is used to send the recommended content to the user; The third generation module is used for any of the recommended phrases or recommended products in the recommended content to: obtain all calculated indicators corresponding to the recommended phrases; obtain the normalized indicator values ​​of the user group for each of the calculated indicators; and perform weighted processing on the indicator values ​​corresponding to all calculated indicators to generate a score corresponding to the recommended phrases. The update module is used to update the order of recommended phrases corresponding to the user group based on the rating of the recommended phrases.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images