Anti-fraud propaganda method and system, storage medium and electronic equipment

Abstract

The invention discloses an anti-fraud propaganda method and system, a storage medium and electronic equipment, and relates to the technical field of information. The method comprises the steps of obtaining an anti-fraud propaganda task list, and determining a propaganda verbal skill library based on historical call records of a current callee; calling a first propaganda verbal skill from a propaganda verbal skill library to inquire the current callee, and acquiring first feedback information of the current callee; determining a key feedback word and an emotional feature in the first feedback information, calling a second propaganda verbal skill from a propaganda verbal skill library based on the key feedback word and the emotional feature to answer, and obtaining second feedback information of the current callee; and determining an anti-fraud propaganda result of the current callee based on the second feedback information, determining a next callee in the task list until anti-fraud propaganda of all callees in the task list is completed, and generating a propaganda result report corresponding to the task list. By implementing the technical scheme provided by the invention, the anti-fraud propaganda efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of information technology, and in particular to an anti-fraud publicity method, system, storage medium and electronic device. Background Art

[0002] With the rapid development of information technology, telecom network fraud has been constantly changing, posing a serious threat to the property safety of the people. In order to raise the public's awareness of anti-fraud, all regions have generally carried out anti-fraud publicity work, popularizing anti-fraud knowledge to the public through telephone calls, text messages and other means.

[0003] At present, anti-fraud publicity work is mainly carried out through manual outbound calls. Staff members promote anti-fraud knowledge to the callees based on a unified script. Although this method can achieve basic publicity purposes, in actual applications, it is impossible to carry out publicity in a timely manner due to limited human resources. In addition, staff members often use a unified script template when communicating with the public, which makes it difficult to carry out personalized publicity based on the actual situation of different groups of people, resulting in low anti-fraud publicity efficiency. Summary of the invention

[0004] The present application provides an anti-fraud publicity method, system, storage medium and electronic device, which can improve the anti-fraud publicity efficiency.

[0005] In a first aspect, the present application provides an anti-fraud publicity method, the method comprising: Obtaining a task list for anti-fraud publicity, wherein the task list includes at least one callee; Determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain a publicity script library corresponding to the anti-fraud publicity level; Retrieving a first promotional speech from the promotional speech library to inquire the current callee, and obtaining first feedback information from the current callee; Determine key feedback words and sentiment features in the first feedback information, and retrieve second promotional words from the promotional words library based on the key feedback words and sentiment features to answer, thereby obtaining second feedback information of the current callee; The anti-fraud publicity result of the current callee is determined based on the second feedback information, and the next callee is determined in the task list until the anti-fraud publicity for all callees in the task list is completed, and a publicity result report corresponding to the task list is generated.

[0006] By adopting the above technical solution, by obtaining a task list for anti-fraud publicity, determining the anti-fraud publicity level based on the historical call records of the callee and obtaining the corresponding publicity script library, the anti-fraud publicity can be carried out in a targeted manner according to the actual situation of different callees; then, by calling the first publicity script in the publicity script library to inquire and obtain the first feedback information of the callee, determining the key feedback words and emotional features in the first feedback information, and calling the second publicity script based on these features to answer, the effect of automatically adjusting the publicity script according to the real-time feedback of the callee is achieved; finally, the anti-fraud publicity result is determined based on the second feedback information and the next callee is automatically determined, until the anti-fraud publicity for all callees in the task list is completed, and a publicity result report is generated, thereby realizing batch and automated anti-fraud publicity without manual intervention, thereby improving the efficiency of anti-fraud publicity.

[0007] In a second aspect of the present application, an anti-fraud publicity system is provided, the system comprising: A task list acquisition module, used to acquire a task list for anti-fraud publicity, wherein the task list includes at least one callee; A speech library determination module is used to determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain the publicity speech library corresponding to the anti-fraud publicity level; The anti-fraud publicity module is used to retrieve a first publicity speech from the publicity speech library to inquire the current callee and obtain first feedback information from the current callee; determine key feedback words and emotional features in the first feedback information, and retrieve a second publicity speech from the publicity speech library based on the key feedback words and emotional features to answer and obtain second feedback information from the current callee; A publicity report generation module is used to determine the anti-fraud publicity result of the current callee based on the second feedback information, determine the next callee in the task list, until the anti-fraud publicity for all callees in the task list is completed, and generate a publicity result report corresponding to the task list.

[0008] In a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above method steps.

[0009] In a fourth aspect of the present application, an electronic device is provided, comprising: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0010] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: The present application obtains a task list for anti-fraud publicity, determines the anti-fraud publicity level based on the historical call records of the callees, and obtains the corresponding publicity script library, so that the anti-fraud publicity can be carried out in a targeted manner according to the actual conditions of different callees; then, the first publicity script is retrieved from the publicity script library for inquiry and the first feedback information of the callee is obtained, key feedback words and emotional features are determined in the first feedback information, and the second publicity script is retrieved based on these features for answering, thereby achieving the effect of automatically adjusting the publicity script according to the real-time feedback of the callee; finally, the anti-fraud publicity result is determined based on the second feedback information and the next callee is automatically determined, until the anti-fraud publicity for all callees in the task list is completed, and a publicity result report is generated, thereby realizing batch and automated anti-fraud publicity without manual intervention, thereby improving the efficiency of anti-fraud publicity. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a flowchart of an anti-fraud publicity method provided in an embodiment of the present application; Figure 2 It is a module schematic diagram of an anti-fraud publicity system provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application.

[0012] Description of reference numerals: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0013] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0014] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0015] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0016] The following will provide a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0017] Please refer to Figure 1 , a flowchart of an anti-fraud propaganda method is proposed, the method can be implemented by a computer program, can be implemented by a single-chip microcomputer, or can be run on an anti-fraud propaganda system, the computer program can be integrated in a computer device, or can be run as an independent tool application, specifically, the method includes steps 10 to 50, the above steps are as follows: Step 10: Obtain a task list for anti-fraud publicity, which includes at least one callee.

[0018] In the embodiment of the present application, the task list refers to a list of target objects that need to undergo anti-fraud publicity, and the list includes information of at least one callee that needs to receive anti-fraud publicity.

[0019] In the embodiment of the present application, the callee refers to the user who needs to receive anti-fraud publicity, including but not limited to users who submit consultation applications through appointment channels such as government service platforms, anti-fraud apps and community service stations.

[0020] Specifically, the occurrence index of fraud cases in each jurisdiction is obtained. Since the frequency and degree of harm of fraud cases in different jurisdictions are different, it is necessary to determine the publicity priority of each jurisdiction based on the occurrence index, and determine the jurisdiction with a publicity priority greater than the preset priority as the target jurisdiction, so that anti-fraud publicity can be given priority to high-risk areas. For the determined target jurisdiction, this application obtains consultation application information from multiple appointment channels such as government service platforms, anti-fraud APPs and community service stations. These channels serve as entrances for users to actively submit consultation needs, and can effectively collect real and effective user information. After obtaining the consultation application information, this information is integrated to generate an anti-fraud publicity task list. The task list contains the information of the callees who need to carry out anti-fraud publicity, laying the foundation for subsequent precise publicity work. In this way, the whole process management from case risk assessment to user information collection is realized, which not only ensures that publicity resources are tilted towards high-risk areas, but also ensures the authenticity and reliability of the callee information, and improves the pertinence and effectiveness of anti-fraud publicity.

[0021] Based on the above embodiment, as an optional embodiment, the step of obtaining the task list of anti-fraud publicity may further include the following steps: Step 101: Obtain the incidence index of fraud cases in each jurisdiction, and determine the publicity priority of each jurisdiction based on the incidence index, and determine the jurisdiction with a publicity priority greater than a preset priority as a target jurisdiction.

[0022] Specifically, due to the different levels of fraud risk in different jurisdictions, it is necessary to reasonably allocate publicity resources. Therefore, it is first necessary to obtain the incidence index of fraud cases in each jurisdiction. By calling up the case database of each jurisdiction, the number of recent fraud cases, the amount involved, the number of victims and other information can be extracted, and the incidence index of each jurisdiction can be obtained by weighted calculation. For example, the weight of the number of cases can be set to 0.4, the weight of the amount involved can be set to 0.3, and the weight of the number of victims can be set to 0.3. These weighted values ​​can be added together to obtain the final incidence index. Next, the incidence index of each jurisdiction is sorted from high to low, and a preset priority threshold is set based on historical experience values, and the jurisdiction with an incidence index higher than the threshold is determined as the target jurisdiction.

[0023] Step 102: Obtain consultation application information for each appointment channel in the target jurisdiction, where the appointment channels include government service platforms, anti-fraud apps, and community service stations.

[0024] Specifically, after determining the target jurisdiction, the system will connect to the data interface of each appointment channel to obtain consultation application information. For the government service platform, the system calls the platform API interface to obtain user information who actively checks the "need anti-fraud publicity and consultation" option during the process of handling government affairs; for anti-fraud APPs, the system extracts the application records submitted by users through the "appointment consultation" function from the APP background server; for community service stations, the system collects the anti-fraud consultation needs registered by residents at the service station through the community management system. The consultation application information obtained includes but is not limited to fields such as user name, contact number, appointment time period, and fraud type of concern.

[0025] Step 103: Generate an anti-fraud publicity task list based on the consultation application information, and the task list includes at least one callee.

[0026] Specifically, after obtaining the consultation application information from various channels, the system first cleans and integrates the information, removes duplicate applications and invalid information, and then sorts the applications according to the order of appointment time. For applications in the same time period, the system will prioritize high-risk groups such as the elderly and high-frequency telephone users. The processed information will be organized into a task list in a standard format. Each task list corresponds to a preset number of callees, such as 30. Each record contains the callee's basic information, appointment information, risk level, etc. The task list is stored in a queue data structure to facilitate subsequent outbound call processing in sequence.

[0027] Step 20: Determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain the publicity script library corresponding to the anti-fraud publicity level.

[0028] In the embodiment of the present application, historical call records refer to the call data of the callee within a preset time period, including but not limited to call duration, call frequency, and historical data of anti-fraud publicity.

[0029] In the embodiment of the present application, the anti-fraud publicity level refers to the publicity difficulty classification obtained based on the analysis of the historical call records of the callee. This level is used to guide the system to select an appropriate publicity strategy. For example, the anti-fraud publicity level can be divided into primary, intermediate and advanced levels. The primary level is suitable for callees who are accepting publicity for the first time or are easy to accept publicity, the intermediate level is suitable for callees who need to be emphasized, and the advanced level is suitable for callees who need repeated publicity or are difficult to accept publicity.

[0030] In the embodiment of the present application, the promotional script library refers to a set of standardized scripts corresponding to different anti-fraud publicity levels. Each script set includes but is not limited to opening scripts, fraud case explanation scripts, prevention points reminder scripts, coping skills explanation scripts and other categories, and each category is provided with multiple specific script contents for different scenarios.

[0031] Specifically, first retrieve the historical call records of the current callee in the past three months, which contain telephone feature information such as the area code of the call number, the type of number segment, the incoming and outgoing call identification, and the call duration data corresponding to each number. Then, the acquired telephone features are matched with the pre-established abnormal number feature library, which stores typical feature data such as the features of overseas numbers with high incidence of fraud, virtual number features, and high-frequency number segment features. The system uses the calculation method of feature vector cosine similarity for matching. When the matching degree exceeds the matching threshold, the number is judged as an abnormal number, and its corresponding call duration is recorded as the abnormal call duration. Subsequently, the total number of abnormal numbers N and the abnormal call duration Ti of each abnormal number are statistically obtained, and the anti-fraud publicity level is determined according to the preset judgment rules. Finally, according to the determined anti-fraud publicity level, the system retrieves the publicity speech library of the corresponding level from the preset graded speech library for subsequent publicity work.

[0032] Based on the above embodiment, as an optional embodiment, the step of determining the anti-fraud publicity level based on the historical call records of the current callee may further include the following steps: Step 201: Acquire the historical call records of the current callee, where the historical call records include the telephone features of each call number and the corresponding call duration.

[0033] Specifically, the operator's communication database is connected through an interface to obtain the historical call records of the current callee in the past three months. The telephone feature information contained in each call record mainly includes: the number's location information (such as country code, province area code, city area code), number type identification (such as fixed-line, mobile phone, virtual number), service provider type (such as China Telecom, China Mobile, China Unicom), call type (calling or called) and other feature data, and also records the duration of each call between the number and the called person.

[0034] Step 202: Match each telephone feature in the historical call records with a preset abnormal number feature library to obtain abnormal numbers with a matching degree exceeding a matching degree threshold and corresponding abnormal call durations.

[0035] Specifically, after obtaining the historical call records, the preset abnormal number feature library will be read for feature matching. The abnormal number feature library stores various types of abnormal number feature vectors, including overseas fraud number features (such as a specific country area code + number segment combination), virtual number features (such as a specific virtual number segment + service provider combination), high-risk area number features (such as number segment features in fraud-prone areas), etc. For each number in the historical call records, the system converts its phone features into feature vectors, and uses a vector space model to calculate the similarity with each feature vector in the abnormal number feature library. The specific calculation method is: first, the phone feature vector is standardized, and then the cosine similarity with the abnormal feature vector is calculated. When the similarity value exceeds the preset matching threshold of 0.8, the number is marked as an abnormal number, and its corresponding call duration is recorded as the abnormal call duration.

[0036] Step 203: Determine the anti-fraud publicity level of the current callee based on the number of abnormal numbers and the abnormal call duration corresponding to each abnormal number.

[0037] Specifically, after completing feature matching, count the number of all marked abnormal numbers N, and record the abnormal call duration Ti corresponding to each abnormal number. Then, the system determines the anti-fraud publicity level according to the following judgment rules: when the number of abnormal numbers N is less than 3, and all abnormal call durations Ti are less than 5 minutes, the system determines that the callee is low risk and determines the anti-fraud publicity level to be primary; when the number of abnormal numbers is between 3 and 5 (including 3 but not including 5), or there is a single abnormal call duration Ti greater than or equal to 5 minutes but less than 10 minutes, the system determines that the callee is medium risk and determines the anti-fraud publicity level to be intermediate; when the number of abnormal numbers is greater than or equal to 5, or there is a single abnormal call duration Ti greater than or equal to 10 minutes, the system determines that the callee is high risk and determines the anti-fraud publicity level to be advanced.

[0038] Step 30: Retrieve the first promotional words from the promotional words library to inquire the current callee, and obtain the first feedback information of the current callee.

[0039] The first promotional speech in the embodiment of the present application refers to the standardized speech content used to establish a call with the callee for the first time, which mainly includes the opening remarks such as identity introduction, purpose of the call, and asking whether the caller is willing to accept anti-fraud publicity. For example, "Hello, I am a staff member of the XX Anti-Fraud Center. We found that you may have been exposed to some abnormal numbers recently. In order to protect the safety of your funds, we would like to provide you with some anti-fraud advice. Is it convenient for you to listen now?" and other standardized expressions.

[0040] In the embodiment of the present application, the first feedback information refers to the response information of the callee to the first promotional speech, including but not limited to the content of the voice response, the characteristics of the tone and intonation, the strength of the willingness to cooperate, etc. For example, the callee may express active cooperation with "OK, I'm willing to listen", or express temporary inability to cooperate with "I don't have time now", or directly hang up the phone to refuse to cooperate. These different types of feedback will be recorded by the system and used to adjust the subsequent speech strategy.

[0041] Specifically, first, according to the current call scenario, the first promotional speech is retrieved from the promotional speech library. The promotional speech library adopts a hierarchical structure design and includes multiple categories such as opening speech, theme promotion speech and closing speech. The opening speech is the first promotional speech. When retrieving the first promotional speech, the system will automatically select the corresponding greeting start according to the current time period (morning / afternoon / evening) to form a personalized complete speech, such as "Good morning, I am a staff member of the XX Anti-Fraud Center...". After completing the retrieval of the first promotional speech, the promotional theme can be determined according to the anti-fraud promotion level, and the target tone parameter can be determined based on the importance of the promotional theme. Then, according to the target tone parameter, the first promotional speech is converted into voice and played to the callee through the communication network. At the same time, the system starts the voice recognition module and the voice feature analysis module to capture the callee's response in real time. The voice recognition module converts the callee's voice reply into text content, and the voice feature analysis module extracts the acoustic feature parameters of the voice, such as pitch, speech speed, and volume. The system integrates these recognition and analysis results into the first feedback information, which includes: response type (such as "agree", "rejection", "hesitation", etc.), tone characteristics (such as "positive", "negative", "neutral", etc.), reply time (the time interval from the end of the script playback to the response) and other multi-dimensional information.

[0042] Based on the above embodiment, as an optional embodiment, the step of calling the first promotional words in the promotional words library to inquire the current callee and obtaining the first feedback information of the current callee may further include the following steps: Step 301: Determine a publicity theme based on the anti-fraud publicity level, and determine a script branch based on the publicity theme.

[0043] Specifically, the corresponding publicity theme is determined from the preset theme mapping table according to the previously determined anti-fraud publicity level. When the anti-fraud publicity level is primary, basic prevention themes are selected, such as "Basic knowledge for preventing telecommunications network fraud" and "Anti-fraud APP installation guide"; when the anti-fraud publicity level is intermediate, warning education themes are selected, such as "Typical fraud case analysis" and "Identification of fraud methods"; when the anti-fraud publicity level is advanced, in-depth communication themes are selected, such as "Personal information protection measures" and "Fund account security protection". The system can select the most suitable publicity theme from the theme set corresponding to the current anti-fraud publicity level based on the preset theme priority and the historical record of whether the callee has received publicity on a certain theme. After determining the publicity theme, read the speech branch corresponding to the theme in the speech library. Each speech branch is a complete speech set designed for a specific theme, including multiple opening speeches, main speeches, and ending speeches. For example, when the topic "Analysis of Typical Fraud Cases" is selected, the opening lines in its speech branches may include: "Recently, there have been many cases of impersonation fraud...", "Do any of your relatives or friends have encountered fraud...", "I want to share with you a real fraud case..." and other alternative lines.

[0044] Step 302: Randomly select an opening speech from the speech branch as the first promotional speech.

[0045] Specifically, after the script branch is determined, a random algorithm is used to select one from the opening script set of the branch as the first promotional script. The purpose of random selection is to avoid using too single a script and to improve the naturalness and diversity of the promotion. The specific random selection process uses a weighted random algorithm. The system sets a usage weight for each opening script. The initial weights are equal. When a script is selected, its weight will be temporarily reduced to ensure that other scripts have an equal chance of being selected. At the same time, the system will record the scripts that have been used on the day to avoid repeating the same script for different callers in a short period of time. For example, suppose there are three opening scripts A, B, and C in the script branch, and the initial weights are all 1. When script A is selected once, its weight is temporarily reduced to 0.5. At this time, in the next random selection, the probability of scripts B and C being selected will be relatively higher. When all scripts have been used in one round, the system will reset the weights of all scripts and start a new round of random selection. In this way, the system ensures the randomness of the use of words and avoids the situation where certain words are overused or not used for a long time.

[0046] Step 303: Determine the target tone parameter based on the importance of the promotional theme, and convert the first promotional speech into voice according to the target tone parameter for playback inquiry.

[0047] Specifically, the preset topic-tone mapping table stores a set of standard tone parameters corresponding to topics of different importance. For example, for basic prevention topics of low importance, set a peaceful and friendly tone parameter: the speech speed is set to 0.8-1.2 times the standard speech speed, the tone fluctuation range is ±15%, and the volume level is medium; for warning education topics of medium importance, set a serious tone parameter: the speech speed is set to 0.9-1.0 times the standard speech speed, the tone fluctuation range is ±10%, and the volume level is medium-high; for in-depth communication topics of high importance, set a solemn and dignified tone parameter: the speech speed is set to 0.7-0.9 times the standard speech speed, the tone fluctuation range is ±5%, and the volume level is high.

[0048] After determining the target tone parameters, the speech synthesis module is called to convert the first propaganda speech into speech. During the speech synthesis process, the target tone parameters are input into the speech synthesis algorithm as control parameters. First, the text is semantically analyzed to identify key words and modal particles, and then the pronunciation characteristics of these words are adjusted according to the target tone parameters; then, the system controls the rhythm and intonation changes of the overall speech through the rhythm processing module; finally, the acoustic parameter synthesis module generates a complete speech that meets the target tone requirements. For example, when playing warning education speech such as "There have been many cases of impersonation and fraud recently", the system will slow down the speech speed appropriately, maintain a stable tone, and slightly increase the volume according to the preset serious tone parameters to highlight the seriousness of the content.

[0049] Step 304: Receive and record the response content of the current called party during the call as the first feedback information.

[0050] Specifically, while the voice is being played, the system starts the voice recognition module and the interaction recording module to capture and record the callee's response in real time. The voice recognition module uses a real-time voice recognition algorithm to convert the callee's voice response into text content. The interaction recording module synchronously records the timestamp, voice waveform characteristics, background noise and other auxiliary information of the response. Specifically, the system classifies and records the callee's response content according to the following dimensions: For example, semantic content: record the textual specific response content, such as "OK, I understand", "I know this", etc.; emotional tendency: analyze the emotional characteristics contained in the response tone through the voice emotion recognition algorithm, such as positive, negative, neutral, etc.; interactive characteristics: record the timeliness of the response (response delay time), completeness (whether there is a speech interruption), initiative (whether there is a follow-up question or supplement), etc.; noise conditions: record the background noise level of the call environment, which is used to judge the call quality and the environment of the callee. These recorded contents will be stored in the system database as the first feedback information for subsequent interaction strategy adjustment and publicity effect evaluation.

[0051] Step 40: Determine key feedback words and emotional features in the first feedback information, and retrieve second promotional words from the promotional words library based on the key feedback words and emotional features to answer, thereby obtaining second feedback information from the current callee.

[0052] In the embodiment of the present application, key feedback words refer to specific words or phrases used to judge the attitude and understanding of the callee. Specifically, they include: words that express understanding (such as "understand", "know", "know"), words that express doubt (such as "what do you mean", "who are you", "why"), words that express attitude (such as "ok", "let's talk about it later", "no time"), etc. The system evaluates the publicity effect and adjusts the subsequent speech strategy by identifying these key feedback words.

[0053] In the embodiment of the present application, emotional features refer to the emotional state indicators extracted from the voice response of the callee, including but not limited to intonation features: the amplitude of the rise and fall of the voice, the stability of the voice; speed features: the change of the speed of speaking, the frequency and duration of pauses; volume features: the change of the strength of the voice, the fluctuation of the volume; sound quality features: the clarity, trembling degree, hoarseness, etc. of the voice. These features together reflect the emotional state of the callee, such as nervousness, relaxation, impatience, seriousness, etc.

[0054] The second promotional language in the embodiment of the present application refers to the detailed promotional content of the next stage automatically selected and generated by the system based on the callee's feedback on the first promotional language.

[0055] The second feedback information in the embodiment of the present application refers to the comprehensive response information of the callee after receiving the second promotional speech, such as, but not limited to, voice response content, emotional response characteristics, interactive behavior characteristics, and degree of understanding and acceptance.

[0056] Specifically, the first feedback information is analyzed by a natural language processing algorithm, and the vocabulary in the feedback content is matched with the preset key feedback word library to extract keywords with judgment significance. At the same time, the voice emotion analysis module of the system will extract the emotional features in the feedback information, judge the positivity of the emotion by analyzing the fundamental frequency change range of the voice, evaluate the tension of speaking by the number of phonemes per unit time, identify the intensity of emotion by the change of sound intensity, and judge the stability of emotion by acoustic features such as harmonic noise ratio. After obtaining the key feedback words and emotional features, the system uses a multi-dimensional matching algorithm to search and match in the publicity speech library. The system will select the corresponding speech branch according to the type of key feedback words identified, such as when identifying doubtful keywords, it will tend to select the explanatory speech branch. At the same time, the expression of the speech is adjusted according to the extracted emotional features, such as selecting detailed knowledge speech when positive emotions are detected, and selecting short guiding speech when hesitant emotions are detected. Finally, from the selected speech branches, combined with the current context and emotional features, a complete second publicity speech is generated. The second promotional script is converted into voice according to the target tone parameters for playback and answering, and then the response of the callee is collected as the second feedback information.

[0057] Based on the above embodiment, as an optional embodiment, the step of determining the key feedback words and the emotional features in the first feedback information may further include the following steps: Step 4011: Perform word segmentation processing on the first feedback information to obtain a phrase sequence.

[0058] Specifically, the first feedback information obtained is segmented, and the text obtained by speech recognition is segmented through a natural language processing engine. A segmentation algorithm based on a combination of a dictionary and statistical probability is used. The text is first segmented using the maximum matching method, and then the segmentation result is optimized in combination with a two-way maximum matching algorithm. During the segmentation process, the system will consider the contextual relationship and semantic connection of the words, and handle special cases including ambiguous words and unregistered words. For example, when the callee's feedback content is "I'm not quite sure what's going on with this matter now", the phrase sequence obtained after the segmentation process is "I / now / don't / quite / clear / this / thing / is / specifically / what's going on".

[0059] Step 4012: Match the phrase sequence with a preset keyword library, and extract phrases whose matching degree exceeds a first threshold as key feedback words.

[0060] Specifically, when the system matches phrases, it will calculate the similarity between the obtained phrase sequence and the pre-established keyword library. The keyword library contains standard phrases of different categories such as the degree of understanding, emotional attitude, and willingness, and each standard phrase is set with a weight value. The system uses the cosine similarity algorithm to calculate the matching degree between the phrase to be matched and the standard phrase, and considers the semantic similarity of the words in combination with the word vector model. During the calculation process, the system will weight the context information such as the position relationship and collocation characteristics of the phrase to improve the accuracy of the match. When the matching degree between the phrase and the standard phrase exceeds the preset first threshold, the phrase will be determined as a key feedback word. The setting of the first threshold is based on a large amount of experimental data and expert experience, usually set between 0.75 and 0.85, and can be dynamically adjusted according to the actual application effect. In the above example, the system may identify "not very clear" as a key feedback word because it has a high matching degree with the standard phrase in the keyword library that indicates the degree of understanding. In addition, "what's going on" may also be identified as a key feedback word because it reflects the callee's state of doubt.

[0061] In order to improve the accuracy of extracting key feedback words, the system will also consider the combination relationship of phrases. When adjacent phrases together form a certain semantic expression, the system will match them as a whole. For example, a phrase combination such as "completely don't understand" may have a higher matching priority than the single "don't understand". At the same time, the system will also filter out some common words that have a high matching degree but do not have actual judgment significance, so as to ensure that the extracted key feedback words have substantial reference value.

[0062] Step 4013: construct a semantic dependency tree of the phrase sequence, and extract the modifying components and predicate components in the semantic dependency tree.

[0063] Specifically, a semantic dependency tree is constructed for the phrase sequence, and a semantic dependency analysis model based on deep learning is used for processing. The model analyzes the dependency relationship between phrases through a neural network algorithm and constructs a hierarchical tree structure. During the construction process, the system first determines the core predicate of the sentence, and then identifies the subject, object, adverbial and complement components related to the core predicate in turn, and finally forms a complete semantic dependency tree. For example, for the phrase sequence "I / now / not quite / clear / this / thing", the system will identify "clear" as the core predicate, "I" as the subject, "thing" as the object, and "not quite" as the adverbial modifier.

[0064] After establishing the semantic dependency tree, the focus is on extracting the modifiers and predicate components. Modifiers mainly include adverbials, degree words, and modal words, which often directly reflect the speaker's subjective attitude and emotional tendency. Predicate components include verbs and their tenses, mood markers, etc. These elements together constitute the core semantics of the sentence. By traversing the semantic dependency tree, the system classifies and stores these components according to their grammatical functions and semantic features. In the above example, "not so" as a modifier represents negation and degree, and "clear" as a predicate component represents cognitive state.

[0065] Step 4014: Determine the mood type based on the modifying component and the predicate component, and determine the emotional feature according to the mood type and the preset mood-emotion mapping relationship.

[0066] Specifically, based on the extracted modifiers and predicate components, the mood type is determined by the mood judgment model. The model comprehensively considers the formal features of negation words, degree words, modal particles and predicates, and divides the mood into basic types such as declarative mood, interrogative mood, imperative mood and exclamatory mood. At the same time, the system will also subdivide the intensity level of the mood, for example, the degree of negation is divided into complete negation, partial negation, etc. For expressions such as "not very clear", the system will determine it as a declarative mood with partial negation. After determining the mood type, the emotional feature is determined according to the preset mood-emotion mapping relationship. This mapping relationship is a knowledge base established based on a large amount of linguistic research and experimental data, which defines the corresponding rules between different mood types and emotional features. Emotional features usually include dimensions such as emotional polarity (positive, negative, neutral) and emotional intensity. The system will combine the specific features of the mood type, such as the degree of negation, the strength of the mood, etc., to calculate the corresponding emotional feature value. For example, a partially negative declarative mood may be mapped to a mildly negative emotional feature. In order to improve the accuracy of emotional feature judgment, the influence of the context can also be considered. By analyzing the semantic associations between the previous and next sentences, the emotional tendency of the current sentence can be understood more accurately. At the same time, the preset emotional dictionary will be combined to perform additional weighting on words with obvious emotional colors. This multi-dimensional analysis method can more comprehensively portray the emotional state of the callee.

[0067] Based on the above embodiment, as an optional embodiment, the step of retrieving the second promotional words from the promotional words library for answering based on the key feedback words and the emotional features may further include the following steps: Step 4021: Construct a multi-level dialogue strategy tree based on the promotional script library. The dialogue strategy tree includes multiple nodes, and each node contains script content and corresponding trigger conditions.

[0068] Specifically, a multi-level dialogue strategy tree is constructed in a top-down manner, and the promotional words are organized in layers according to themes, purposes, applicable scenarios and other dimensions. Theme categories are set at the top-level nodes of the tree, such as fraud warnings, safety tips, etc.; the second-level nodes correspond to different promotional purposes, such as knowledge popularization, risk warnings, behavior guidance, etc.; the third-level nodes are subdivided into specific words content. Each node contains the content of the words and trigger conditions, which are composed of multiple parameters such as key feedback word features, emotional features and scene features. For example, the trigger conditions of a node may include the key feedback word features of "low level of understanding", the emotional features of "mild doubts", and the scene features of "first contact".

[0069] Step 4022: Match the key feedback words and sentiment features with the triggering conditions of each node, and determine the optimal dialogue node in the dialogue strategy tree based on the matching results.

[0070] Specifically, according to the extracted key feedback words and emotional features, a dynamic programming algorithm is used to search for paths in the dialogue strategy tree. First, the current feedback features are matched with the trigger conditions of each node. The matching calculation involves feature comparisons in multiple dimensions. For each node, the system calculates a comprehensive score, which is composed of weighted factors such as feature matching, node weight, and historical feedback effect. For example, when it is detected that the callee shows difficulty in understanding a concept, the system will give priority to nodes containing detailed explanations and select appropriate expressions based on emotional features. During the search process, the system will consider the combined effects of the current node and subsequent nodes at the same time, and predict the possible development path of the dialogue through forward reasoning. The system uses a recursive algorithm to calculate the expected benefits of different paths and selects the path with the highest overall benefit. This method not only considers the adaptability of the current answer, but also takes into account the coherence of the subsequent dialogue and the possibility of achieving the goal. After comprehensive calculation, the node with the highest score is finally selected as the optimal dialogue node.

[0071] Step 4023: Select the speech content corresponding to the optimal dialogue node as the second promotional speech, and convert the second promotional speech into voice according to the target tone parameters for playing and answering.

[0072] Specifically, after obtaining the optimal dialogue node, the system extracts the corresponding speech content of the node as the second promotional speech. Then, the speech content is converted into voice according to the same target tone parameters as mentioned above. The target tone parameters include speech speed, intonation, volume, pauses and other elements. These parameters can also be dynamically adjusted according to the current emotional characteristics and dialogue strategies. For example, when the callee is detected to be anxious, the system will reduce the speech speed and adopt a calmer tone; when important information needs to be emphasized, the system will appropriately increase the volume and add emphasis, and finally convert the second promotional speech into voice according to the target tone parameters for playback and answer.

[0073] Step 50: Determine the anti-fraud publicity result for the current callee based on the second feedback information, determine the next callee in the task list, until the anti-fraud publicity for all callees in the task list is completed, and generate a publicity result report corresponding to the task list.

[0074] Specifically, the second feedback information obtained is subjected to deep semantic analysis, such as using a natural language processing model based on deep learning to perform semantic understanding of the feedback text through a multi-layer neural network. First, the text is converted into a word vector representation, and then the important semantic fragments in the text are identified through the attention mechanism to extract key semantic features including the cognitive state, attitude tendency and behavioral intention of the callee. For example, when the callee answers "I understand this fraud method and will definitely take precautions in the future", the system will extract key semantic features such as "understand" indicating cognitive state, "definitely" indicating attitude strength, and "pay attention to precautions" indicating behavioral intention.

[0075] The system pre-constructs a three-dimensional semantic evaluation matrix, which includes three evaluation dimensions: cognitive level, acceptance level, and willingness to prevent. Detailed scoring standards and judgment rules are set for each dimension. The cognitive level dimension evaluates the callee's level of understanding of anti-fraud knowledge, and the scoring standards involve aspects such as conceptual understanding, principle mastery, and application ability; the acceptance level dimension evaluates the callee's attitude towards the promotional content, including indicators such as emotional tendency, depth of identification, and initiative; the willingness to prevent dimension evaluates the callee's willingness to take preventive measures, examining factors such as behavioral commitment, specific plans, and execution determination.

[0076] After obtaining the key semantic features, the system uses a scoring algorithm to score the three dimensions separately. First, the extracted semantic features are matched with the scoring criteria of each dimension, and the preliminary scores are obtained through fuzzy calculation. Then, the system adjusts the scores considering the strength and credibility of the semantic features. For example, when a strong word such as "complete understanding" is detected, the cognitive level score will be increased accordingly; when a specific description of preventive measures is found, the preventive willingness score will receive additional points. The scoring range for each dimension is 0-100 points, and the scoring process uses precise quantitative standards to ensure the objectivity and comparability of the evaluation results.

[0077] Among them, the system presets weight parameters for three dimensions, which reflect the importance of each dimension in the overall evaluation. Generally speaking, the weight of the willingness to prevent is the highest, followed by the degree of cognition, and the weight of the degree of acceptance is relatively low. The system uses a weighted average algorithm to multiply the score value of each dimension by the corresponding weight and sum them to obtain the final publicity effect score. The calculation formula is: effect score = Σ(dimension score × dimension weight). For example, if the scores of the three dimensions are 85 points, 90 points, and 95 points respectively, and the weights are 0.3, 0.3, and 0.4 respectively, the final effect score is 90.5 points. Based on the calculated effect score, the system determines the anti-fraud publicity results according to the preset score range. Generally, scores above 90 are judged as excellent effects, indicating that the publicity has achieved significant results; scores of 75-90 are judged as good effects, indicating that the publicity goals have been basically achieved; scores of 60-75 are judged as general effects, which require follow-up; scores below 60 are judged as poor effects, and the publicity strategy needs to be adjusted and re-launched. At the same time, the system will update the publicity status of the current callee as "completed" and record the completion time and effect level.

[0078] Then, the information of the next person to be called is obtained from the task list. During the selection process, the system will consider multiple factors to prioritize. First, check the timeliness requirements of the task, and give priority to tasks with clear time limits; second, consider the regional distribution, and try to maintain the continuity of tasks in the same area to improve execution efficiency; finally, factors such as historical connection rate will be considered, and objects with a higher probability of connection will be given priority in a specific time period. The system determines the optimal task execution order through multi-dimensional priority calculations. Execute the promotional tasks in the task list in sequence until all tasks are completed. During the execution process, the system will count the completion status of the tasks in real time, including key indicators such as the number of completed tasks, average completion time, and cumulative effect scores. When certain abnormal situations are detected, such as multiple consecutive call failures or the effect score continues to be lower than expected, the system will trigger an early warning mechanism to indicate the need for manual intervention or strategy adjustment.

[0079] After completing all tasks, the system automatically generates a publicity result report. The report generation process is divided into three stages: data aggregation, analysis and calculation, and report writing. In the data aggregation stage, the system extracts all relevant data from the task execution database, including task completion status, effect evaluation results, abnormal records, etc. In the analysis and calculation stage, the system uses statistical analysis algorithms to calculate the mean, distribution characteristics and change trends of various indicators, and identify cases with significant effects and prominent problems. The report content includes an overview of the overall execution status, detailed statistical data, key case analysis, and improvement suggestions. In the overall overview, the system will generate core indicators such as task completion rate, average effect score, and abnormal situation statistics; in the statistical data section, the system will generate detailed data tables and charts according to different dimensions (such as time period, region, population characteristics, etc.); in the key case analysis, the system will select several cases with the best and worst effects and analyze their characteristics and reasons; in the improvement suggestion section, the system will automatically generate targeted optimization suggestions based on the data analysis results.

[0080] See also Figure 2 , is a schematic diagram of a module of an anti-fraud publicity system provided in an embodiment of the present application, wherein the system includes: A task list acquisition module, used to acquire a task list for anti-fraud publicity, wherein the task list includes at least one callee; A speech library determination module is used to determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain the publicity speech library corresponding to the anti-fraud publicity level; The anti-fraud publicity module is used to retrieve a first publicity speech from the publicity speech library to inquire the current callee and obtain first feedback information from the current callee; determine key feedback words and emotional features in the first feedback information, and retrieve a second publicity speech from the publicity speech library based on the key feedback words and emotional features to answer and obtain second feedback information from the current callee; A publicity report generation module is used to determine the anti-fraud publicity result of the current callee based on the second feedback information, determine the next callee in the task list, until the anti-fraud publicity for all callees in the task list is completed, and generate a publicity result report corresponding to the task list.

[0081] Optionally, the task list acquisition module is further used to obtain the incidence index of fraud cases in each jurisdiction, and determine the publicity priority of each jurisdiction according to the incidence index, and determine the jurisdiction whose publicity priority is greater than the preset priority as the target jurisdiction; Obtain consultation application information for each appointment channel in the target jurisdiction, wherein the appointment channels include government service platforms, anti-fraud apps, and community service stations; An anti-fraud publicity task list is generated based on the consultation application information, and the task list includes at least one callee.

[0082] Optionally, the speech library determination module is further used to obtain historical call records of the current callee, wherein the historical call records include telephone features of each call number and corresponding call duration; Matching each of the telephone features in the historical call records with a preset abnormal number feature library to obtain abnormal numbers with a matching degree exceeding a matching degree threshold and corresponding abnormal call durations; The anti-fraud publicity level of the current called party is determined based on the number of the abnormal numbers and the abnormal call durations corresponding to the abnormal numbers.

[0083] Optionally, the anti-fraud publicity module is further used to determine a publicity theme based on the anti-fraud publicity level, and determine a speech branch based on the publicity theme; Randomly selecting an opening speech from the speech branch as the first promotional speech; Determine a target tone parameter based on the importance of the promotional topic, and convert the first promotional speech into voice for playing inquiry according to the target tone parameter; Receive and record the response content of the current called party during the call as the first feedback information.

[0084] Optionally, the anti-fraud publicity module is further used to perform word segmentation processing on the first feedback information to obtain a phrase sequence; Matching the phrase sequence with a preset keyword library, and extracting phrases whose matching degree exceeds a first threshold as key feedback words; Constructing a semantic dependency tree of the phrase sequence, and extracting modifying components and predicate components in the semantic dependency tree; The mood type is determined based on the modifying component and the predicate component, and the emotional feature is determined according to the mood type and a preset mood-emotion mapping relationship.

[0085] Optionally, the anti-fraud publicity module is further used to construct a multi-level dialogue strategy tree based on the publicity speech library, wherein the dialogue strategy tree includes a plurality of nodes, each of which contains speech content and corresponding trigger conditions; Matching the key feedback words and sentiment features with the triggering conditions of each of the nodes, and determining the optimal dialogue node in the dialogue strategy tree based on the matching results; The speech content corresponding to the optimal dialogue node is selected as the second promotional speech, and the second promotional speech is converted into voice according to the target tone parameter for playing answer.

[0086] Optionally, the publicity report generating module is further used to perform semantic analysis on the second feedback information to extract key semantic features; Constructing a semantic evaluation matrix, wherein the semantic evaluation matrix includes scoring criteria for three evaluation dimensions: degree of cognition, degree of acceptance, and willingness to prevent; Scoring each of the evaluation dimensions based on the key semantic features to obtain a scoring value for each of the evaluation dimensions; Perform weighted calculation on the score values ​​of each evaluation dimension according to the preset dimension weights to obtain a publicity effect score; The anti-fraud publicity result for the current callee is determined based on the effect score.

[0087] It should be noted that: when the system provided in the above embodiment realizes its functions, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0088] An embodiment of the present application also provides a computer storage medium, which can store multiple instructions. The instructions are suitable for being loaded by a processor and executing an anti-fraud publicity method of the above embodiment. The specific execution process can be found in the specific description of the above embodiment, which will not be repeated here.

[0089] Please refer to Figure 3 The application also discloses an electronic device. Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0090] The communication bus 302 is used to realize the connection and communication between these components.

[0091] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0092] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0093] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0094] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally also be at least one storage device located away from the aforementioned processor 301. Refer to Figure 3 , as a computer storage medium, the memory 305 may include an operating system, a network communication module, a user interface module, and an application program for an anti-fraud publicity method.

[0095] exist Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call an application program storing an anti-fraud propaganda method in the memory 305, and when executed by one or more processors 301, the electronic device 300 executes one or more of the methods described in the above embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited to the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0096] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0098] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0099] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0100] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0101] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0102] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An anti-fraud publicity method, characterized in that: The method comprises: Obtaining a task list for anti-fraud publicity, wherein the task list includes at least one callee; Determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain a publicity script library corresponding to the anti-fraud publicity level; Retrieving a first promotional speech from the promotional speech library to inquire the current callee, and obtaining first feedback information from the current callee; Determining key feedback words and sentiment features in the first feedback information, and retrieving a second promotional speech from the promotional speech library based on the key feedback words and the sentiment features to answer, thereby obtaining second feedback information of the current callee; The anti-fraud publicity result of the current callee is determined based on the second feedback information, and the next callee is determined in the task list until the anti-fraud publicity for all callees in the task list is completed, and a publicity result report corresponding to the task list is generated.

2. The anti-fraud publicity method according to claim 1, characterized in that: The task list for obtaining anti-fraud publicity includes: Obtaining the incidence index of fraud cases in each jurisdiction, and determining the publicity priority of each jurisdiction according to the incidence index, and determining the jurisdiction whose publicity priority is greater than the preset priority as the target jurisdiction; Obtain consultation application information for each appointment channel in the target jurisdiction, wherein the appointment channels include government service platforms, anti-fraud apps, and community service stations; An anti-fraud publicity task list is generated based on the consultation application information, and the task list includes at least one callee.

3. The anti-fraud publicity method according to claim 1, characterized in that: The step of determining the anti-fraud publicity level based on the historical call records of the current callee includes: Obtaining the historical call records of the current callee, wherein the historical call records include the telephone characteristics of each call number and the corresponding call duration; Matching each of the telephone features in the historical call records with a preset abnormal number feature library to obtain abnormal numbers with a matching degree exceeding a matching degree threshold and corresponding abnormal call durations; The anti-fraud publicity level of the current called party is determined based on the number of the abnormal numbers and the abnormal call durations corresponding to the abnormal numbers.

4. The anti-fraud publicity method according to claim 1, characterized in that: The step of retrieving a first promotional speech from the promotional speech library to inquire the current callee and obtaining first feedback information from the current callee includes: Determine a publicity theme based on the anti-fraud publicity level, and determine a script branch based on the publicity theme; Randomly selecting an opening speech from the speech branch as the first promotional speech; Determine a target tone parameter based on the importance of the promotional topic, and convert the first promotional speech into voice for playing inquiry according to the target tone parameter; Receive and record the response content of the current called party during the call as the first feedback information.

5. The anti-fraud publicity method according to claim 1, characterized in that: The determining of key feedback words and sentiment features in the first feedback information includes: Performing word segmentation processing on the first feedback information to obtain a phrase sequence; Matching the phrase sequence with a preset keyword library, and extracting phrases whose matching degree exceeds a first threshold as key feedback words; Constructing a semantic dependency tree of the phrase sequence, and extracting modifying components and predicate components in the semantic dependency tree; The mood type is determined based on the modifying component and the predicate component, and the emotional feature is determined according to the mood type and a preset mood-emotion mapping relationship.

6. The anti-fraud publicity method according to claim 4, characterized in that: The method of retrieving a second promotional speech from the promotional speech library based on the key feedback words and the emotional features to answer includes: Constructing a multi-level dialogue strategy tree based on the promotional speech library, wherein the dialogue strategy tree includes a plurality of nodes, each of which includes speech content and corresponding triggering conditions; Matching the key feedback words and the emotional features with the triggering conditions of each of the nodes, and determining the optimal dialogue node in the dialogue strategy tree based on the matching results; The speech content corresponding to the optimal dialogue node is selected as the second promotional speech, and the second promotional speech is converted into voice according to the target tone parameter for playing and answering.

7. The anti-fraud publicity method according to claim 1, characterized in that: The determining the anti-fraud publicity result of the current callee based on the second feedback information includes: Performing semantic analysis on the second feedback information to extract key semantic features; Constructing a semantic evaluation matrix, wherein the semantic evaluation matrix includes scoring criteria for three evaluation dimensions: degree of cognition, degree of acceptance, and willingness to prevent; Scoring each of the evaluation dimensions based on the key semantic features to obtain a scoring value for each of the evaluation dimensions; Perform weighted calculation on the score values ​​of each evaluation dimension according to the preset dimension weights to obtain a publicity effect score; The anti-fraud publicity result for the current callee is determined based on the publicity effect score. 8.An anti-fraud publicity system, characterized in that: The system comprises: A task list acquisition module, used to acquire a task list for anti-fraud publicity, wherein the task list includes at least one callee; A speech library determination module is used to determine the anti-fraud publicity level based on the historical call records of the current callee, and obtain the publicity speech library corresponding to the anti-fraud publicity level; The anti-fraud publicity module is used to retrieve a first publicity speech from the publicity speech library to inquire the current callee and obtain first feedback information from the current callee; determine key feedback words and emotional features in the first feedback information, and retrieve a second publicity speech from the publicity speech library based on the key feedback words and emotional features to answer and obtain second feedback information from the current callee; A publicity report generation module is used to determine the anti-fraud publicity result of the current callee based on the second feedback information, determine the next callee in the task list, until the anti-fraud publicity for all callees in the task list is completed, and generate a publicity result report corresponding to the task list.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

Images