Intention level identification method and device, computer device, and storage medium

Abstract

The application relates to a will level recognition method and device, computer equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the following steps: an interactive robot performs intelligent interaction on a target user according to an interaction strategy; based on the content in call interaction information, the weight corresponding to each keyword is calculated through a will weight calculation strategy; the will degree score of the target user is calculated through a preset will degree score calculation method; and the will level of the target user is determined according to a preset will level discrimination algorithm, the interaction end node information, the mood index of the target user and the will degree score. The will level can effectively reflect the interaction effect, and the efficiency of interaction effect evaluation is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer device, and storage medium for intention-level recognition. Background Technology

[0002] With the development of technologies such as Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and Natural Language Understanding (NLU), intelligent interaction technology has emerged. Intelligent interactive robots can automatically initiate robot telephone interaction tasks based on business scenarios.

[0003] In traditional technologies, intelligent interactive robots utilize automatic speech recognition, text-to-speech, and natural language understanding technologies to conduct intelligent human-computer interaction.

[0004] However, current intelligent interaction technologies lack methods for analyzing user intentions (i.e., interaction effects) after human-computer interaction occurs, still requiring manual analysis, resulting in low efficiency in evaluating interaction effects. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer device, and computer-readable storage medium for identifying the level of user intention after the intelligent interaction has ended, in order to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for identifying the level of intent. The method includes:

[0007] Acquire interaction information obtained from the interaction between the interactive robot and the target user;

[0008] Based on the interaction information, the mood index corresponding to the target user is determined, and each willingness keyword and interaction end node information are extracted from the interaction information.

[0009] The intention level data of the target user is calculated by using a preset intention level algorithm, the weight of each intention keyword, and the frequency of occurrence of each intention keyword;

[0010] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0011] In one embodiment, before acquiring the interaction information obtained from the interaction between the interactive robot and the target user, the method further includes:

[0012] Obtain the target user's account information;

[0013] The target user's account information is input into a pre-trained interaction strategy selection model to obtain the target interaction strategy corresponding to the target user;

[0014] Based on the target interaction strategy, the interactive robot interacts with the target user.

[0015] In one embodiment, determining the mood index corresponding to the target user based on the interaction information includes:

[0016] The interaction information is analyzed and processed based on a preset sentiment analysis algorithm to obtain the mood index corresponding to the target user.

[0017] In one embodiment, the step of calculating the target user's willingness level data using a preset willingness level algorithm, the weights corresponding to each willingness keyword, and the frequency of occurrence of each willingness keyword includes:

[0018] For each desired keyword, the product of the frequency of occurrence of the desired keyword and the desired keyword coefficient is determined to obtain the first calculation result;

[0019] The first calculation result corresponding to each of the stated intention keywords is weighted and summed with the weight corresponding to each of the stated intention keywords to obtain the intention level data of the target user.

[0020] In one embodiment, the method further includes:

[0021] Obtain sample interaction information between each sample user and the interactive robot;

[0022] Extract sample intention keywords from the sample interaction information, classify the sample intention keywords, and obtain the intention keyword set corresponding to each type;

[0023] Linear fitting is performed based on the pre-stored mapping formula, the resource index values ​​of each sample user, the frequency of occurrence of sample intention keywords contained in each intention keyword set, and the intention coefficient corresponding to each intention keyword set to obtain the weight corresponding to each intention keyword set; wherein, the mapping formula is used to reflect the mapping relationship between the resource index values, the frequency of occurrence of each type of intention keyword, the intention coefficient corresponding to each type of intention keyword, and the weight corresponding to each type of intention keyword;

[0024] Based on the weights corresponding to each set of desired keywords, a correspondence between desired keywords and their weights is established.

[0025] In one embodiment, the method further includes:

[0026] The account resource data of sample users is obtained, and the account resource data is processed through a resource analysis algorithm to determine the resource indicator values.

[0027] In one embodiment, after determining the target user's willingness level using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data, the method further includes:

[0028] Based on the interaction information, the interaction results of each type are identified to obtain the identification results;

[0029] Based on the identification results, target interaction strategy feedback information corresponding to the target user is generated.

[0030] Secondly, this application also provides a willingness-level identification device. The device includes:

[0031] The first acquisition module is used to acquire interaction information obtained by the interactive robot interacting with the target user.

[0032] The extraction module is used to determine the mood index of the target user based on the interaction information, and to extract various types of intention keywords and interaction end node information from the interaction information;

[0033] The calculation module is used to calculate the target user's willingness level data by using a preset willingness level algorithm, the weights corresponding to each type of willingness keyword, and the frequency of occurrence of each type of willingness keyword;

[0034] The determination module is used to determine the target user's willingness level by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0035] In one embodiment, the device further includes:

[0036] The second acquisition module is used to acquire the target user's account information;

[0037] The input module is used to input the account information of the target user into a pre-trained interaction strategy selection model to obtain the target interaction strategy corresponding to the target user;

[0038] The interaction module interacts with the target user through the interactive robot, based on the target interaction strategy.

[0039] In one embodiment, the extraction module is specifically used for:

[0040] The interaction information is analyzed and processed based on a preset sentiment analysis algorithm to obtain the mood index corresponding to the target user.

[0041] In one embodiment, the computing module is specifically used for:

[0042] Based on the preset intention level algorithm, the frequency of occurrence of each intention keyword, the intention coefficient of each intention keyword, and the weight corresponding to each intention keyword, the intention level data of the target user is obtained by weighted summation.

[0043] In one embodiment, the device further includes:

[0044] The third acquisition module is used to acquire sample interaction information between each sample user and the interactive robot;

[0045] The classification module is used to extract sample intention keywords from the sample interaction information, classify the sample intention keywords, and obtain a set of intention keywords corresponding to each type.

[0046] The fitting module is used to perform linear fitting based on a pre-stored mapping formula, the resource index values ​​of each sample user, the frequency of occurrence of sample intention keywords contained in each intention keyword set, and the intention coefficient corresponding to each intention keyword set, to obtain the weight corresponding to each intention keyword set; wherein, the mapping formula is used to reflect the mapping relationship between the resource index values, the frequency of occurrence of each type of intention keyword, the intention coefficient corresponding to each type of intention keyword, and the weight corresponding to each type of intention keyword;

[0047] A module is established to establish a correspondence between desired keywords and their corresponding weights based on the weights of each set of desired keywords.

[0048] In one embodiment, the device further includes:

[0049] The fourth acquisition module is used to acquire the account resource data of sample users, and process the account resource data through a resource analysis algorithm to obtain resource indicator values.

[0050] In one embodiment, the device further includes:

[0051] The recognition module is used to recognize the various types of interaction results based on the interaction information, and obtain the recognition results;

[0052] The generation module is used to generate target interaction strategy feedback information corresponding to the target user based on the recognition results.

[0053] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0054] Acquire interaction information obtained from the interaction between the interactive robot and the target user;

[0055] Based on the interaction information, the mood index corresponding to the target user is determined, and each willingness keyword and interaction end node information are extracted from the interaction information.

[0056] The intention level data of the target user is calculated by using a preset intention level algorithm, the weight of each intention keyword, and the frequency of occurrence of each intention keyword;

[0057] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0058] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0059] Acquire interaction information obtained from the interaction between the interactive robot and the target user; the interaction information includes interactive information;

[0060] Based on the interaction information, the mood index corresponding to the target user is determined, and each willingness keyword and interaction end node information are extracted from the interaction information.

[0061] The intention level data of the target user is calculated by using a preset intention level algorithm, the weight of each intention keyword, and the frequency of occurrence of each intention keyword;

[0062] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0063] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0064] Acquire interaction information obtained from the interaction between the interactive robot and the target user; the interaction information includes interactive information;

[0065] Based on the interaction information, the mood index corresponding to the target user is determined, and each willingness keyword and interaction end node information are extracted from the interaction information.

[0066] The intention level data of the target user is calculated by using a preset intention level algorithm, the weight of each intention keyword, and the frequency of occurrence of each intention keyword;

[0067] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0068] The aforementioned method, apparatus, computer equipment, storage medium, and computer program product for identifying intention levels acquire interaction information obtained from interactions between an interactive robot and a target user. Based on this interaction information, the target user's mood index is determined. In addition, intention keywords and interaction end-point information are extracted from the interaction information. Using a preset intention level algorithm, the weights of each intention keyword, and the frequency of occurrence of each intention keyword, the target user's intention level data is calculated. Finally, using a preset intention level discrimination algorithm, interaction end-point information, mood index, and intention level data, the target user's intention level is determined. In this solution, the target user's intention level can be identified based on the interaction information between the interactive robot and the target user. This intention level effectively reflects the interaction effect without requiring manual analysis, thus improving the efficiency of interaction effect evaluation. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating a method for identifying intention levels in one embodiment;

[0070] Figure 2 This is a flowchart illustrating a method for identifying intention levels in one embodiment;

[0071] Figure 3 This is a flowchart illustrating a method for identifying intention levels in one embodiment;

[0072] Figure 4 This is a flowchart illustrating a method for identifying intention levels in one embodiment;

[0073] Figure 5 This is a flowchart illustrating a method for identifying intention levels in one embodiment;

[0074] Figure 6 This is a structural block diagram of a willingness-level recognition device in one embodiment;

[0075] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0077] In one embodiment, such as Figure 1 As shown, a method for identifying intent levels is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal communicates with the server via a network. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The server can be a standalone server or a server cluster consisting of multiple servers.

[0078] In this embodiment, the method includes the following steps:

[0079] Step 102: Obtain the interaction information obtained from the interaction between the interactive robot and the target user.

[0080] In this embodiment, the terminal can access a user account database to obtain user account information. The terminal then filters users who meet preset follow-up criteria based on the account information as target users. Specifically, it can select users who have conducted business within a predetermined time period (i.e., newly registered users) from all users based on their account information. For example, users whose account opening date is earlier than their card issuance date and whose account type is Class I can be selected as target users. The terminal can pre-store target interaction strategies, invoke an interaction robot to intelligently interact with target users according to the target interaction strategies, and then obtain and store the voice interaction information during the interaction process. Optionally, the interaction information can be voice interaction information, text chat interaction information, etc., and this embodiment does not limit the scope.

[0081] For example, in this embodiment, each target interaction strategy includes multiple interaction node information, and each interaction node information includes an interaction node identifier and pre-recorded multi-round dialogue. The interaction node identifier can be multiple interaction node identifiers within a preset identifier range, such as nodes 1.1 to 9.1. The interaction node type can include clarification nodes and technical nodes. The interaction node name can include, but is not limited to, opening remarks, preliminary introduction, end_not the person in question, end_busy, and end failed. The dialogue attributes can include, but are not limited to, first round clarification, second round clarification, multiple rounds, and closing remarks. Specifically, interaction node identifier 1.1 can include a clarification node opening remarks, and the first round clarification dialogue can include, but is not limited to, opening remarks such as: "Hello, this is Organization A. Welcome to become our customer. Are you Mr. / Ms. XX?"

[0082] After the interaction begins, the robot uses the pre-recorded first-round script from interaction node 1.1 to interact with the target user and records the current interaction node. Based on pre-stored response samples, the robot determines whether the target user's response matches the sample. If it matches, it jumps to the next interaction node corresponding to the response and uses the pre-recorded first-round script from that node to interact with the target user. This continues until a preset interaction end node is reached, completing the intelligent outbound call interaction. The robot can record the interaction nodes and the voice interaction text file.

[0083] The target interaction strategy may include multiple interaction nodes, and the information of each interaction node includes the interaction node identifier and pre-recorded multi-round scripts; the user account database includes the user's card opening date and account opening date; the interaction information may include interactive voice files, human-computer interaction rounds and interaction end points; the interactive robot may include an outbound call robot, which is an intelligent outbound call program.

[0084] Step 104: Determine the mood index corresponding to the target user based on the interaction information, and extract the intention keywords and interaction end node information from the interaction information.

[0085] In this embodiment, the terminal extracts the interactive voice file from the voice interaction information. Based on the interactive voice file, it uses automatic speech recognition technology to convert the voice information contained in the interactive voice file into text information, obtaining a voice interaction text file. Simultaneously, the terminal analyzes and processes the voice interaction information to obtain the target user's mood index.

[0086] The terminal uses a sentence separation method to recognize the voice interaction text file, distinguishing between the interactive robot's statements and the target user's statements. Specifically, the sentence separation method involves the terminal recognizing the interactive robot's dialogue in the voice interaction text file based on pre-recorded scripts, thus obtaining the interactive robot's statements and the target user's statements.

[0087] The terminal then employs keyword extraction technology from natural language processing to extract the target user's statement and obtain the desired keywords. Subsequently, the terminal categorizes these desired keywords using a preset classification method, resulting in preset types of desired keywords. These desired keyword types can include positive and negative types. Positive types further include: a first positive type and a second positive type; negative types further include: a first negative type and a second negative type.

[0088] In one example, the terminal can pre-store a set of desired keywords for each type. After identifying a desired keyword, the terminal can determine whether the desired keyword exists in one of the desired keyword sets. If it exists, the desired keyword belongs to that desired keyword set. If it does not exist, the terminal calculates the similarity between the identified desired keyword and the desired keywords contained in each desired keyword set, and identifies the desired keyword as the desired keyword type corresponding to the desired keyword set.

[0089] Specifically, the preset classification method can be: based on the degree of positive intention of the intention keywords, they can be divided into a first positive type and a second positive type; based on the degree of negative intention of the intention keywords, they can be divided into a first negative type and a second negative type; wherein, the degree of positive intention of the first positive type intention keywords is greater than that of the second positive type intention keywords; and the degree of negative intention of the first negative type intention keywords is less than that of the second negative type intention keywords.

[0090] For example, if the desired keyword is a word with a positive intention, such as "interested," "good," or "okay," it is added to the first set of positive desired keywords; if the desired keyword is empty (i.e., the target user does not reply or the target user is offline), it is added to the second set of positive desired keywords; if the desired keyword is a word with a negative intention, such as "not use," "not interested," or "not needed," it is added to the first set of negative desired keywords; if the desired keyword is a word with a very negative intention, such as "sick" or "has a problem," it is added to the second set of negative desired keywords.

[0091] The terminal can also identify human-computer interaction nodes in the interactive information and record the interaction nodes before the end as interaction end node information.

[0092] Step 106: Calculate the target user's willingness level data using a preset willingness level algorithm, the weight of each willingness keyword, and the frequency of each willingness keyword.

[0093] In this embodiment, the terminal can store weights corresponding to various types of intention keywords, and can also store intention coefficients corresponding to various types of intention keywords. The weights can be obtained by linear fitting using a pre-stored mapping formula, resource index values ​​of each sample user, the frequency of occurrence of sample intention keywords in each intention keyword set, and the intention coefficients corresponding to each intention keyword set; this will be explained in detail later. The intention coefficients can be pre-set by the terminal based on each intention keyword set. For example, the intention coefficient corresponding to the first positive intention keyword set is set as a first value; the intention coefficient corresponding to the second positive intention keyword set is set as a second value; the intention coefficient corresponding to the first negative intention keyword set is set as a third value; and the intention coefficient corresponding to the second negative intention keyword set is set as a fourth value; wherein the relationship between the first and fourth values ​​is from largest to smallest.

[0094] For each type of desired keyword, the terminal can identify keywords belonging to that type from the identified desired keywords. For each desired keyword in that set, the terminal counts the number of times that keyword appears in the interactive information, and then sums the occurrence counts of all desired keywords in the set to obtain the total occurrence count of that type of desired keyword. Then, the terminal uses a preset desiredness degree algorithm to calculate the desiredness coefficient, the total occurrence count of each type of desired keyword, and the weight of each type of desired keyword to obtain the desiredness degree data of the target user.

[0095] Step 108: Determine the target user's willingness level by using a preset willingness level discrimination algorithm, interaction end node information, mood index, and willingness level data.

[0096] In this embodiment, the terminal determines the target user's willingness level based on interaction end node information, mood index, and willingness level data using a preset willingness level discrimination algorithm.

[0097] The preset intention level discrimination algorithm is as follows: the terminal identifies whether the interaction end node information represents a preset node of interest. If the interaction end node information represents a preset node of interest, the terminal determines that the target user's intention level is interested; if the terminal identifies that the interaction end node information represents a preset node of disinterest, the terminal determines that the interaction end node information represents a preset node of disinterest; if the interaction end node information does not represent a preset node of disinterest, the terminal determines that the target user's intention level is disinterested by judging the target user's intention level index through a preset intention level index calculation method.

[0098] Specifically, the calculation method for the willingness level index is as follows:

[0099] A = B * 0.5 + C

[0100] Among them, A is the willingness level index, B is the mood index, and C is the willingness degree data.

[0101] For example, the terminal identifies the target user's willingness level index. If the willingness level index is greater than 5, the terminal judges the target user's willingness level as interested; if the willingness level index is between 2 and 4, the terminal judges the target user's willingness level as neutral; if the willingness level index is less than 2, the terminal judges the target user's willingness level as uninterested.

[0102] The aforementioned method for identifying intention levels, by analyzing and extracting interaction information, can obtain the user's mood index and intention keywords. By using various types of intention keywords, their corresponding weights, and intention coefficients, the accuracy of user intention analysis can be improved. By calculating the user's intention level data using an intention degree algorithm, and combining this data with the mood index and interaction termination information, the method can accurately and clearly identify the user's intention level during human-computer interaction, thus improving the efficiency of interaction effect evaluation.

[0103] In one embodiment, such as Figure 2 As shown, before step 102, the procedure also includes:

[0104] Step 202: Obtain the target user's account information.

[0105] In this embodiment of the application, the terminal obtains the target user's account information through the target user account information database.

[0106] The target user's account information includes, but is not limited to, preset field data, such as user name, user account type, mobile phone number, card activation date, and account opening date.

[0107] Step 204: Input the target user's account information into the pre-trained interaction strategy selection model to obtain the target interaction strategy corresponding to the target user.

[0108] In this embodiment of the application, the terminal inputs the target user's account information into a pre-trained interaction strategy selection model to obtain the target interaction strategy corresponding to the target user.

[0109] The interaction strategy selection model can be a model trained based on a deep neural network, or a model established based on the mapping relationship between account information and the target interaction strategy. The mapping relationship between account information and the target interaction strategy is a predefined mapping relationship on the terminal.

[0110] For example, the terminal invokes the interaction strategy selection model, compares the card opening date and the account opening date. If the card opening date is less than the account opening date and the account type is Class I, then the target interaction strategy is set to the first interaction strategy; if the card opening date is equal to the account opening date and the account type is Class II, then the target interaction strategy is set to the second interaction strategy; if the card opening date is equal to the account opening date and the account type is Class I, then the target interaction strategy is set to the third interaction strategy.

[0111] Step 206: Based on the target interaction strategy, interact with the target user through an interactive robot.

[0112] In this embodiment, each target interaction strategy includes multiple interaction node information, which may include, but is not limited to, interaction node identifiers, interaction node types, interaction node names, dialogue attributes, and pre-recorded multi-round dialogues. The interaction node identifiers may include multiple interaction node identifiers within a preset identifier range, such as nodes 1.1 to 9.1. The interaction node type may include clarification nodes and technical nodes. The interaction node name may include, but is not limited to, opening remarks, preliminary introduction, end_not the person in question, end_busy, and end failed. The dialogue attributes may include, but are not limited to, first round clarification, second round clarification, multiple rounds, and closing remarks. Specifically, interaction node identifier 1.1 may include a clarification node opening remarks, and the first round clarification dialogue may include, but is not limited to, opening remarks such as: "Hello, this is Organization A. Welcome to become our customer. Are you Mr. / Ms. XX?"

[0113] After the interaction begins, the terminal invokes the interactive robot and uses the pre-recorded first-round script from interaction node 1.1 to interact with the target user. The terminal records the node where the interactive script information is located during the interaction. Based on pre-stored response samples, the interactive robot uses a preset similarity algorithm to determine whether the target user's response matches the content of the response sample. If it matches the preset response sample, it jumps to the next interaction node corresponding to the response. The terminal invokes the interactive robot and uses the pre-recorded first-round script from that interaction node to interact with the target user until a preset interaction end node is reached, completing the intelligent outbound call interaction. If it does not match the preset response sample, it jumps to the preset end node. The interactive robot can record the interaction node identifiers of the interactive scripts and the voice interaction text file. Any algorithm with similarity calculation function can be applied in this application, and this embodiment does not limit this.

[0114] In this embodiment, the target user account information is processed by a pre-trained interaction strategy selection model, which achieves the effect of setting a target interaction strategy that is more suitable for the target user's conditions.

[0115] In one embodiment, step 104 includes:

[0116] The interaction information is analyzed and processed based on a preset sentiment analysis algorithm to obtain the mood index corresponding to the target user.

[0117] In this embodiment, the terminal processes the voice interaction text file by calling a preset sentiment analysis algorithm to obtain a mood description text for the voice interaction text file. The terminal pre-stores the correspondence between each mood description text and a mood index. The terminal then uses the sentiment analysis algorithm to convert the voice interaction text file into mood description text, obtaining the corresponding mood index for that file. Specifically, the mood index is a preset range of values, reflecting the user's mood from low to high; a higher mood index indicates a better user mood.

[0118] Among them, the sentiment analysis algorithm can be based on Naive Bayes, deep learning LSTM (Long Short-Term Memory model), and pre-trained VADER (Valence-Aware Dictionary and Threaction Reasoner) model.

[0119] In this embodiment, the voice interaction text is analyzed and processed using a sentiment analysis algorithm. The numerical value from 0 to 5 is used to intuitively reflect the user's mood index, which facilitates the subsequent calculation of the level of willingness.

[0120] In one embodiment, such as Figure 3 As shown, step 106 includes:

[0121] Step 302: For each desired keyword, determine the product of the frequency of occurrence of the desired keyword and the desired keyword's desired coefficient to obtain the first calculation result.

[0122] In this embodiment, the willingness coefficient can be a willingness coefficient preset by the terminal based on each set of willingness keywords. For example, the willingness coefficient corresponding to the first set of positive willingness keywords is set as the first value; the willingness coefficient corresponding to the second set of positive willingness keywords is set as the second value; the willingness coefficient corresponding to the first set of negative willingness keywords is set as the third value; and the willingness coefficient corresponding to the second set of negative willingness keywords is set as the fourth value; wherein the relationship between the first value and the fourth value is from large to small.

[0123] The terminal then uses a preset calculation method to calculate the frequency of occurrence of each type of keyword.

[0124] The terminal obtains the product of the frequency of occurrence of each desired keyword and the desired keyword's desired coefficient, as the first calculation result for that desired keyword. For example, the terminal obtains the product of the frequency of occurrence of a first-positive-type desired keyword and the desired coefficient of a first-type positive desired keyword, as the first calculation result for the first-positive-type desired keyword; the terminal obtains the product of the frequency of occurrence of a second-positive-type desired keyword and the desired coefficient of a second-positive-type desired keyword, as the first calculation result for the second-positive-type desired keyword, and so on, to obtain the first calculation result for each type of desired keyword.

[0125] In this application, any algorithm with the function of calculating the number of items can be used for the preset calculation method, and the embodiments of this application do not limit it.

[0126] Step 304: The first calculation result corresponding to each intention keyword is weighted and summed with the weight corresponding to each intention keyword to obtain the intention level data of the target user.

[0127] In this embodiment of the application, the terminal calculates the target user's degree of willingness data by using a preset willingness degree algorithm, based on the weights of each type of willingness keyword obtained in advance and the first calculation results corresponding to each willingness keyword, and by using weighted summation calculation.

[0128] Specifically, the preset algorithm for the degree of willingness is as follows:

[0129] γ = P1*a1 + P2*a2 + N1*a3 + N2*a4

[0130] Where γ represents the degree of willingness data, P1 represents the first calculation result of the first positive type of willingness keyword, P2 represents the first calculation result of the second positive type of willingness keyword, N1 represents the first calculation result of the first negative type of willingness keyword, N2 represents the first calculation result of the second negative type of willingness keyword, and a1, a2, a3, and a4 represent the weights corresponding to each type of willingness keyword.

[0131] In this embodiment, the terminal uses a preset intention degree algorithm to perform a weighted summation calculation on the weights corresponding to each type of intention keyword, the number of times each type of keyword appears, and the intention coefficients corresponding to each type of intention keyword. By introducing weights, a more balanced intention degree data can be obtained.

[0132] In one embodiment, such as Figure 4 As shown, it also includes:

[0133] Step 402: Obtain sample interaction information between each sample user and the interactive robot.

[0134] In this embodiment, the terminal acquires all interaction information between each sample user and the interactive robot, which is used as sample interaction information. The sample users can be randomly selected by the terminal accessing an account database.

[0135] Step 404: Extract sample intention keywords from sample interaction information, classify the sample intention keywords, and obtain the intention keyword set corresponding to each type.

[0136] In this embodiment, the terminal extracts the sample interaction voice file from the sample interaction information, and uses automatic speech recognition (ASR) technology to process the sample interaction voice file into text to obtain the sample voice interaction text file.

[0137] The terminal recognizes the sample voice interaction text file, distinguishing between the interactive robot's statements and the sample user statements, resulting in multiple sample user statements. The terminal then employs keyword extraction technology from Natural Language Processing (NLP) to extract keywords from the sample user statements, obtaining individual sample intention keywords. Subsequently, the terminal categorizes these intention keywords using a preset classification method, resulting in sets of intention keywords corresponding to each type. For example, the preset classification method could be: categorizing intention keywords into a first positive type and a second positive type based on their degree of positive intention; or categorizing them into a first negative type and a second negative type based on their degree of negative intention; wherein, the first positive type intention keywords have a higher degree of positive intention than the second positive type, and the first negative type intention keywords have a lower degree of negative intention than the second negative type.

[0138] For example, if the desired keyword is a word with a positive intention such as "interested," "good," or "okay," it is added to the first set of positive desired keywords; if the desired keyword is empty, it is added to the second set of positive desired keywords; if the desired keyword is a word with a negative intention such as "not use," "not interested," or "don't need," it is added to the first set of negative desired keywords; if the desired keyword is a word with a very negative intention, such as "sick" or "has a problem," it is added to the second set of negative desired keywords.

[0139] For the preset judgment method, any algorithm that has the function of judging the same two can be applied in this application, and the embodiments of this application do not limit it.

[0140] Step 406: Based on the pre-stored mapping relationship, the resource index values ​​of each sample user, the frequency of occurrence of sample intention keywords in each intention keyword set, and the intention coefficient corresponding to each intention keyword set, perform linear fitting to obtain the weights corresponding to each intention keyword set.

[0141] The mapping relationship is used to reflect the mapping relationship between resource indicator values, the number of occurrences of sample intention keywords contained in each intention keyword set, the intention coefficient corresponding to each intention keyword set, and the weight corresponding to each intention keyword set.

[0142] In this embodiment, the terminal obtains the weights of each set of desired keywords by performing linear fitting based on the pre-stored mapping formula, the resource index values ​​of each sample user, the number of occurrences of each type of desired keyword set, and the desired coefficients of each set of desired keywords.

[0143] Specifically, the pre-stored mapping relationship is as follows:

[0144] R=P1*2*a1+P2*1*a2+N1*(-1)*a3+N2*(-2)*a4

[0145] Where R is the resource indicator value, P1 is the number of times the first positive intention keyword appears, P2 is the number of times the second positive intention keyword appears, N1 is the number of times the first negative intention keyword appears, N2 is the number of times the second negative intention keyword appears, and a1, a2, a3, and a4 are the weights corresponding to each type of intention keyword.

[0146] For each type of desired keyword, the terminal can input the target user's resource indicator value, the frequency of occurrence of the desired keyword, and the desired coefficient of the desired keyword into the above mapping formula to obtain a mapping formula for the target user's resource indicator value, the frequency of occurrence of the desired keyword, the desired coefficient of the desired keyword, and the weights corresponding to each set of desired keywords. The terminal performs linear fitting on multiple sets of mapping formulas to obtain the weights corresponding to each set of desired keywords. Optionally, the above mapping formula is only an example provided by the embodiments of this application, and the parameter values ​​in the mapping formula can be set by those skilled in the art based on experience. The embodiments of this application do not limit the mapping formula.

[0147] Step 408: Establish the correspondence between the desired keywords and their weights based on the weights of each set of desired keywords.

[0148] In this embodiment, the terminal sets the weight of each sample keyword in each set of desired keywords to the weight of the corresponding set of desired keywords. The terminal sets the weight of each sample keyword to the weight of the desired keyword of the corresponding type. Specifically, the terminal can pre-select and store sets of desired keywords of each type. After identifying a desired keyword, the terminal can determine whether the desired keyword exists in one of the desired keyword sets. If it exists, the desired keyword belongs to that set. If it does not exist, the terminal calculates the similarity between the identified desired keyword and the desired keywords contained in each set of desired keywords, and identifies the desired keyword as the desired keyword type corresponding to the set of desired keywords.

[0149] In this embodiment, the terminal performs linear fitting on the resource index values ​​of each sample user, the number of occurrences of each type of desired keyword, and the desired coefficient corresponding to each type of desired keyword through a pre-stored mapping formula. This yields the weight of each type of desired keyword, which serves as the weight for subsequent weighted summation calculations.

[0150] In one embodiment, prior to step 406, the method further includes:

[0151] We obtain account resource data from sample users, process the data using resource analysis algorithms, and obtain resource indicator values.

[0152] In this embodiment, the terminal obtains the account resource data of each sample user through a sample user account database, processes the account resource data using a resource analysis algorithm, and thus determines the resource indicator value. For example, the account resource data may include the sample user's average daily resource value and product quantity. The terminal calls the resource analysis algorithm to calculate the resource indicator value of the sample user. Specifically, the terminal uses a resource value calculation method to obtain the resource value of each sample user, and then uses a mapping algorithm to map all resource values ​​to a preset numerical range, for example, mapping all resource values ​​to numbers from 0 to 4. The mapped values ​​are used as the sample user's resource indicator value.

[0153] R = Q / 10000 + P

[0154] Where R is the resource value, Q is the average daily resource value, and P is the product quantity.

[0155] In this embodiment, the terminal obtains the resource index values ​​of the sample users through a resource analysis algorithm, which facilitates subsequent weight calculation.

[0156] In one embodiment, such as Figure 5 As shown, after step 108, the following steps are also included:

[0157] Step 502: Based on the interaction information, identify the interaction results of each type to obtain the identification results.

[0158] In this embodiment, after determining the target user's willingness level, the terminal identifies the interaction result based on the interaction end node information. The interaction results can include target interaction strategy type, follow-up result type, whether the user is the actual user, and problem record type. For example, if the terminal identifies the interaction end node information as node 4.1, the follow-up result is assigned a success value, and whether the user is the actual user is assigned a non-actual user value; if the terminal identifies the interaction end node information as node 4.2, the follow-up result is assigned a failure value, and the problem record is an incorrect phone number; if the terminal identifies the interaction end node information as node 6.1, 7.1, 8.1, or 9.1, the follow-up result is assigned a success value, and whether the user is the actual user is assigned a "yes" value.

[0159] The terminal displays the identification results based on the various types of interaction results. These interaction results can include the target interaction strategy type, successful / failed callback, whether the person is the correct individual, and issue records. For example, issue records can include whether the phone number is incorrect and whether the interaction was successfully initiated.

[0160] Step 504: Based on the recognition results, generate target interaction strategy feedback information corresponding to the target user.

[0161] In this embodiment of the application, the terminal generates target interaction strategy feedback information corresponding to the target user based on the identification result. The feedback information may include target interaction strategy type, target user account information, whether it is the user, success / failure of return visit, problem record, and other information.

[0162] In this embodiment, after the terminal determines the target user's level of willingness, it identifies and marks the follow-up results to facilitate subsequent processing by staff.

[0163] For example, after the terminal generates feedback information, if the terminal identifies the target user's target interaction strategy as the first interaction strategy and the feedback information is "failed to return visit", then the target user is marked as a user to be manually verified; if the terminal identifies the target user's target interaction strategy as the second interaction strategy and the feedback information is "successful return visit" and "interest level", then the target user will be pushed SMS service, etc.

[0164] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0165] Based on the same inventive concept, this application also provides a willingness-level identification device for implementing the willingness-level identification method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more willingness-level identification device embodiments provided below can be found in the limitations of the willingness-level identification method described above, and will not be repeated here.

[0166] In one embodiment, such as Figure 6 As shown, a willingness-level recognition device 600 is provided, comprising: a first acquisition module 602, an extraction module 604, a calculation module 606, and a determination module 608, wherein:

[0167] The first acquisition module 602 is used to acquire interaction information obtained by the interactive robot interacting with the target user.

[0168] Extraction module 604 is used to determine the mood index corresponding to the target user based on the interaction information, and extract various types of intention keywords and interaction end node information from the interaction information;

[0169] The calculation module 606 is used to calculate the target user's willingness level data by using a preset willingness level algorithm, the weights corresponding to each type of willingness keyword, and the number of times each type of willingness keyword appears;

[0170] The determination module 608 is used to determine the intention level of the target user by using a preset intention level discrimination algorithm, the interaction end node information, the mood index and the intention degree data.

[0171] The aforementioned intention level recognition device can obtain the user's mood index and intention keywords by analyzing and extracting interaction information. By using various types of intention keywords, their corresponding weights, and intention coefficients, the accuracy of user intention analysis can be improved. The intention level algorithm calculates the user's intention level data, and combining this data with the mood index and interaction termination information can accurately and clearly identify the user's intention level during human-computer interaction, thus improving the efficiency of interaction effect evaluation.

[0172] In one embodiment, the apparatus further includes:

[0173] The second acquisition module is used to acquire the target user's account information;

[0174] The input module is used to input the account information of the target user into a pre-trained interaction strategy selection model to obtain the target interaction strategy corresponding to the target user;

[0175] The interaction module interacts with the target user through the interactive robot, based on the target interaction strategy.

[0176] In one embodiment, the extraction module 604 is specifically used for:

[0177] The interaction information is analyzed and processed based on a preset sentiment analysis algorithm to obtain the mood index corresponding to the target user.

[0178] In one embodiment, the computing module 606 is specifically used for:

[0179] Based on the preset intention level algorithm, the frequency of occurrence of each intention keyword, the intention coefficient of each intention keyword, and the weight corresponding to each intention keyword, the intention level data of the target user is obtained by weighted summation.

[0180] In one embodiment, the apparatus further includes:

[0181] The third acquisition module is used to acquire sample interaction information between each sample user and the interactive robot;

[0182] The extraction module is used to extract sample intention keywords from the sample interaction information;

[0183] The classification module is used to classify the sample's intention keywords to obtain a set of intention keywords corresponding to each type;

[0184] The fitting module is used to perform linear fitting based on a pre-stored mapping formula, the resource index values ​​of each sample user, the frequency of occurrence of sample intention keywords contained in each intention keyword set, and the intention coefficient corresponding to each intention keyword set, to obtain the weight corresponding to each intention keyword set; wherein, the mapping formula is used to reflect the mapping relationship between the resource index values, the frequency of occurrence of each type of intention keyword, the intention coefficient corresponding to each type of intention keyword, and the weight corresponding to each type of intention keyword;

[0185] A module is established to establish a correspondence between desired keywords and their corresponding weights based on the weights of each set of desired keywords.

[0186] In one embodiment, the apparatus further includes:

[0187] The fourth acquisition module is used to acquire the account resource data of sample users, and process the account resource data through a resource analysis algorithm to obtain resource indicator values.

[0188] In one embodiment, the apparatus further includes:

[0189] The recognition module is used to recognize the various types of interaction results based on the interaction information, and obtain the recognition results;

[0190] The generation module is used to generate target interaction strategy feedback information corresponding to the target user based on the recognition results.

[0191] Each module in the aforementioned intention-level recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.

[0192] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a willingness-based identification method. The display screen can be an LCD screen or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0193] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0194] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0195] Acquire interaction information obtained from the interaction between the interactive robot and the target user; the interaction information includes interactive information;

[0196] Based on the interaction information, the mood index corresponding to the target user is determined, and various types of intention keywords and interaction end node information are extracted from the interaction information.

[0197] The intention level data of the target user is calculated by using a preset intention level algorithm, the weights corresponding to each type of intention keyword, and the frequency of occurrence of each type of intention keyword;

[0198] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0199] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0200] Acquire interaction information obtained from the interaction between the interactive robot and the target user; the interaction information includes interactive information;

[0201] Based on the interaction information, the mood index corresponding to the target user is determined, and various types of intention keywords and interaction end node information are extracted from the interaction information.

[0202] The intention level data of the target user is calculated by using a preset intention level algorithm, the weights corresponding to each type of intention keyword, and the frequency of occurrence of each type of intention keyword;

[0203] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0204] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0205] Acquire interaction information obtained from the interaction between the interactive robot and the target user; the interaction information includes interactive information;

[0206] Based on the interaction information, the mood index corresponding to the target user is determined, and various types of intention keywords and interaction end node information are extracted from the interaction information.

[0207] The intention level data of the target user is calculated by using a preset intention level algorithm, the weights corresponding to each type of intention keyword, and the frequency of occurrence of each type of intention keyword;

[0208] The target user's willingness level is determined by using a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data.

[0209] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0210] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0211] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0212] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method of identifying a level of intent, characterized by, The method comprises: obtaining interaction information obtained by an interactive robot interacting with a target user, the interaction information comprising an interaction voice file and an interaction end node; determining a mood index corresponding to the target user based on the interaction information, and extracting each intention keyword and interaction end node information from the interaction information; calculating intention degree data of the target user by a preset intention degree algorithm, a weight corresponding to each intention keyword, and a number of occurrences of each intention keyword; determining an intention level of the target user by a preset intention level discrimination algorithm, the interaction end node information, the mood index, and the intention degree data. The method further comprises: obtaining sample interaction information of each sample user and the interactive robot; extracting sample intention keywords in the sample interaction information, classifying the sample intention keywords, and obtaining an intention keyword set corresponding to each type; performing linear fitting processing on a mapping relationship formula, resource index values of each sample user, a number of occurrences of sample intention keywords included in each intention keyword set, and an intention coefficient corresponding to each intention keyword set, to obtain a weight corresponding to each intention keyword set, wherein the mapping relationship formula is used to reflect a mapping relationship among the resource index values, the number of occurrences of sample intention keywords included in each intention keyword set, the intention coefficient corresponding to each intention keyword set, and the weight corresponding to each intention keyword set; establishing a corresponding relationship between intention keywords and weights according to the weight corresponding to each intention keyword set.

2. The method of claim 1, wherein, Before the obtaining of the interaction information, the method further comprises: obtaining account information of the target user; inputting the account information of the target user into a pre-trained interaction strategy selection model to obtain a target interaction strategy corresponding to the target user; interacting with the target user through the interactive robot based on the target interaction strategy.

3. The method of claim 1, wherein, The determination of the mood index corresponding to the target user based on the interaction information comprises: performing analysis and processing on the interaction information based on a preset sentiment analysis algorithm to obtain the mood index corresponding to the target user.

4. The method of claim 1, wherein, The calculation of the intention degree data of the target user by a preset intention degree algorithm, a weight corresponding to each intention keyword, and a number of occurrences of each intention keyword comprises: for each intention keyword, determining a product of the number of occurrences of the intention keyword and an intention coefficient of the intention keyword to obtain a first calculation result; performing weighted summation calculation on the first calculation result corresponding to each intention keyword and the weight corresponding to each intention keyword to obtain the intention degree data of the target user.

5. The method of claim 1, wherein, The method further comprises: obtaining account resource data of a sample user, and processing the account resource data by a resource analysis algorithm to obtain resource index values.

6. The method of claim 1, wherein, The method further includes: Based on the interaction information, identifying each type of interaction result to obtain an identification result; Based on each identification result, generating target interaction strategy feedback information corresponding to the target user.

7. A will level identifying apparatus characterized by comprising: The apparatus includes: A first obtaining module configured to obtain interaction information obtained by an interactive robot interacting with a target user, the interaction information including an interactive voice file and an interaction end node; An extracting module configured to determine a mood index corresponding to the target user based on the interaction information, and extract each willingness keyword and interaction end node information from the interaction information; A calculating module configured to calculate willingness degree data of the target user by a preset willingness degree algorithm, a weight corresponding to each willingness keyword, and a number of occurrences of each willingness keyword; A determining module configured to determine a willingness level of the target user by a preset willingness level discrimination algorithm, the interaction end node information, the mood index, and the willingness degree data; A third obtaining module configured to obtain sample interaction information of each sample user and the interactive robot; The extracting module is further configured to extract sample willingness keywords in the sample interaction information, classify the sample willingness keywords, and obtain a set of willingness keywords corresponding to each type; A fitting module configured to perform linear fitting processing on each set of willingness keywords according to a mapping relationship formula, resource index values of each sample user, a number of occurrences of sample willingness keywords included in each set of willingness keywords, and a willingness coefficient corresponding to each set of willingness keywords, to obtain a weight corresponding to each set of willingness keywords; wherein the mapping relationship formula is used to reflect a mapping relationship between the resource index values, the number of occurrences of sample willingness keywords included in each set of willingness keywords, the willingness coefficient corresponding to each set of willingness keywords, and the weight corresponding to each set of willingness keywords; According to the weight corresponding to each set of willingness keywords, a correspondence between a willingness keyword and a weight is established.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images