Feedback information processing method and system and computer program product

By collecting and processing user statements, using whitelists, classification models and clustering technology to extract feedback problem descriptions, the problem of extracting product problems in user online feedback information is solved, efficient and intelligent problem discovery and resolution are achieved, and product operations are improved efficiency.

CN119027191APending Publication Date: 2024-11-26NETEASE (HANGZHOU) NETWORK CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411164588.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing technology is difficult to extract product problems from users' online feedback information efficiently and intelligently, making it difficult to solve product problems in a timely manner and affects operations.

Method used

By collecting user statements, filtering and classification based on whitelists and classification models, extracting feedback problem descriptions, and generating class cluster abstracts through clustering and language models, presenting feedback problems, and creating task items in response to user follow-up requests.

Benefits of technology

It has achieved efficient and intelligent discovery of problems from users' online speeches, timely solve product problems, reduce negative impacts on operations, and improve product optimization efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027191A_ABST
    Figure CN119027191A_ABST
Patent Text Reader

Abstract

The invention provides a feedback information processing method and system and a computer program product, and relates to the technical field of big data. The feedback information processing method comprises the steps of collecting a first user statement; processing the first user statement, and extracting one or more feedback problem descriptions; presenting at least part of feedback problem descriptions in the one or more feedback problem descriptions through one or more feedback data pages; and in response to a feedback problem follow-up request initiated through at least one feedback data page in the one or more feedback data pages, creating a task item and sending the task item to a second user side associated with the task item.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a feedback information processing method, system and computer program product. Background Art

[0002] Online feedback from users (such as product consumers) is an important way for merchants (such as product providers) to understand product problems. If users' online feedback can be collected efficiently and proactively, it will help merchants solve product problems and optimize products in a timely manner. On the other hand, it can also timely avoid serious negative impacts on operations caused by product problems.

[0003] To this end, some embodiments of this specification are intended to provide a feedback information processing method, so as to efficiently and intelligently discover problems from users' online comments and solve them in a timely manner. Summary of the invention

[0004] One or more embodiments of the present specification provide a feedback information processing method, including: collecting first user statements, wherein the first user statements include statements reflecting the first user's remarks; processing the first user statements to extract one or more feedback problem descriptions; presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages; and in response to a feedback problem follow-up request initiated through at least one of the one or more feedback data pages, creating a task item and sending it to a second user terminal associated with the task item.

[0005] According to the method described in some embodiments of this specification, the first user statement comes from one or more of the following data: online speech, chat data, first user question, and first user suggestion.

[0006] According to the method described in some embodiments of the present specification, the processing of the first user statement to extract one or more feedback problem descriptions includes: filtering the first user statement based on a whitelist to obtain a filtered first user statement; classifying the filtered first user statement based on a first classification model to obtain first user statements belonging to a target category and a non-target category, and using the first user statement in the target category as the filtered first user statement; obtaining one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters based on the filtered first user statement; generating a cluster summary of the problem feedback cluster, and removing the first user statement corresponding to the problem feedback cluster whose cluster summary does not belong to the target category.

[0007] According to the method described in some embodiments of the present specification, the filtering of the first user sentence based on the whitelist to obtain the filtered first user sentence includes: removing the first user sentence whose number of repetitions is greater than a first set threshold in the first user sentence to obtain the remaining first user sentence; removing the first user sentence containing a preset whitelist keyword in the remaining first user sentence, and / or removing the first user sentence whose semantic similarity with the preset whitelist keyword in the remaining first user sentence is greater than a second set threshold to obtain the filtered first user sentence; the processing of the first user sentence to extract one or more feedback problem descriptions also includes: before obtaining one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters based on the filtered first user sentence, processing the first user sentence whose word count is greater than a third set threshold in the filtered first user sentence based on the first language model to obtain the first user sentence whose word count is not greater than a fourth set threshold; wherein the fourth set threshold is less than or equal to the third set threshold.

[0008] According to the method described in some embodiments of the present specification, the removing of the first user sentence corresponding to the question feedback cluster whose cluster summary does not belong to the target category includes: obtaining web page data through a search engine based on the cluster summary of the question feedback cluster, and determining whether the cluster summary of the question feedback cluster belongs to the target category based on the web page data through a second language model; when the cluster summary of the question feedback cluster does not belong to the target category, removing the first user sentence corresponding to the question feedback cluster.

[0009] According to the method described in some embodiments of the present specification, the obtaining of one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters based on the filtered first user statement includes: processing the filtered first user statement by a first clustering method to obtain one or more first problem feedback clusters and one or more first feedback problem descriptions corresponding to the first problem feedback clusters; processing the filtered first user statement by a second clustering method to obtain one or more second problem feedback clusters and one or more second feedback problem descriptions corresponding to the second problem feedback clusters; wherein the first clustering method and the second clustering method correspond to different semantic levels and / or semantic classification granularities.

[0010] According to the method described in some embodiments of the present specification, the first user statement after filtering is processed by a first clustering method to obtain one or more first problem feedback clusters and one or more first feedback problem descriptions corresponding to the first problem feedback clusters, including: clustering the first user statement after filtering to obtain the one or more first problem feedback clusters; generating a cluster summary of the first problem feedback cluster based on one or more first user statements in the first problem feedback cluster whose statement confidence satisfies a first set condition; wherein the statement confidence of the first user statement is determined based on keywords extracted from the first user statement and one or more of the following information: portrait information of the first user related to the first user statement, source of the first user statement; and obtaining the one or more first feedback problem descriptions based on the cluster summary of the one or more first problem feedback clusters.

[0011] According to the method described in some embodiments of the present specification, presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages includes: presenting at least part of the one or more first feedback problem descriptions and a corresponding expansion control through a first feedback data page; in response to a triggering operation of the expansion control on the first feedback data page, presenting a first user statement in a problem feedback cluster corresponding to the first feedback problem description corresponding to the expansion control.

[0012] According to the method described in some embodiments of the present specification, the method of processing the filtered first user sentences by a second clustering method to obtain one or more second problem feedback clusters and one or more second feedback problem descriptions corresponding to the second problem feedback clusters includes: classifying the filtered first user sentences based on a second classification model to obtain first user sentences belonging to one or more problem labels; obtaining the one or more second feedback problem descriptions based on the one or more problem labels; clustering the first user sentences belonging to the same problem label to obtain one or more second problem feedback clusters belonging to the problem label; generating a cluster summary of the second problem feedback cluster based on one or more first user sentences in the second problem feedback cluster whose sentence confidence meets the second set condition; wherein the sentence confidence of the first user sentence is determined based on keywords extracted from the first user sentence and one or more of the following information: portrait information of the first user related to the first user sentence, and the source of the first user sentence.

[0013] According to the method described in some embodiments of the present specification, the first user statement after filtering is classified based on the second classification model to obtain the first user statement belonging to one or more question tags, including: determining classification scores for the first user statement after filtering respectively through the second classification model; classifying the first user statement after filtering based on the classification scores and the score ranges corresponding to the one or more question tags to obtain the first user statement belonging to one or more question tags; wherein the score ranges corresponding to the one or more question tags are adjustable.

[0014] According to the method described in some embodiments of the present specification, the presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages includes: presenting at least part of the one or more second feedback problem descriptions and the corresponding expansion control through the second feedback data page; in response to the triggering operation of the expansion control on the second feedback data page, presenting a cluster summary of the problem feedback cluster belonging to the second feedback problem description corresponding to the expansion control, and / or presenting a first user statement belonging to the second feedback problem description corresponding to the expansion control.

[0015] According to the method described in some embodiments of the present specification, a feedback problem follow-up control corresponding to the feedback problem description is also presented on the one or more feedback data pages; the response to a feedback problem follow-up request initiated through at least one feedback data page among the one or more feedback data pages, creating a task item and sending it to a second user terminal associated with the task item includes: responding to a triggering operation of a feedback problem follow-up control in the at least one feedback data page, creating the task item based on the feedback problem description corresponding to the feedback problem follow-up control.

[0016] The method described in some embodiments of the present specification also includes: presenting task item information and adjusting task progress controls through a task item page; wherein the task item information includes one or more of the following information: task name, second user information associated with the task item, task progress, and one or more first user statements related to the feedback problem description; in response to a triggering operation of adjusting the task progress control, presenting one or more first user statements related to the feedback problem description and their corresponding first user information and a reply control through a task processing page; in response to a triggering operation of the reply control, obtaining reply content and sending it to a first user terminal corresponding to at least some of the first user statements in the one or more first user statements related to the feedback problem description.

[0017] The method described in some embodiments of the present specification also includes: presenting a first user statement and its corresponding verification control related to at least part of the feedback problem description through the feedback data page; displaying source data and / or image data of the first user statement in response to a triggering operation of the verification control corresponding to the first user statement; presenting verification controls corresponding to at least part of the feedback problem description respectively through the feedback data page; and displaying portrait information of the first user corresponding to the feedback problem description in response to a triggering operation of the verification control corresponding to the feedback problem description.

[0018] According to the method described in some embodiments of the present specification, it also includes: when at least part of the one or more feedback problem descriptions meet the preset alarm conditions corresponding to one or more alarm types, sending an alarm message to one or more second user terminals through the alarm window; the alarm message includes one or more of the following information: alarm type, feedback problem description related to the alarm message, one or more first user statements related to the feedback problem description; the one or more alarm types include one or more of the following types: the emergence of new feedback problem descriptions, a sharp increase in the number of first user statements related to the feedback problem description, the feedback problem description is related to the first user who is the focus of attention, and the list is broadcast regularly.

[0019] The method described in some embodiments of the present specification also includes: when at least part of the first user statements in the first user statement meets the alarm condition that the feedback problem description involves key focus content, sending an alarm message to one or more second user terminals through the alarm window; the alarm message includes one or more of the following information: the alarm type, the key focus content related to the alarm information, and one or more first user statements containing the key focus content related to the alarm information.

[0020] The method described in some embodiments of the present specification also includes: in response to a priority configuration operation corresponding to one or more feedback problem descriptions, setting the priority of the corresponding feedback problem description; when at least part of the one or more feedback problem descriptions meet the preset alarm conditions and priority conditions corresponding to one or more alarm types, sending an alarm message to one or more second user terminals through the alarm window; the alarm message includes one or more of the following information: an alarm type, a feedback problem description related to the alarm information, and one or more first user statements related to the feedback problem description.

[0021] According to the method described in some embodiments of the present specification, the alarm window also includes a feedback question follow-up control; the method also includes: in response to a triggering operation of the feedback question follow-up control, creating a task item based on the alarm information and sending it to a second user terminal associated with the task item.

[0022] According to the method described in some embodiments of the present specification, the feedback issue follow-up control includes a second user information input box and a sending control; the response to the triggering operation of the feedback issue follow-up control, creating a task item based on the alarm information and sending it to the second user terminal associated with the task item, includes: obtaining the second user information through the second user information input box; in response to the triggering operation of the sending control, sending the task item to the second user terminal corresponding to the second user information, and switching the display properties of the alarm window to a first state to indicate that the task process is in progress.

[0023] According to the method described in some embodiments of the present specification, the method also includes: in response to the triggering operation of the feedback question follow-up control, switching the display properties of the alarm window to a first state to indicate that the task process is in progress, displaying the second user information associated with the task item and displaying the processing control and / or the completion control in the alarm window; in response to the triggering operation of the processing control, obtaining the second user information corresponding to the second user end that initiated the triggering operation and additionally displaying it in the alarm window; in response to the triggering operation of the completion control, switching the display properties of the alarm window to a second state to indicate that the task process is completed.

[0024] According to the method described in some embodiments of the present specification, the method also includes: creating a task item based on the alarm information, and sending it to a second user terminal associated with the task item determined based on the correspondence between preset problem keywords and the second user; switching the display properties of the alarm window to a first state to indicate that the task process is in processing, displaying the second user information associated with the task item in the alarm window and displaying processing controls and / or completion controls; in response to a trigger operation of the processing control, obtaining the second user information corresponding to the second user terminal that initiated the trigger operation and displaying it in the alarm window; in response to a trigger operation of the completion control, switching the display properties of the alarm window to a second state to indicate that the task process is completed.

[0025] According to the method described in some embodiments of this specification, the alarm window is presented in a chat interface; wherein the chat interface includes a group chat interface.

[0026] One or more embodiments of the present specification also provide a feedback information processing system, including: a collection module, used to collect first user statements, wherein the first user statements include statements reflecting the first user's remarks; a processing module, used to process the first user statements and extract one or more feedback problem descriptions; a display module, used to present at least part of the one or more feedback problem descriptions through one or more feedback data pages; and a task module, used to create a task item and send it to a second user terminal associated with the task item in response to a feedback problem follow-up request initiated through at least one of the one or more feedback data pages.

[0027] One or more embodiments of the present specification further provide a computer program product, including computer codes, which can implement any of the aforementioned methods when at least part of the computer codes is executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] This specification will be further explained by way of exemplary embodiments, which will be described in detail by way of the accompanying drawings. The same numbers in the drawings represent the same structures or steps.

[0029] Figure 1 It is an application scenario diagram of the feedback information processing system shown in some embodiments of this specification.

[0030] Figure 2 It is an exemplary flow chart of a feedback information processing method according to some embodiments of this specification.

[0031] Figure 3 This is an exemplary flowchart of the step of extracting a feedback question description based on a first user sentence according to some embodiments of this specification.

[0032] Figure 4 It is a schematic diagram of a feedback data page according to some embodiments of this specification.

[0033] Figure 5 It is a schematic diagram of a feedback data page according to other embodiments of this specification.

[0034] Figure 6 It is a schematic diagram of a task item page according to some embodiments of the present specification.

[0035] Figure 7 This is a schematic diagram of the first alarm window shown in some embodiments of this specification.

[0036] Figure 8 This is a schematic diagram of a second alarm window according to some embodiments of this specification.

[0037] Fig. 9 This is a schematic diagram of the third alarm window shown in some embodiments of this specification.

[0038] Fig.10 It is a schematic diagram of the fourth alarm window shown in some embodiments of this specification.

[0039] Fig.11 is an exemplary block diagram of a feedback information processing system according to some embodiments of the present specification. DETAILED DESCRIPTION

[0040] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the embodiments will be described in detail below with reference to the accompanying drawings. Obviously, the contents described below are some examples or embodiments of this specification. For ordinary technicians in this field, without paying creative work, the technical solutions or means disclosed in this specification can also be applied to other scenarios based on these technical contents.

[0041] It should be understood that the "system", "device", "unit" and / or "module" used in this specification is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0042] Unless otherwise specified, technical terms used in this specification to describe components, elements, etc. do not refer to the singular, but may also include the plural. Generally speaking, terms such as "include", "comprise", etc. only indicate that the steps, elements, or components that have been clearly identified are included, and these steps, elements, and components do not constitute an exclusive list, such as the method or device described may also include other steps or components.

[0043] Flowcharts are used in this specification to illustrate the operation steps performed by the device or system of the relevant embodiments, but unless otherwise specified, the order used to describe these steps should not be understood as a limitation on the order in which the steps are performed. A person of ordinary skill in the art can adjust the order in which these steps are performed based on the knowledge and information conveyed by the embodiments of this specification, and the adjustment includes but is not limited to swapping the order, merging multiple steps, and splitting a certain step.

[0044] User feedback plays an irreplaceable role in helping businesses improve their products, optimize services, and enhance user experience. There are many ways to obtain user feedback, such as inviting users to comment on the purchase page of a product or service, asking users about their post-use experience by phone, opening specific chat channels (such as customer service interfaces) to obtain users' suggestions on products or services, etc. Research surveys show that more than half of users prefer to provide product or service feedback through channels they like or are accustomed to using. With the popularization of the Internet, people are becoming more and more accustomed to expressing themselves through the Internet. Taking games as an example, most players like to express their opinions on game content, game skills, and even game problems through forums, new media platforms with short videos as the main medium, and instant messaging tools. This means that Internet data contains a lot of information related to user feedback. If this information can be effectively used, it will help businesses solve product problems and optimize products in a timely manner.

[0045] On the other hand, when there are problems with services or products, users’ online comments may have a serious negative impact on the business operations. For example, there are abnormal points in the game scene of a certain game, and players can easily win through these abnormal points. This problem has attracted a large number of players to give bad reviews to the game, and many players have quit the game. If the online feedback of users can be obtained in time, the game merchants will be able to find and solve the problems in the first time, thereby effectively avoiding more losses.

[0046] To this end, some embodiments of this specification are intended to provide a feedback information processing method, so as to efficiently and intelligently discover problems from users' online comments and solve them in a timely manner.

[0047] In some embodiments, users can be understood as consumers or users of products, or can be called first users or C-end users. Among them, products can be physical goods that can be sold and circulated in the market, such as clothes, food, computers, daily necessities, etc.; products can also be service products such as entertainment services, medical services, financial services, etc. In some embodiments, the party providing products can be called a merchant.

[0048] Figure 1 is an application scenario diagram of the feedback information processing system shown in some embodiments of this specification. Figure 1 As shown, scenario 100 may include a processing device 110 , a network 120 , a storage medium 130 , a first user terminal 140 , and a second user terminal 150 .

[0049] In some embodiments, the processing device 110 may include one or more processors or processing engines (e.g., a single-chip processing engine or a multi-chip processing engine). The processor or processing engine may execute computer instructions to implement the feedback information processing method shown in one or more embodiments of this specification. As an example only, the processor or processing engine may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), an application-specific instruction processor (ASIP), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic circuit (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, etc., or any combination thereof.

[0050] The network 120 may include any suitable network that can facilitate information and / or data exchange between components in the scene 100. In some embodiments, one or more components in the scene 100 (e.g., the processing device 110, the storage medium 130, etc.) may exchange information and / or data through the network 120. For example, the processing device 110 obtains data in the storage medium 130 through the network 120. For another example, the processing device 110 may receive and collect the first user speech of the first user terminal 140 through the network 120. For another example, the processing device 110 may push task information created based on the first user's online feedback to the second user terminal 150 through the network 120. In some embodiments, the network 120 may be any one or more of a wired network or a wireless network. For example, the network 120 may include a cable network, an optical fiber network, a telecommunications network, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network (ZigBee), a near field communication (NFC), an in-device bus, an in-device line, a cable connection, etc., or any combination thereof. The network connection between the parts may be in one of the above-mentioned ways or in multiple ways. In some embodiments, the network may be in various topological structures such as point-to-point, shared, central, or a combination of multiple topological structures.

[0051] The storage medium 130 can be used to store data and / or instructions related to the scene 100. In some embodiments, the storage medium 130 can store data and / or information obtained from the first user terminal 140, etc. In some embodiments, the storage medium 130 may include one or more storage components, each of which may be an independent device or a part of other devices. In some embodiments, the storage medium 130 may be set in the processing device 110. In some embodiments, the storage medium 130 may include a random access memory (RAM), a read-only memory (ROM), a large capacity memory, a removable memory, a volatile read-write memory, etc. or any combination thereof. Exemplarily, the large capacity storage may include a magnetic disk, an optical disk, a solid-state disk, etc. In some embodiments, the storage medium 130 may be implemented on a cloud platform.

[0052] The first user terminal 140 may be a terminal used or owned by the first user. The terminal may be a terminal device such as a smart phone, a tablet computer, a laptop computer, a touch screen, a personal computer (PC), a personal digital assistant (PDA), etc. In some embodiments, the terminal may also include a client, which may be a game application client, a web browser client, a new media application client, or an instant messaging client, etc. The first user may use the first user terminal 140 to purchase goods or services through the network 120, or may comment on the purchased goods or services through the first user terminal 140. In some embodiments, the first user terminal 140 may also transmit relevant information and data through the network 120 and other components in the application scenario 100 (e.g., the storage medium 130).

[0053] The second user terminal 150 may be a terminal used or owned by another user group. The user may be a user who uses the feedback information processing results provided in some embodiments of this specification, such as a merchant or a staff member of the merchant. In order to distinguish, it may be called a second user, or a B-end user. Taking the game application scenario as an example, the first user may be a game player, and the second user may be a staff member of the game merchant, such as a game planner, an engineer, etc. Taking the physical commodity sales application scenario as an example, the first user may be a consumer of the physical commodity, and the second user may be a staff member of the merchant who produces the physical commodity, such as a product manager, a designer, etc. Similar to the first user terminal 140, the terminal used or owned by the second user may be a terminal device such as a smart phone, a tablet computer, a laptop, a touch screen, a personal computer (PC, Personal Computer), a personal digital assistant (Personal Digital Assistant, PDA), etc. In some embodiments, the second user's terminal may also include a client, which may be a game application client, a web browser client, a new media application client, or an instant messaging client, etc. The second user can use the second user terminal 150 to provide goods or services through the network 120, and can also obtain feedback from the first user through the network 120 to improve the goods or services. In some embodiments, the second user terminal 150 can also transmit relevant information and data through the network 120 and other components in the application scenario 100 (for example, the storage medium 130).

[0054] The following mainly uses game application scenarios as examples to illustrate some embodiments of this specification, but it should not be understood as limiting the scope of this specification. The feedback information processing method disclosed in the embodiments of this specification can be applied to feedback information processing of any product.

[0055] Figure 2 is an exemplary flow chart of a feedback information processing method according to some embodiments of this specification. In some embodiments, Figure 2 The process 200 shown may be performed by a processing device (such as Figure 1 Specifically, the process 200 may be implemented by a feedback information processing system 1100 deployed on the processing device. Figure 2 As shown, process 200 may include the following steps.

[0056] Step 210 , collecting the first user statement. In some embodiments, step 210 may be implemented by the collection module 1110 .

[0057] The first user statement may be a statement that reflects the first user's remarks. Specifically, the first user statement may be a statement made by the first user at a certain point in time. As an example, game player a posted a comment about game A in a certain game forum, "The picture quality of this game is impeccable, and the skin is relatively cheap among all mobile games." For another example, the first user b posted a short video on a short video platform that analyzes and compares the performance of multiple home appliances. For another example, game player c reported the problem of frequent freezes in game C through the quick feedback interface in the game.

[0058] In some embodiments, one or more first user statements may come from one or more of the following data: online speech, chat data, first user questions, and first user suggestions. Among them, online speech may be speech published by the first user in a forum post bar, and chat data may be speech published by the first user when communicating with other users in an instant messaging application. The first user question or the first user suggestion may be speech published by the first user through a customer service or feedback channel.

[0059] In some embodiments, the first user statement can be obtained from the web page under the specified link through crawler technology. In other embodiments, the first user statement can be obtained from the first user's blog or short video homepage based on pre-set keywords through crawler technology. For another example, the chat log of the first user in the instant messaging application can be obtained, and the first user statement can be extracted from it. Among them, the instant messaging application can be a dedicated chat application, or it can be a chat application integrated in other service platforms, such as an in-game chat application. For another example, the first user statement can be obtained from the customer service log of the platform, etc.

[0060] In some embodiments, the first user statement may be brief, such as "Why is it stuck again?", "The picture quality is terrible", etc., or it may be detailed and complete, such as a first user posting a 400-word game review on a game forum. In some embodiments, the first user statement may be in text form, or in audio, video, etc. In some embodiments, the audio data (such as the audio in a short video) may be processed by speech recognition technology to obtain the first user statement in text form.

[0061] Some embodiments of the present specification can actively collect first user statements from multiple channels such as web pages, new media platforms, chat data, user feedback, etc., so as to timely and comprehensively discover problems or anomalies in the product from user opinions.

[0062] Step 220 , processing the first user statement to extract one or more feedback question descriptions. In some embodiments, step 220 may be implemented by the processing module 1120 .

[0063] The feedback problem description may reflect product-related problems or suggestions. In terms of expression, the feedback problem description may be a short sentence within a preset number of words (such as 10 words, 20 words), or may be called a feedback problem topic. Exemplarily, the feedback problem topic may be "Stone click problem, freeze and inability to click". In other embodiments, the feedback problem description may be a phrase or a problem label. Exemplarily, the problem label may be "download", "fever", etc. In some embodiments, one or more first user statements may be processed to extract feedback problem descriptions with relatively clear and concentrated semantics.

[0064] The collected first user statements may have a large amount of data and contain a lot of invalid effects, such as a lot of content that is irrelevant to the product problem or suggestion. Therefore, before obtaining the feedback problem description, the collected first user data may be filtered to obtain one or more filtered first user statements, and then one or more feedback problem descriptions may be obtained based on the filtered one or more first user statements, so as to improve the accuracy and confidence of the feedback problem description.

[0065] In some embodiments, invalid first user statements may be filtered out based on a whitelist, and then semantically classified using a classification technique to remove first user statements outside of the question category of interest from the classification results.

[0066] Figure 3 is an exemplary flow chart of the steps of extracting feedback question description based on the first user sentence according to some embodiments of this specification. Figure 3 As shown, process 300 may include the following steps.

[0067] Step 310: Filter the first user statement based on the whitelist to obtain the filtered first user statement.

[0068] The whitelist may reflect the semantics of invalid information that is not related to product problems or suggestions, which may be keywords, phrases, or sentences. In some embodiments, the whitelist may be adaptively counted or extracted from the first user's sentence by a processing device, or referred to as a system whitelist. In other embodiments, an artificial whitelist may be obtained by manually setting keywords.

[0069] Taking the system whitelist as an example, it can be considered that when the number of repeated occurrences of the collected first user statement is high, the statement may be relatively conventional or unrelated to product problems or suggestions. Therefore, in some embodiments, the processing device or processing module 1120 can remove the first user statement whose number of repeated occurrences is greater than the first set threshold value from the collected first user statements. As an example only, the collected first user statement a can be obtained, and the text matching can be performed one by one in the remaining collected first user statements. Whenever the same first user statement is found in the remaining first user statements, the number of repeated occurrences of the first user statement a is increased by 1, and so on, until the aforementioned remaining first user statements are traversed. When the number of repeated occurrences of the first user statement a is greater than the first set threshold value, the first user statement a and the remaining first user statements that are the same as it can be removed. Thereafter, the first user statement b can be text matched with the remaining first user statements, and the number of repeated occurrences can be counted. When the number of repeated occurrences of the first user statement b is also greater than the first set threshold value, the first user statement b and the remaining first user statements that are the same as it are removed. In this way, through multiple rounds of matching and removal of the system whitelist, the sentences with a large number of repetitions in the first user's sentences can be filtered out. This method is suitable for the rapid cleaning of a large number of first user sentences with more noise (such as first user sentences from chat data). In specific applications, the text matching conditions can be strengthened (such as setting the text matching conditions to be the same as the first user sentences participating in the matching) to avoid accidental deletion. In this way, the filtering efficiency can be improved while ensuring the filtering accuracy. In some embodiments, the first set threshold can be flexibly set according to actual conditions, such as 500, 1000, 20000, etc., and the first user sentences with a number of repetitions greater than the first set threshold can be further used as a system whitelist and stored. When there is a newly collected first user sentence, the pre-stored system whitelist can be used for text matching to determine whether the new first user sentence should be removed or retained.

[0070] In some embodiments, manually configured preset whitelist keywords can be obtained through the whitelist configuration page, such as the second user can enter one or more preset whitelist keywords and store them on the whitelist configuration page. Exemplary manual whitelists can be "can't beat", "no connection", "no coins have been charged", etc. In some embodiments, the first user statements containing preset whitelist keywords in one or more first user statements can be removed, and / or the first user statements whose semantic similarity with the preset whitelist keywords in one or more first user statements is greater than the second set threshold are removed. As an example, the collected first user statements can be text-matched with the preset whitelist keywords one by one, and when the preset whitelist keywords appear in the first user statement, the statement is removed. For another example, the semantic similarity between the first user statement and the preset whitelist keywords can be calculated, and when the semantic similarity is greater than the second set threshold, the first user statement is removed. As an example only, the first user statement and the preset whitelist keywords can be processed by the natural language processing model respectively, and the corresponding embedding vectors are obtained, and the distance between the corresponding embedding vectors of the two is calculated to determine the semantic similarity between the two, wherein the semantic similarity can be negatively correlated with the distance of the embedding vector. The second set threshold value can be flexibly set according to actual conditions. For example, the second set threshold value can be 80%, 95%, etc.

[0071] In some embodiments, the collected first user sentences can be removed once using a system whitelist, and then the remaining first user sentences can be removed twice using a manual whitelist, thereby obtaining one or more first user sentences after filtering. Furthermore, when performing manual whitelist removal, the first user sentences can be removed by either detecting text inclusion or calculating semantic similarity, or the two can be combined to perform a second removal of the first user sentences. The system whitelist removal method is efficient but has a limited scope of application, and the manual whitelist removal method is more accurate. The combination of the system whitelist and manual whitelist methods can take into account both the efficiency and quality of invalid information removal.

[0072] Step 320, classify the filtered first user sentences based on the first classification model to obtain first user sentences belonging to the target category and the non-target category, and use the first user sentences in the target category as the filtered first user sentences.

[0073] In some embodiments, the target category may be a category that the second user is concerned about or interested in, such as a question category and a suggestion category. The non-target category is a category that the second user is not concerned about. In some embodiments, the first classification model can be obtained through supervised training. Specifically, sample user sentences can be obtained, and classification labels can be marked for these sample user sentences. For example, the classification labels may include question categories, suggestion categories, and other categories. The initial classification model is trained with sample user sentences and their classification labels to obtain a first classification model. During the use stage, the first classification model can process the first user sentence, specifically, it can process the first user sentence after filtering out in step 310, and divide it into a question category, a suggestion category, or other categories. In some embodiments, the initial classification model may be a neural network, a BERT model, a large language model, or the like.

[0074] In the classification labels of the first classification model, target classification labels of interest can be determined, such as the problem category and the suggestion category. The first user statements in these two categories are likely to involve product problem feedback or improvement suggestions, and can be retained as the filtered first user statements, and the first user statements in other categories can be deleted.

[0075] The first user statement filtering step shown in step 310 and step 320 combines the two methods of invalid information filtering and useful information screening to improve the statement filtering efficiency and accuracy.

[0076] Step 330: Based on the filtered first user sentence, obtain one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters.

[0077] In some embodiments, one or more feedback question topics can be obtained based on the filtered one or more first user sentences. In some embodiments, the filtered first user sentences can be processed by a first clustering method to obtain one or more first question feedback clusters and one or more first feedback question descriptions corresponding to the first question feedback clusters; the filtered first user sentences can be processed by a second clustering method to obtain one or more second question feedback clusters and one or more second feedback question descriptions corresponding to the second question feedback clusters; wherein the first clustering method and the second clustering method correspond to different semantic levels and / or semantic classification granularity.

[0078] In some embodiments, the sentences can be divided into one or more semantic levels, the semantics of the upper layer can be the superordinate concept of the semantics of the lower layer, and the semantics of the same layer can include one or more parallel concepts. As an example only, "stuttering" is the first semantic level, and "stuttering of high-definition image quality", "stuttering when collision occurs (in the game)", and "stuttering of voice when connected to the microphone" are three parallel concepts in the second semantic level. In some embodiments, the sentences can be divided into different numbers of semantic levels. For example, the semantic level obtained by processing the filtered first user sentence by the first clustering method can be less than the semantic level obtained by processing the filtered first user sentence by the second clustering method. As an example only, by processing the filtered first user sentence by the first clustering method, one or more question feedback clusters and the first user sentence belonging to each question feedback cluster can be obtained. By processing the filtered first user sentence by the second clustering method, one or more question tags, one or more question feedback clusters belonging to each question tag, and the first user sentence belonging to each question feedback cluster can be obtained; or it can be intuitively understood that the first clustering method corresponds to two semantic levels and the second clustering method corresponds to three semantic levels.

[0079] In some embodiments, the semantics of a sentence can be divided into different categories according to different classification granularities. The semantic classification granularity can be coarse or fine. It can be intuitively understood that the more categories there are, the finer the semantic classification granularity, and vice versa. As an example only, the first clustering method can cluster the filtered first user sentences once to obtain one or more question feedback clusters. The second clustering method can first divide the filtered first user sentences into several categories, and then cluster the first user sentences in each category. It can be intuitively understood that the semantic classification granularity corresponding to the second clustering method is finer than the semantic classification granularity corresponding to the first clustering method.

[0080] In some embodiments, processing the filtered first user sentences by the first clustering method to obtain one or more first problem feedback clusters and one or more first feedback problem descriptions corresponding to the first problem feedback clusters may further include: clustering the filtered one or more first user sentences to obtain one or more first problem feedback clusters; generating cluster summaries of the first problem feedback clusters based on one or more first user sentences in the first problem feedback clusters whose sentence confidences meet the first set condition; and obtaining one or more feedback problem topics based on the cluster summaries of the one or more first problem feedback clusters. Among them, the feedback problem topic may be used as the first feedback problem description.

[0081] In some embodiments, the first user sentence can be directly used as clustering data for clustering. As an example, the first user sentence can be encoded to obtain a sentence vector, and then the sentence vector can be processed using an unsupervised clustering algorithm to obtain one or more first question feedback clusters. Exemplary clustering algorithms can be K-means algorithm, DBSCAN algorithm, Mean-Shift algorithm, etc. In some embodiments, keywords can be extracted from the first user sentence first, and the keywords of the first user sentence can be used as clustering data for clustering. As an example, one or more keywords can be extracted from the first user sentence using an entity naming recognition algorithm, a UIE (Universal Information Extraction) extraction framework, etc. In some embodiments, the extracted keywords can also be identified to distinguish which of the following categories the keywords belong to: product proper nouns, description objects, problem descriptions, etc. Further, the keywords extracted from the same first user sentence can be spliced, and the splicing result can be used as clustering data. Similarly, the keywords of the first user sentence can be encoded to obtain keyword vectors, the keyword vectors in the same first user sentence can be spliced ​​or superimposed, and then the spliced ​​or superimposed vectors can be processed using an unsupervised clustering algorithm to obtain one or more first question feedback clusters.

[0082] In some embodiments, a cluster summary can be determined for the first question feedback cluster. The cluster summary can be a semantic induction or summary of the first user statements in the same question feedback cluster. In some embodiments, one or more first user statements in the first question feedback cluster whose statement confidence meets the first set condition can be processed by a text summary algorithm to generate a cluster summary of the question feedback cluster. Exemplary text summary algorithms can be TextRank algorithms, topic models, large language models, etc. In some embodiments, the first set condition can be the first n first user statements when the first user statements in the first question feedback cluster are arranged in descending order of statement confidence. Exemplarily, n can be 20, 50, 150, etc.

[0083] In some embodiments, a statement confidence may be calculated for the first user statement. The statement confidence may characterize the amount of useful information in the first user statement. As an example, the greater the statement confidence, the higher the possibility that the corresponding first user statement reflects a product problem or suggestion. In some embodiments, one or more keywords of the filtered first user statement may be extracted; the statement confidence of the first user statement is determined based on the keywords of the first user statement and one or more of the following information: the portrait information of the first user related to the first user statement, and the source of the first user statement.

[0084] More information about extracting keywords of the first user's sentence can be found in the above description of clustering based on keywords of the first user's sentence, which will not be repeated here. As an example, when more keywords of product proper nouns are extracted from the first user's sentence, the sentence confidence is higher, or when more keywords describing objects are extracted from the first user's sentence, the sentence confidence is higher, or when more keywords describing problems are extracted from the first user's sentence, the sentence confidence is higher.

[0085] The portrait information of the first user can be represented by the specific values ​​of one or more portrait tags. One or more portrait tags can describe the first user from different dimensions. For example, the portrait tags may include one or more of the following: online time, recent activity, payment ability, etc. The specific values ​​of different portrait tags may correspond to different individual scores. In some embodiments, the portrait tag of the first user and its corresponding individual score can be determined. The individual scores of the portrait tags of the first user are accumulated to obtain the score corresponding to the portrait information of the first user. In some embodiments, the statement confidence of the first user's statement is positively correlated with the score corresponding to the portrait information of the first user.

[0086] In some embodiments, the source score may be set according to the credibility or authority of the source of the first user's statement. Exemplarily, the source scores of chat data, online speech, and user feedback (such as the first user's question and / or the first user's suggestion) may be increased in sequence. In some embodiments, the statement confidence of the first user's statement is positively correlated with its source score.

[0087] In some embodiments, the sentence confidence of the first user's sentence can be determined based on the keyword score of the first user's sentence and the first user's portrait score. In other embodiments, the sentence confidence of the first user's sentence can also be determined based on the keyword score of the first user's sentence, the first user's portrait score, and the source score of the first user's sentence.

[0088] In some embodiments, a cluster summary may be calculated for each first question feedback cluster obtained by clustering, thereby obtaining cluster summaries of one or more first question feedback clusters and using them as one or more question feedback topics.

[0089] In some embodiments, processing the filtered first user sentences by a second clustering method to obtain one or more second problem feedback clusters and one or more second feedback problem descriptions corresponding to the second problem feedback clusters may further include: classifying the filtered one or more first user sentences based on a second classification model to obtain first user sentences belonging to one or more question tags; obtaining one or more second feedback problem descriptions based on one or more question tags; clustering the first user sentences belonging to the same question tag to obtain one or more second problem feedback clusters belonging to the question tag; generating a cluster summary of the second problem feedback cluster based on one or more first user sentences in the second problem feedback cluster whose sentence confidence meets the second set condition.

[0090] In some embodiments, the question label can be used as the second feedback question description. In some embodiments, the filtered first user statement can be classified in a more fine-grained manner to obtain its corresponding question label. Specifically, a second classification model can be used for classification. In some embodiments, the second classification model can be obtained through supervised training. Specifically, sample user statements can be obtained, and classification labels can be marked for these sample user statements. For example, the classification labels can include multiple question labels. The initial classification model is trained with sample user statements and their classification labels to obtain a second classification model. During the use stage, the second classification model can process the first user statement, specifically, it can process the first user statement after filtering in step 310 and step 320, and divide it into more than one question label. In some embodiments, the initial classification model can be a neural network, a BERT model, a large language model, etc., and the question labels can be manually labeled based on the product problems or suggestions obtained by the current business.

[0091] In some embodiments, the second classification model may process the first user statement to determine its classification score. As an example, the classification score of the first user statement may be a value in the range of [0,100]. Further, based on the classification score and the score range corresponding to one or more question tags, the filtered one or more first user statements may be classified to obtain first user statements belonging to one or more question tags. Among them, the score range corresponding to one or more question tags is adjustable. As an example, the score range corresponding to question tag 1 may be pre-configured to be [0,10], the score range corresponding to question tag 2 may be [10,35], and so on. In some embodiments, the score range corresponding to each question tag may be adjusted according to actual needs in the AI ​​recall strength configuration page to adjust the recall strength of the question tag. For example, the score range of a question tag may be narrowed to reduce its recall strength, so that the first user statement is less likely to be classified under the question tag. For another example, the score range of a question tag may be expanded to increase its recall strength, so that the first user statement is more easily classified under the question tag.

[0092] In some embodiments, in addition to the training labels from the second classification model, the question labels may also come from keywords extracted from the first user sentences (or simply referred to as extracted keywords). Specifically, after some first user sentences are processed by the second classification model, they may be difficult to be classified into existing training labels. In this case, classification may be performed based on the extracted keywords of these first user sentences that cannot be classified. As an example, the extracted keywords of the first user sentences that cannot be classified may be counted by word frequency (such as the number of occurrences), and the extracted keywords with a larger word frequency may be used as question labels, and the first user sentences containing the extracted keywords may be classified under the question labels.

[0093] In some embodiments, the first user statements belonging to the same question label can be clustered and abstracted according to the clustering method described above, thereby obtaining one or more second question feedback clusters and their cluster summaries belonging to each question label. More specifically, the setting conditions used in this embodiment to determine the first user statements involved in the summary calculation can be the same as or different from the aforementioned first setting conditions. For ease of distinction, the aforementioned setting conditions used in this embodiment are referred to as second setting conditions. Exemplarily, the second setting condition can be the first m first user statements in the question feedback cluster when the first user statements are arranged in descending order of statement confidence, where m can be 3, 5, 10, etc.

[0094] Some embodiments of the present specification process the filtered first user sentences separately by clustering them in accordance with different semantic levels and / or semantic classification granularities, so as to obtain feedback problem descriptions displayed at different semantic levels or semantic granularities, thereby meeting the semantic needs of different second users to view the feedback information of the first user.

[0095] Step 340 , generating a cluster summary of the question feedback cluster, and removing the first user sentence corresponding to the question feedback cluster whose cluster summary does not belong to the target category.

[0096] In some embodiments, the processing device or processing module 1120 may also perform secondary filtering on the clustered first user statement. Specifically, web page data may be obtained through a search engine based on the cluster summary of the problem feedback cluster obtained by clustering, and the cluster summary of the problem feedback cluster may be determined based on the web page data through a second language model to determine whether it belongs to the target category; when the cluster summary of the problem feedback cluster does not belong to the target category, the problem feedback cluster and its corresponding first user statement are removed. More instructions on generating the cluster summary of the problem feedback cluster can be found in step 330, which will not be repeated here. In some embodiments, the cluster summary of each problem feedback cluster generated in step 330 may be judged to determine whether it belongs to the target category. Among them, the target category may be a category that the second user is concerned about or interested in, such as a problem category and / or a suggestion category. As an example, the search engine API may be called through AI Agent (artificial intelligence agent, an intelligent entity that can autonomously perceive the environment, make decisions and perform tasks) technology, and the search engine may be searched based on the cluster summary of the problem feedback cluster to obtain web page data. Furthermore, the AI ​​Agent can transmit the web page data to the second language model, allowing the second language model to determine whether the corresponding cluster summary is related to the product problem or suggestion based on the web page data. In some embodiments, the second language model can be a large language model such as GPT, Qwen4B, RoBerta, etc.

[0097] As an example, the cluster summary of a problem feedback cluster is "Wo Tu is difficult and cannot be upgraded". The AI ​​Agent can search for web page data related to Wo Tu through network retrieval. The second language model can determine based on the searched web page data that Wo Tu is difficult and cannot be upgraded. It is not a defect of the game product, but a state of the game player. Then the cluster and its first user statement can be deleted. In this way, the clustering results can be filtered twice, and the first user statements that are not related to product problems or suggestions can be removed from a deeper semantic level, thereby improving the accuracy of the feedback problem description and avoiding the subsequent waste of task processing resources.

[0098] Considering that the collected first user sentences are of different lengths, the feedback problem description obtained by the same processing method may be unstable in accuracy. Alternatively, for a relatively lengthy first user sentence, if a semantic extraction can be performed before classification or clustering, it is helpful to improve the accuracy of clustering or classification. In view of this, in some optional embodiments, before obtaining one or more feedback problem descriptions based on one or more filtered first user sentences (such as before step 330), the processing device or processing module 1120 can also process the first user sentence with a word count greater than the third set threshold in the filtered one or more first user sentences based on the first language model to obtain a first user sentence with a word count not greater than the fourth set threshold; wherein the fourth set threshold is less than or equal to the third set threshold. In other words, the lengthy first user sentence can be rewritten by the first language model to shorten it. In some embodiments, the first language model can be a large language model such as GPT, Qwen4B, RoBerta, etc. In some embodiments, the first language model can be fine-tuned based on the sample user sentence to make it more suitable for rewriting long sentences in a specific application scenario into short sentences. In some embodiments, the third set threshold and the fourth set threshold can be set according to actual conditions. As an example, the third set threshold can be 20, 50, 100, etc., and the fourth set threshold can be 5, 10, 15, 20, etc.

[0099] Step 230 : Present at least part of the one or more feedback problem descriptions through one or more feedback data pages. In some embodiments, step 230 may be implemented by the display module 1130 .

[0100] In some embodiments, the feedback problem description obtained through step 220 can be presented through a feedback data page, specifically, can be presented through one or more feedback data pages in the feedback information processing system. As mentioned above, the feedback problem description can be a first feedback problem description (such as a feedback problem topic) or a second feedback problem description (such as a problem tag). In some embodiments, the feedback problem topic obtained through processing can be presented through the first feedback data page, and the problem tag obtained through processing can be presented through the second feedback data page.

[0101] Figure 4 is a schematic diagram of a feedback data page according to some embodiments of this specification. Figure 4 The illustrated page presents one or more first feedback question descriptions (eg, feedback question topics). Figure 4 "XXXX1" and "XXXX2" in the table represent feedback question topic 1 and feedback question topic 2 respectively. Each feedback question topic has an expansion control, such as Figure 4The expansion control may be triggered, and the triggering operation may include, for example, single-clicking, double-clicking, etc. In some embodiments, in response to the triggering operation of the expansion control, the display module 1130 may further present the first user statement in the first question feedback cluster of the feedback question topic corresponding to the expansion control. Figure 4 Taking the feedback question topic 1 in the example, when its corresponding expansion control is triggered, the first user sentence "1.xxxxxx", "2.xxxxxx", etc. in the first question feedback cluster corresponding to the topic can be expanded and presented below the feedback question topic 1. In some embodiments, the form of the expansion control can change to indicate whether the current first feedback question topic is in an expanded state or a collapsed state. For example, when the feedback question topic is expanded, Figure 4 The triangle symbol in the feedback question topic is facing upwards, and when the feedback question topic is collapsed, the triangle symbol is facing downwards. In some embodiments, the expansion control corresponding to the feedback question topic 1 can also be triggered again to collapse and hide the expanded first user statement.

[0102] Figure 5 is a schematic diagram of a feedback data page according to other embodiments of this specification. Figure 5 The illustrated page presents one or more second feedback question descriptions (such as question tags). Figure 5 "XX1" and "XX2" in the question represent question label 1 and question label 2 respectively. Each question label has an expansion control, such as Figure 5 The triangle symbol in the middle. The expansion control can be triggered, and illustratively, the triggering operation can include single-clicking, double-clicking, etc. In some embodiments, in response to the triggering operation of the expansion control, the display module 1130 can further present the cluster summaries of one or more second question feedback clusters under the question label corresponding to the expansion control. For example, when the expansion control corresponding to question label 1 is triggered, the cluster summary "XXXX1" of the second question feedback cluster 1 under the question label and the cluster summary "XXXX2" of the second question feedback cluster 2 can be expanded and presented below the question label 1. In some embodiments, the cluster summary of each second question feedback cluster can be further expanded to display the first user statement in the corresponding second question feedback cluster. As Figure 5 As shown, the cluster summary "XXXX1" of the second problem feedback cluster 1 also presents the first user statements "1.xxxxxx", "2.xxxxxx", etc. under the cluster, and the cluster summary "XXXX2" of the second problem feedback cluster 2 also presents the first user statements "1.xxxxxx", "2.xxxxxx", etc. under the cluster. In some other embodiments, in response to Figure 5In response to the triggering operation of the expansion control in the display module 1130, the display module 1130 may further present one or more first user statements under the question tag corresponding to the expansion control. Figure 5 , when the expansion control corresponding to question label 1 is triggered, the first user statement "1.xxxxxx", "2.xxxxxx", etc. of the second question feedback cluster 1 and the first user statement "1.xxxxxx", "2.xxxxxx", etc. of the second question feedback cluster 2 can be expanded directly below question label 1 "XX1", and the cluster summaries "XXXX1" and "XXXX2" in the figure are no longer presented. Figure 5 Similar to the expansion control shown, Figure 5 The shape of the expansion control shown may also change to indicate whether the current feedback question topic is in an expanded state or a collapsed state.

[0103] It should be understood that Figure 4 and Figure 5 This is only an example of a feedback data page. In some other embodiments, additional information items or controls may be added to the page based on actual needs. Figure 4 The page shown may also present relevant keywords, feedback volume, feedback time information, etc. corresponding to each feedback question topic.

[0104] In some embodiments, the number of feedback question topics or question tags is relatively large, and only a portion of the processed feedback question topics or a portion of the question tags may be presented in the corresponding feedback data page.

[0105] In some embodiments, hot topics can be selected from the feedback problem topics according to preset rules to be presented on the first feedback data page. As an example, the cluster confidence can be calculated for the first problem feedback clusters corresponding to each feedback problem topic, and then several first problem feedback clusters with larger cluster confidence can be selected, and their cluster summaries can be presented as hot topics on the first feedback data page. In other embodiments, the feedback problem topics corresponding to each first problem feedback cluster can be divided into hot topics and ordinary feedback based on the cluster confidence of each first problem feedback cluster, and hot topics or ordinary feedback or both types of feedback problem topics can be presented in full on the first feedback data page according to the needs of the second user. The following is an example of cluster confidence calculation for the first problem feedback cluster. Taking the first problem feedback cluster 1 as an example, several (such as the first 10) first user statements with larger statement confidence in the first problem feedback cluster 1 are determined, and the average statement confidence of the aforementioned several first user statements is calculated. The feedback amount of the first problem feedback cluster 1 is counted, and specifically, the total number of first user statements under the cluster can be counted. The cluster confidence of the first question feedback cluster 1 is determined based on the average sentence confidence and the feedback amount. For example, the cluster confidence is positively correlated with the average sentence confidence and the feedback amount.

[0106] In some embodiments, hot keywords can be selected from the question tags according to preset rules to be presented on the second feedback data page. As an example, cluster confidences can be calculated for the second question feedback clusters under each question tag, and then several second question feedback clusters with larger cluster confidences can be selected, and the question tags to which they belong are presented as hot keywords on the second feedback data page.

[0107] In some embodiments, the first user statement obtained after filtering can also be presented through a third feedback data page. The third feedback data page can also be provided with a search control (such as a search term input box and a query button). The second user can operate the search control to directly search and view the first user statement after filtering.

[0108] In some embodiments, the first user statement from the key focus group in the first user statement obtained after filtering can also be presented through the fourth feedback data page. In some embodiments, the key focus group can be the first user who meets the preset conditions (such as a high consumption amount, a significant influence on the future trend of the product, etc.). In some embodiments, the second user can also configure the key focus group through the key focus group configuration page. In some embodiments, the fourth feedback data page can also present a reply control, and the second user can use the reply control to send a description or handling of the question or suggestion to the first user in one or more key focus groups. More descriptions about the configuration of the key focus group and replying to the first user can also be found in the following text, which will not be repeated here.

[0109] Step 240 , in response to a feedback question follow-up request initiated through at least one of the one or more feedback data pages, create a task item and send it to a second user terminal associated with the task item. In some embodiments, step 240 may be implemented by task module 1140 .

[0110] In order to handle product-related problems or suggestions in a timely manner, in some embodiments of this specification, when the second user views the feedback problem description through the feedback data page, he can directly initiate a feedback problem follow-up request through the feedback data page to promptly follow up and handle more important or urgent problems or suggestions. In some embodiments, a feedback problem follow-up control corresponding to the feedback problem description can also be presented on the feedback data page; in response to the triggering operation of the feedback problem follow-up control in the feedback data page, a task item can be created based on the feedback problem description corresponding to the feedback problem follow-up control.

[0111] As an example, Figure 4In the page shown, a "question follow-up" control may be presented, which corresponds to the topic of the feedback question one by one. Figure 5 In the page shown, the "follow up on the problem" control can correspond to the cluster summary under the problem label one by one. The triggering operation of the "follow up on the problem" control can include a single click or a double click, etc. In response to the triggering operation of the "follow up on the problem" control, a task editing window can be further displayed, and the second user can fill in the task information related to the task item through the window. For example, the second user can fill in the task name, task description, information of the second user related to the task item (such as name, work number, email address, etc.), task progress (such as the specific link where the current task item is located), etc. in the window, wherein the second user related to the task item can be the second user who follows up or processes the task item. In some embodiments, the task editing window can also present one or more first user statements under the corresponding feedback problem description, and the second user can add or delete these first user statements, or do not process them. When the second user completes the task editing, click the send or submit control in the task editing window to complete the creation of the task item, and send the task item to the second user terminal of the second user related to the task item.

[0112] The second user terminal can receive task item reminder information, which can be a reminder message from the feedback information processing system, a reminder message from an order tracking system other than the feedback information processing system, or a reminder message from a work communication system, so that the second user related to the task item can receive the task item information in a timely manner.

[0113] In some embodiments, the display module 1130 can present the information of the task item through the task item page. Specifically, the information of the task item can be presented through the task item page in the feedback information processing system. The second user can log in to the feedback information processing system described in some embodiments of this specification through the account password, and view the information of the task item on its private task item page in the system. The second user can also view the public task item page in the system in the feedback information processing system, and the page contains the information of all the task items in the system. In some embodiments, the information of the task item presented on the task item page can be one or more of the task item information edited when creating the task item. Figure 6 1 is a schematic diagram of a task item page according to some embodiments of this specification. Figure 6 As shown, the task item page presents task items with task names "TTT1" and "TTT2", and for each task item, relevant second user information (specifically, name, work number, email address, etc.) and task progress are also presented.

[0114] In some embodiments, the processing device or display module 1130 may also adjust the task progress by presenting a task item page with a control (such as Figure 6 In some embodiments, in response to the triggering operation of adjusting the task progress control, the processing device or display module 1130 may further present a task processing page, and the task processing page may present a task progress bar. The second user may enter or select the process corresponding to the current task item in the task progress bar. An exemplary task item process may be a problem verification stage, a problem handling stage, or completed. In some embodiments, the task processing page may also present a feedback problem description and a reply control corresponding to the task item. Corresponding to the feedback problem description, the task processing page may further present one or more first user statements related thereto and their corresponding first user information. Exemplarily, the first user information may include the identity identification, portrait information, etc. of the first user. The second user may trigger the reply control to reply to the problem or suggested processing status of one or more first users related to the feedback problem description. In some embodiments, one or more first user statements related to the feedback problem description may be the first user statements in the clustering or classification results corresponding to the feedback problem description, and accordingly, one or more first users related to the feedback problem description may be the speakers of the first user statements in the clustering or classification results corresponding to the feedback problem description. The processing device or task module 1140 may, in response to the triggering operation of the reply control, obtain the reply content and send it to the first user terminal corresponding to at least part of the first user statements in the one or more first user statements related to the feedback problem description. As an example, in response to the triggering operation of the reply control, a reply content editing window may be further presented, in which the second user may edit the reply information, and the second user may also select one or more first user statements presented, and then select one or more first users to be replied. When the second user completes the reply content editing and determines the first user to be replied, the reply content may be sent to the first user terminal of the selected first user by clicking the send control in the reply content editing window. The first user may view the reply content at the first user terminal. Taking the game application scenario as an example, the reply content may be presented to the first user in the form of a customer service dialogue in the game application of the first user.

[0115] In some embodiments, in order to accurately and fully handle the problem or suggestion, the feedback information processing system can provide a more complete verification mechanism for the second user. Before processing the task item, the second user can first investigate the problem or suggestion reflected in the feedback problem description, and then determine a better solution. In view of this, in some embodiments, the processing device or display module 1130 can present a verification control corresponding to one or more first user statements related to the feedback problem description on the task item page. Figure 6 Each task item has an expansion control corresponding to it. In response to the triggering operation of the expansion control, one or more first user statements "1.xxxxxx", "2.xxxxxx", etc. corresponding to the task item can be expanded and presented. A verification control can be presented after each first user statement, such as Figure 6The magnifying glass icon shown. In response to the triggering operation of the verification control, the source data of the first user statement corresponding to the verification control is displayed. As an example, when the first user statement comes from a quick feedback channel, the source data displayed may be a feedback log or a customer service conversation log containing the first user statement; when the first user statement comes from a web page post bar, the source data displayed may be web page data, and the verification control may be associated with a web page link, and when the second user clicks the verification control, the web page corresponding to the web page link may be directly jumped to in the browser; when the first user statement comes from chat data, the source data displayed may be a chat conversation content or a chat log containing the first user statement. In some embodiments, in response to the triggering operation of the verification control, the image data of the first user statement corresponding to the verification control may also be displayed. As an example, the first user may reflect the problems or suggestions of the product to the customer service of the product merchant through the quick feedback channel, and may provide image data such as pictures and videos as evidence. When acquiring the first user statement, the feedback information processing system may also acquire and store the image data associated with it. Further, in response to the triggering operation of the verification control, the image data related to the first user statement may be presented to the second user. In some embodiments, in response to the triggering operation of the verification control, the relevant data of the first user statement corresponding to the verification control in the verification system can also be displayed. The verification system can be a system for verification independent of the feedback information processing system, and the data related to the first user or his speech statement can be stored and managed in the verification system. In response to the triggering operation of the verification control, the first user information and the speaking time of the first user statement can be retrieved in the verification system as query conditions, and then the data related to the first user statement recorded in the verification system can be obtained. In some other embodiments, the processing device or display module 1130 can also present the verification control corresponding to the task item or the feedback problem description corresponding to the task item in the task item page, and in response to the triggering operation of the verification control, display the portrait information of the first user corresponding to the task item or the feedback problem description corresponding to the verification control. The first user corresponding to the task item or feedback problem description can be the first user corresponding to the first user statement contained in the classification or clustering result corresponding to the task item or feedback problem description. As an example, the distribution information of different portrait label values ​​of the first user corresponding to the aforementioned task item or feedback problem description under one or more portrait labels can be presented. Taking the online time portrait label as an example, the online time can be divided into four levels: new user, low frequency, leisure, and loyal. The distribution ratio of new user, low frequency, leisure, and loyal under the online time portrait label of the first user corresponding to a task item or feedback problem description can be further presented. The first user portrait information of the task item or feedback problem description can help the second user quickly determine the importance of the problem or suggestion or the root cause, thereby improving the efficiency of problem solving.

[0116] In some embodiments, different icons may be used to represent verification controls. For example, verification controls related to data sources may be represented by icons. Figure 6 The link icon shown in the figure indicates that the verification control involving image data can be represented by a rectangular icon, the verification control involving cross-system verification can be represented by a magnifying glass, the verification control at the task item or feedback problem description level can be represented by a pie icon, and so on. In some embodiments, the verification control can also be presented on the feedback data page. That is to say, the second user can also verify the first user's statement when browsing the feedback data page. As an example, the corresponding verification control can be presented at the first user's statement or the feedback problem description on the feedback data page. The relevant description of the verification control in the feedback data page and the corresponding verification content can be found in the figure below. Figure 6 The verification controls in the relevant instructions are consistent with the verification content and will not be repeated here.

[0117] In addition to presenting feedback information through a page, some embodiments of this specification also establish an alarm mechanism to actively push alarm information to related second users to ensure that questions or suggestions can receive more timely responses. In some embodiments, when at least part of the feedback problem descriptions in one or more feedback problem descriptions meet the preset alarm conditions corresponding to one or more alarm types, an alarm message is sent to one or more second user terminals through an alarm window. In some embodiments, this step can be implemented by the alarm module 1150.

[0118] In some embodiments, the alarm types may include the emergence of new feedback problem descriptions, a sharp increase in the number of first user statements related to the feedback problem descriptions, the feedback problem descriptions being related to the first user who is the focus of attention, regular broadcasting of lists, the feedback problem descriptions involving focus content, etc.

[0119] In some embodiments, the appearance of a new feedback problem description means that after the newly collected first user statement is processed by the aforementioned steps 210 and 220, a feedback problem description that has not appeared before is obtained. Specifically, it may be that a new feedback problem topic or a new problem tag has appeared. The preset alarm conditions of this type of alarm can be constrained from both time and feedback volume. In some embodiments, the feedback volume can be the number of first user statements under a certain feedback problem topic or a certain problem tag. As an example, the preset alarm condition may be that the problem tag has not appeared in the last n days, and the feedback volume is not less than m. For example, the preset alarm condition may be a hot topic with a feedback volume of not less than m, and the semantics are not similar to the feedback problem topic in the last n days, etc. Among them, n and m can be flexibly set as needed.

[0120] In some embodiments, a sharp increase in the number of first user statements related to the feedback problem description means that the rate of increase in the number of first user statements under the feedback problem description exceeds the historical period. The preset alarm conditions of this type of alarm can be constrained from aspects such as the historical period and the rate of increase. As an example, the preset alarm condition may be that the amount of feedback on the topic of the feedback problem on the day is greater than m, and is j times the amount of feedback on the topic of the similar feedback problem on the previous day. For another example, the preset alarm condition may be that the amount of feedback on the question tag in the most recent preset period is greater than m, and is k1 times and k2 times higher than the previous two periods, respectively, where the preset period, m, j, k1, and k2 can be flexibly set as needed. In some embodiments, the preset alarm condition compares the feedback amount change trend of the past two time periods to avoid abnormal changes in the feedback amount caused by specific events (such as product activities, version releases), and improve the accuracy of the alarm.

[0121] In some embodiments, the feedback problem description related to the first user of focus may refer to the feedback problem topic or question tag containing the first user statement from the first user of focus. In some embodiments, the second user can pre-configure the first user of focus. As an example, the second user can configure the first user of interest or importance on the configuration page of the focus group. Specifically, the second user can create a new group name on the configuration page of the focus group, and then describe the characteristics of the group from the portrait tag dimension. After the configuration is completed, the processing device or the feedback information processing system can determine whether the first user statement comes from the pre-configured focus group through the portrait information of the first user of the first user statement. In some embodiments, the preset alarm condition corresponding to this type of alarm type can be that the feedback problem topic or question tag contains the first user statement from the focus group within a preset time period, and the feedback amount is not less than m. Among them, the preset time period and m can be flexibly configured as needed.

[0122] In some embodiments, the timed broadcast of the list is to broadcast the list of hot topics or hot keywords at a specified point. As an example, the preset alarm condition can be to broadcast the top j hot topics or hot keywords once at a certain interval or at a preset time node. Among them, the interval time, time node and j can be flexibly set as needed.

[0123] In some embodiments, the feedback problem description involving key focus content may refer to the first user sentence containing the key focus content. In some embodiments, the second user can pre-configure the key focus content. As an example, the second user can configure the key focus content's vocabulary, short sentences, etc. on the configuration page of the key focus content. After completing the configuration, the processing device or the feedback information processing system can determine whether the first user sentence contains the key focus content by text matching in the collected first user sentence or the filtered first user sentence. In some embodiments, the preset alarm condition corresponding to this type of alarm type can be that the feedback problem topic or question tag contains the key focus content within a preset time period, and the feedback amount is not less than m. Among them, the preset time period and m can be flexibly configured as needed.

[0124] In some embodiments, the alarm type and its preset alarm conditions can be pre-configured as needed. For example, the second user can configure the alarm conditions corresponding to the aforementioned alarm type in the alarm configuration page of the feedback information system, or can configure a new alarm type and corresponding alarm conditions according to actual needs.

[0125] In some embodiments, the alarm information presented in the alarm window includes one or more of the following information: the alarm type, the feedback problem description related to the alarm information, one or more first user statements related to the feedback problem description, or one or more of the following information: the alarm type, the focus content related to the alarm information, and one or more first user statements containing the focus content related to the alarm information. In some embodiments, the alarm window can be a pop-up window in the feedback information system, or it can be other systems outside the feedback information system, such as a message card of the work communication system. The work communication system can be a chat tool, through which staff or second users can conveniently communicate at work, deploy tasks, and follow up. Message cards are a way of presenting information in the work communication system. Message cards can be pushed to specific second users individually, or they can be published in a chat group and received by multiple second users at the same time.

[0126] In some embodiments, the alarm window may also include a feedback question follow-up control; in response to the triggering operation of the feedback question follow-up control, the processing device or task module 1140 may create a task item based on the alarm information and send it to the second user terminal associated with the task item. The task item created through the alarm window can also be viewed in the task item page of the feedback information processing system.

[0127] Figure 7 This is a schematic diagram of the first alarm window shown in some embodiments of this specification. Figure 7The first alarm window shown may include the words of the alarm type "XXX!", the feedback problem description and the related first user statement. In some embodiments, the feedback problem follow-up control in the alarm window may further include a second user information input box 701 and a send control. The second user who receives the alarm window can fill in the second user information (such as the second user's email address, the account in the work communication system, etc.) of the follow-up or processing task item in the second user information input box 701. After clicking the send control, the processing device or task module 1140 will create a task item and send the task item to the corresponding second user end. In some embodiments, the specific method of creating a task item from the alarm window is similar to the process of creating a task item from the feedback data page, which will not be repeated here. In other embodiments, the method of creating a task item from the alarm window can be more simplified, such as in response to the triggering operation of the feedback problem follow-up control in the alarm window, the task module 1140 can automatically collect and fill in the relevant information of the task item to complete the creation of the task item.

[0128] Continue to refer Figure 7 , the alarm window can also display an ignore control at the same time. When the second user determines that the alarm information has little impact, the ignore control can be triggered to set the task process of the task item to completed or ignored. In some embodiments, the processing device or task module 1140 can also switch the display properties of the alarm window to intuitively indicate the task process of the task item. For example, when the display property is in the first state, it indicates that the task process is in processing, when the display property is in the second state, it indicates that the task process is completed, and when the display property is in the third state, it indicates that the task process is ignored. The display properties may include properties such as the shape, color, and shading of the window. As an example, before the task item is created, the color of area 702 in the alarm window is blue. After the task item is created, the color of area 702 can be switched to orange (first state). When the alarm information is ignored, the color of area 702 can be switched to gray (third state).

[0129] Figure 8 is a schematic diagram of a second alarm window according to some embodiments of this specification. In some embodiments, Figure 8 The second alarm window shown can be Figure 7 As an example, when the first alarm window is switched to Figure 7 After the first alarm window shown in the figure creates a task item, the processing device or task module 1140 can make the alarm window appear Figure 8 Status shown. Figure 8 The area 802 in the window is orange, indicating that the task process is in progress. Figure 8 The area 802 shown may also present second user information related to the task item (such as "Zhang San") to indicate who is following up or handling the current task item. Figure 8 The second user information input box 801 and the send control can continue to be presented in the window shown. Figure 8 After filling in the new second user information in the second user information input box 801 and clicking the send control, a new follow-up or handler can be added to the current task item. Figure 8 As shown, the alarm window also displays a processing control and a completion control. In response to the triggering operation of the processing control, the processing device or task module 1140 can obtain the second user information corresponding to the second user end that initiated the triggering operation and add it as a new follow-up or handler of the current task item. In response to the triggering operation of the completion control, the processing device or task module 1140 can modify the task progress of the task item to completed. It should be understood that in some embodiments, Figure 8 The processing controls and completion controls shown may also be presented alternatively.

[0130] Fig. 9 is a schematic diagram of a third alarm window according to some embodiments of this specification. In some embodiments, Fig. 9 The third alarm window shown can be Figure 8 As an example, when the first alarm window is switched to Figure 8 After adding a handler or follow-up person, or when the second user information input box and the sending control of the second alarm window are Figure 8 After adding a handler or follow-up person to the processing control of the second alarm window shown in the figure, the alarm window can present Fig. 9 Status shown. Fig. 9 The window shown is the same as Figure 8 The main difference is Fig. 9 The area 902 in the window shown will additionally display the newly added second user information (such as "Li Si").

[0131] Fig.10 is a schematic diagram of a fourth alarm window according to some embodiments of this specification. In some embodiments, Fig.10 The fourth alarm window shown can be Figure 8 or Fig. 9 As an example, when the alarm is triggered Figure 8 (or Fig. 9 ) After completing the control of the alarm window shown in the figure, the alarm window can be presented Fig.10 Status shown. Fig.10 The color of area 1002 in the window shown changes to green, indicating that the current task item has been completed.

[0132] In some embodiments, while sending the alarm information through the alarm window, the processing device or task module 1140 can automatically create a task item through the alarm window. The processing device or task module 1140 can automatically collect and fill in the relevant information of the task item based on the alarm information, and can determine the second user associated with the current task item based on the correspondence between the preset question keywords and the second user, and directly send the task item to the second user terminal of the second user. At this time, the alarm window received by the second user can be directly Figure 8 Afterwards, the second user can interact with the controls in the window to obtain Fig. 9 or Fig.10 In some embodiments, the correspondence between the problem keyword and the second user can be pre-set through the on-duty person configuration page in the feedback information system. When the feedback problem description or the first user statement related to the alarm information includes the problem keyword, the follow-up person or the processor of the task item created based on the alarm information can be set as the second user corresponding to the aforementioned problem keyword.

[0133] In some embodiments, the alarm window can be presented in a chat interface. Among them, the chat interface can be an interface in an instant messaging application, and users can use the chat interface to perform instant messaging. As an example only, the chat interface may include an information input area and a chat content display area, and the user (specifically, the second user) can input the text, voice, image and other data to be sent through the information input area, and the chat content display area can present the data sent by the user and the data sent from other users participating in the chat. In some embodiments, the chat interface may be a private message interface or a group chat interface. As an example, the alarm module 1150 may send the alarm window to a specified chat interface, such as a group chat interface, through the interface of the instant messaging application or through a pre-registered chat account. In some embodiments, the group chat interface may be an interface for chatting with more than three second users, and the second user may view the alarm window in the chat content display area of ​​the group chat interface and interact through the controls in the alarm window to follow up on the corresponding task items.

[0134] In some embodiments of this specification, while presenting alarm information through an alarm window, it is also possible to conveniently and quickly create task items related to the alarm information and update the task progress of the task items, so that the alarm information can be followed up in time. The alarm window can be presented in the chat interface, so that relevant staff can understand the alarm information and problem follow-up in time.

[0135] In some embodiments, the priority of the feedback problem description can also be configured. In response to the priority configuration operation corresponding to one or more feedback problem descriptions, the processing device or the alarm module 1150 can set the priority of the corresponding feedback problem description. For example, the second user can adjust the priority of each problem label to high, normal or low through the priority configuration page in the feedback information system. When the alarm is triggered, the alarm module 1150 will detect whether the priority condition of the alarm type is met in addition to detecting whether the preset alarm condition of the alarm type is met. The priority condition corresponding to the alarm type can be preset. Taking the alarm type related to the feedback problem description and the first user who focuses on it as an example, the alarm condition of the alarm type can be pre-configured, and the priority condition can be set to high. When the feedback problem description with a normal or low priority meets the alarm condition, the alarm module 1150 will not send an alarm message through the alarm window. In this way, the second user can control the priority of the alarm according to the actual situation to reduce false alarms.

[0136] In some embodiments, the second user can also modify the priority of the feedback problem description by clicking on it on the feedback data page. As an example, in the second feedback data page, each problem tag is presented with a click control. After the second user clicks on the click control, the priority of the corresponding problem tag will be lowered by one level, such as from normal to low level, or from high level to normal. In some embodiments, in response to the triggering operation of the click control, an inquiry window can also be displayed to guide the second user to fill in the reason for the click, so as to optimize or improve the feedback information processing system. For example, when a second user believes that a problem tag is not closely related to a product problem or suggestion, the feedback information processing system can remove the problem tag in the subsequent classification process.

[0137] Fig.11 is an exemplary block diagram of a feedback information processing system according to some embodiments of this specification. Fig.11 As shown, the feedback information processing system 1100 may include a collection module 1110 , a processing module 1120 , a display module 1130 , and a task module 1140 . In some optional embodiments, the feedback information processing system 1100 may further include an alarm module 1150 .

[0138] The collection module 1110 is used to collect first user sentences, where the first user sentences include sentences reflecting the first user's remarks.

[0139] The processing module 1120 is used to process the first user statement and extract one or more feedback question descriptions.

[0140] The display module 1130 is configured to present at least a portion of the one or more feedback problem descriptions through one or more feedback data pages.

[0141] The task module 1140 is used to create a task item and send it to a second user terminal associated with the task item in response to a feedback question follow-up request initiated through at least one feedback data page of the one or more feedback data pages.

[0142] In some optional embodiments, the alarm module 1150 is used to send an alarm message to one or more second user terminals through an alarm window when at least part of the one or more feedback problem descriptions meet preset alarm conditions corresponding to one or more alarm types; or to send an alarm message to one or more second user terminals through an alarm window when at least part of the one or more first user statements meet the alarm condition that the feedback problem description involves key focus content.

[0143] For more information about each module, see Figures 2~10 The relevant description of will not be repeated here. It should be understood that Fig.11 The system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or control codes contained in a processor, such as a disk, CD or DVD-ROM, etc., and such codes are provided in a carrier medium or a memory of a programmable device. The system and its modules of this specification can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (e.g., firmware).

[0144] It should be noted that the above description of the system and its modules is only for convenience of description and does not limit this specification to the scope of the embodiments. It is understandable that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine the modules without deviating from this principle to form a subsystem connected to other modules. Or some modules may be split to obtain more modules or multiple units under the module. Such variations are all within the scope of the disclosure of this specification.

[0145] Some embodiments of this specification also provide a computer program product, including computer instructions or computer codes. When at least part of the computer instructions or the computer codes are executed by a processor, the feedback information processing method shown in some embodiments of this specification can be implemented. In some embodiments, the computer program product may only involve computer instructions or computer codes, which may be carried by a storage medium or a processing device. In other embodiments, the computer program product may also be a storage medium or a processing device containing the aforementioned computer instructions or computer codes. The processing device may include one or more processors, and a storage medium. For more information about processors, processing devices, and storage media, please refer to Figure 1 The relevant description will not be repeated here.

[0146] The beneficial effects that may be brought about by the embodiments of this specification include but are not limited to: (1) Actively collect first user statements from multiple channels such as web pages, new media platforms, chat data, user feedback, etc., so as to timely and comprehensively discover problems or anomalies in the product from user opinions. (2) Combine the two methods of invalid information filtering and useful information screening to improve the filtering efficiency and accuracy of the statements. (3) Direct clustering or classification combined with clustering is used to process the filtered first user statements, so that feedback problem descriptions displayed at different information levels can be obtained to meet different user needs. (4) Secondary filtering is performed on the clustering results to remove first user statements that are not related to product problems or suggestions from a deeper semantic level, improve the accuracy of the feedback problem description, and avoid wasting task processing resources. (5) The processing results are presented in two ways: page display and alarm push, taking into account the completeness and timeliness of information presentation. It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

[0147] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are taught in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

Claims

1. A feedback information processing method, characterized in that: include: Collecting first user statements, wherein the first user statements include statements reflecting the first user's remarks; Processing the first user statement to extract one or more feedback question descriptions; presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages; In response to a feedback question follow-up request initiated through at least one of the one or more feedback data pages, a task item is created and sent to a second user terminal associated with the task item.

2. The method according to claim 1, characterized in that The first user statement comes from one or more of the following data: online speech, chat data, first user question, and first user suggestion.

3. The method according to claim 1, characterized in that: The processing of the first user statement to extract one or more feedback question descriptions includes: Filtering the first user statement based on the whitelist to obtain the filtered first user statement; Classifying the filtered first user sentences based on the first classification model to obtain first user sentences belonging to a target category and a non-target category, and using the first user sentences in the target category as the filtered first user sentences; Based on the filtered first user sentence, obtaining one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters; A cluster summary of the question feedback cluster is generated, and the first user sentence corresponding to the question feedback cluster whose cluster summary does not belong to the target category is removed.

4. The method according to claim 3, characterized in that The filtering the first user statement based on the whitelist to obtain the filtered first user statement includes: Removing first user sentences whose number of repetitions in the first user sentences is greater than a first set threshold, to obtain remaining first user sentences; Removing the first user sentences containing the preset whitelist keywords from the remaining first user sentences, and / or removing the first user sentences whose semantic similarity with the preset whitelist keywords is greater than a second set threshold from the remaining first user sentences, to obtain the filtered first user sentences; The processing of the first user statement to extract one or more feedback question descriptions further includes: Before obtaining one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters based on the filtered first user sentence, the first user sentence whose word count is greater than a third set threshold is processed based on the first language model to obtain a first user sentence whose word count is not greater than a fourth set threshold; wherein the fourth set threshold is less than or equal to the third set threshold.

5. The method according to claim 3, characterized in that: The step of removing the first user statement corresponding to the question feedback cluster whose cluster summary does not belong to the target category includes: Based on the cluster summary of the question feedback cluster, web page data is obtained through a search engine, and whether the cluster summary of the question feedback cluster belongs to the target category is determined based on the web page data through a second language model; when the cluster summary of the question feedback cluster does not belong to the target category, the first user sentence corresponding to the question feedback cluster is removed.

6. The method according to claim 3, characterized in that: The obtaining, based on the filtered first user statement, one or more problem feedback clusters and one or more feedback problem descriptions corresponding to the problem feedback clusters includes: Processing the filtered first user sentences in a first clustering manner to obtain one or more first question feedback clusters and one or more first feedback question descriptions corresponding to the first question feedback clusters; Processing the filtered first user sentence in a second clustering manner to obtain one or more second question feedback clusters and one or more second feedback question descriptions corresponding to the second question feedback clusters; The first clustering method and the second clustering method correspond to different semantic levels and / or semantic classification granularities.

7. The method according to claim 6, characterized in that The step of processing the filtered first user sentence in a first clustering manner to obtain one or more first question feedback clusters and one or more first feedback question descriptions corresponding to the first question feedback clusters includes: Clustering the filtered first user sentences to obtain the one or more first question feedback clusters; Based on one or more first user sentences in the first question feedback cluster whose sentence confidences meet the first set condition, a cluster summary of the first question feedback cluster is generated; wherein the sentence confidences of the first user sentences are determined based on keywords extracted from the first user sentences and one or more of the following information: portrait information of the first user related to the first user sentences, and the source of the first user sentences; The one or more first feedback question descriptions are obtained based on the cluster summaries of the one or more first question feedback clusters.

8. The method according to claim 7, characterized in that The presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages includes: Presenting at least part of the one or more first feedback question descriptions and corresponding expansion controls through a first feedback data page; In response to a triggering operation of an expansion control on the first feedback data page, a first user statement in a problem feedback cluster corresponding to a first feedback problem description corresponding to the expansion control is presented.

9. The method according to claim 3, characterized in that: The step of processing the filtered first user sentence in a second clustering manner to obtain one or more second question feedback clusters and one or more second feedback question descriptions corresponding to the second question feedback clusters includes: Classifying the filtered first user sentences based on the second classification model to obtain first user sentences belonging to one or more question tags; Obtaining the one or more second feedback question descriptions based on the one or more question tags; Clustering the first user sentences belonging to the same question label to obtain one or more second question feedback clusters belonging to the question label; Based on one or more first user statements in the second question feedback cluster whose statement confidences meet the second set condition, a cluster summary of the second question feedback cluster is generated; wherein the statement confidence of the first user statement is determined based on keywords extracted from the first user statement and one or more of the following information: portrait information of the first user related to the first user statement, and the source of the first user statement.

10. The method according to claim 9, characterized in that The step of classifying the filtered first user sentences based on the second classification model to obtain first user sentences belonging to one or more question tags includes: Determine classification scores for the filtered first user sentences respectively by using the second classification model; Based on the classification scores and the score ranges corresponding to the one or more question tags, classify the filtered first user statements to obtain the first user statements belonging to the one or more question tags; The score range corresponding to the one or more question tags is adjustable.

11. The method according to claim 9, characterized in that The presenting at least part of the one or more feedback problem descriptions through one or more feedback data pages includes: Presenting at least part of the one or more second feedback question descriptions and corresponding expansion controls through the second feedback data page; In response to the triggering operation of the expansion control on the second feedback data page, a cluster summary of the problem feedback cluster belonging to the second feedback problem description corresponding to the expansion control is presented, and / or a first user statement belonging to the second feedback problem description corresponding to the expansion control is presented.

12. The method according to claim 1, characterized in that The one or more feedback data pages also present a feedback question follow-up control corresponding to the feedback question description; The step of creating a task item and sending it to a second user terminal associated with the task item in response to a feedback question follow-up request initiated through at least one of the one or more feedback data pages comprises: In response to a triggering operation of a feedback question follow-up control in the at least one feedback data page, the task item is created based on a feedback question description corresponding to the feedback question follow-up control.

13. The method according to claim 12, characterized in that Also includes: Presenting task item information and adjusting task progress controls through a task item page; wherein the task item information includes one or more of the following information: task name, second user information associated with the task item, task progress, and one or more first user statements related to the feedback problem description; In response to the triggering operation of adjusting the task progress control, presenting one or more first user statements related to the feedback problem description and their corresponding first user information and a reply control through a task processing page; In response to the triggering operation of the reply control, the reply content is obtained and sent to a first user terminal corresponding to at least part of the first user statements in one or more first user statements related to the feedback question description.

14. The method according to claim 1, characterized in that Also includes: Presenting, through the feedback data page, a first user statement related to at least part of the feedback question description and a corresponding verification control thereof; In response to a triggering operation of a verification control corresponding to the first user statement, displaying source data and / or image data of the first user statement; The feedback data page presents verification controls corresponding to at least part of the feedback problem description; in response to a triggering operation of the verification control corresponding to the feedback problem description, the portrait information of the first user corresponding to the feedback problem description is displayed.

15. The method according to claim 1, characterized in that Also includes: When at least part of the one or more feedback problem descriptions meets the preset alarm conditions corresponding to one or more alarm types, an alarm message is sent to one or more second user terminals through an alarm window; the alarm message includes one or more of the following information: an alarm type, a feedback problem description related to the alarm message, and one or more first user statements related to the feedback problem description; The one or more alarm types include one or more of the following types: the emergence of a new feedback problem description, a sharp increase in the number of first user statements related to the feedback problem description, the feedback problem description is related to the first user who is the focus of attention, and the list is broadcasted regularly.

16. The method according to claim 1, characterized in that Also includes: When at least part of the first user statements in the first user statements meet the alarm condition that the feedback problem description involves key focus content, an alarm message is sent to one or more second user terminals through the alarm window; the alarm message includes one or more of the following information: the alarm type, the key focus content related to the alarm information, and one or more first user statements containing the key focus content related to the alarm information.

17. The method according to claim 1, characterized in that Also includes: In response to a priority configuration operation corresponding to one or more feedback problem descriptions, setting the priority of the corresponding feedback problem descriptions; When at least part of the one or more feedback problem descriptions meet the preset alarm conditions and priority conditions corresponding to one or more alarm types, an alarm message is sent to one or more second user terminals through the alarm window; the alarm message includes one or more of the following information: the alarm type, the feedback problem description related to the alarm message, and one or more first user statements related to the feedback problem description.

18. The method according to any one of claims 15 to 17, characterized in that: The alarm window also includes a feedback problem follow-up control; The method further comprises: In response to the triggering operation of the feedback question follow-up control, a task item is created based on the alarm information and sent to a second user terminal associated with the task item.

19. The method according to claim 18, characterized in that The feedback question follow-up control includes a second user information input box and a sending control; The step of creating a task item based on the alarm information and sending the task item to a second user terminal associated with the task item in response to the triggering operation of the feedback question follow-up control comprises: Acquire the second user information through the second user information input box; In response to the triggering operation of the sending control, the task item is sent to the second user terminal corresponding to the second user information, and the display attribute of the alarm window is switched to a first state to indicate that the task process is being processed.

20. The method according to claim 18, characterized in that The method further comprises: In response to a triggering operation of the feedback question follow-up control, the display attribute of the alarm window is switched to a first state to indicate that the task process is in progress, and the second user information associated with the task item and the processing control and / or the completion control are displayed in the alarm window; In response to the triggering operation of the processing control, obtaining second user information corresponding to the second user terminal that initiates the triggering operation and additionally displaying it in the alarm window; In response to the triggering operation of the completion control, the display attribute of the alarm window is switched to a second state to indicate that the task process is completed.

21. The method according to any one of claims 15 to 17, characterized in that: The method further comprises: Creating a task item based on the alarm information, and sending it to a second user terminal associated with the task item determined based on a correspondence between a preset problem keyword and a second user; Switching the display attribute of the alarm window to a first state to indicate that the task process is being processed, displaying the second user information associated with the task item and displaying a processing control and / or a completion control in the alarm window; In response to the triggering operation of the processing control, obtaining second user information corresponding to the second user terminal that initiates the triggering operation and additionally displaying it in the alarm window; In response to the triggering operation of the completion control, the display attribute of the alarm window is switched to a second state to indicate that the task process is completed.

22. The method according to any one of claims 15 to 17, characterized in that: The alarm window is presented in a chat interface; wherein the chat interface includes a group chat interface.

23. A feedback information processing system, characterized in that: include: A collection module, configured to collect first user statements, wherein the first user statements include statements reflecting the first user's remarks; A processing module, configured to process the first user statement and extract one or more feedback question descriptions; A display module, configured to present at least a portion of the one or more feedback problem descriptions through one or more feedback data pages; The task module is used to create a task item and send it to a second user terminal associated with the task item in response to a feedback issue follow-up request initiated through at least one feedback data page of the one or more feedback data pages.

24. A computer program product, comprising computer codes, which, when at least part of the computer codes are executed by a processor, can implement the method according to any one of claims 1 to 22.