An AI intelligent calling method

Abstract

The present invention provides an AI intelligent calling method, comprising the steps of: obtaining voice content when promoting a current product to a current user, obtaining a call time period, purchase intention, and the current user's call scenario based on the voice content; obtaining a user profile of the current user, correlating the above analysis content, and obtaining a voice tag; when it is necessary to promote a target product, matching a first target tag from all voice tags, determining whether the number of current users in the first target tag reaches a target number, and if so, using the current user in the first target tag as the target user; otherwise, matching a second target tag based on the user profile and the call scenario, and combining the two as the target user; calling the target user based on the call time period in the target tag corresponding to the target user to promote the target product. The present invention can ensure the accuracy and communication rate of voice calls.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to an AI intelligent calling method. Background Art

[0002] Due to the high labor costs associated with traditional telemarketing, intelligent telemarketing systems based on intelligent voice interaction technology are gradually replacing traditional manual dialing. However, existing technologies still have many limitations in practical applications, which restrict their efficiency and user experience.

[0003] With the rise of big data and artificial intelligence (AI), researchers are beginning to explore AI technologies to address issues such as low voice recognition accuracy and limited emotional input in smart telemarketing systems, thereby increasing user acceptance of intelligent voice marketing. However, existing technologies remain unable to address low call completion rates and poor accuracy. Summary of the Invention

[0004] In order to solve the above problems in the prior art, the present invention provides an AI intelligent calling method to improve the accuracy and communication rate of voice calls.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] In a first aspect, the present invention provides an AI intelligent calling method, comprising the steps of:

[0007] Acquire the voice content when promoting the current product to the current user, and obtain the call time period, purchase intention, and call scenario of the current user based on the voice content;

[0008] Obtaining a user profile of the current user, and associating the current product, the call time period, the purchase intention, the call scenario, and the user profile with the current user to obtain a voice tag;

[0009] When it is necessary to promote a target product, a voice tag whose category of the current product is the same as that of the target product and whose purchase intention reaches the purchase recommendation level is matched from all voice tags as a first target tag. When it is determined whether the number of current users in the first target tag reaches the target number, if so, the current user in the first target tag is used as the target user. Otherwise, after removing the voice tags that are the same as the current user in the first target tag from all voice tags, a second target tag is matched based on the user portrait and the call scenario from the remaining behavior tags. The current user in the second target tag and the current user in the first target tag are used together as target users.

[0010] The target promotion time of the target user is determined according to the call time period in the target tag corresponding to the target user, and the target user is called during the target promotion time to promote the target product.

[0011] The beneficial effects of the present invention are as follows: by analyzing user call content in real time, the present invention dynamically identifies their call scenarios and purchase intentions, generates voice tags containing the product, scenario, time period, and profile, and prioritizes users with matching purchase intentions for similar products, achieving precise user matching. Furthermore, the present invention recommends promotional times based on the user's historical scenarios and time periods, avoiding blind calls and thus improving the accuracy and communication rate of voice calls. Furthermore, when the target population is insufficient, the present invention expands the number of users whose profiles and scenarios match, ensuring not only coverage of the target users but also the accuracy and communication rate of voice calls.

[0012] Optionally, the method further comprises the steps of:

[0013] When the target product has an associated scene, the voice tag containing the associated scene and the purchase intention reaching the purchase recommendation level will also be used as the first target tag.

[0014] According to the above description, scene association recognition can enhance the relevance between target products and scenes, thereby improving conversion rates.

[0015] Optionally, matching a second target tag according to the user portrait and the call scenario includes the steps of:

[0016] The call scenes that appear at a preset frequency ratio in all first target tags are selected as candidate scenes, and the voice tags that match the user portrait and the target product in the remaining behavior tags, whose call scenes belong to the candidate scenes and whose purchase intention reaches the purchase recommendation level are selected as second target tags.

[0017] According to the above description, we can count high-frequency scenarios and screen users whose portraits match the products in the same scenarios to ensure that the expanded target users receive sales calls in scenarios where they have the intention to buy, thereby ensuring the accuracy and communication rate of voice calls.

[0018] Optionally, the method further comprises the steps of:

[0019] When there is a new user, obtain the user portrait of the new user, match the voice tag corresponding to the user portrait of the new user from all voice tags, the category of the current product is the same as the category of the target product, and the purchase intention reaches the purchase recommendation level as the third target tag, filter out the call scene with the highest frequency from all third target tags as the alternative scene, obtain the call time period of the current user with the same user portrait as the new user in the alternative scene, use the time with the highest frequency in all the obtained call time periods as the new promotion time for the new user, and call the new user at the new promotion time to promote the target product.

[0020] According to the above description, it is necessary to clarify the scenario category rather than relying on the time points in the user's historical behavior data. When a new user has no historical records, first determine the scenario in which the user portrait is most suitable for promotion, and then determine the time point corresponding to the user portrait in the scenario suitable for promotion to carry out promotion, thereby solving the cold start problem of new users and improving the communication rate of new users.

[0021] Optionally, the step of using the time with the highest frequency among all acquired call time periods as the new promotion time for the new user comprises the following steps:

[0022] The whole day is divided into multiple time windows, and the occurrence frequency of all acquired call time periods in each time window is calculated. The time window with the highest occurrence frequency is used as the new promotion time for the new user.

[0023] Optionally, the step corresponding to the user profile of the new user includes:

[0024] If the tag set in the user portrait of the new user is included in the tag set of the user portrait in the voice tag, then the user portrait in the voice tag corresponds to the user portrait of the new user.

[0025] Optionally, the method further comprises the steps of:

[0026] If, after the call time node of the first target tag, the same voice tag as the current user of the first target tag appears for a preset number of consecutive times and the purchase intention reaches the minimum communication level, the priority level of the first target tag will be adjusted to the second target tag or lower than the second target tag.

[0027] According to the above description, by dynamically optimizing call priorities and allocating resources to high-willing users, the overall conversion efficiency is improved, and the accuracy and communication rate of voice calls are guaranteed.

[0028] Optionally, the purchase intention includes a product purchase tendency or a user communication intention or all of the above.

[0029] Optionally, the identification of the call scenario includes at least one of the following methods:

[0030] Determine the call scenario by analyzing keywords in the voice content;

[0031] Alternatively, the call scene is determined by analyzing the background noise in the voice content;

[0032] Alternatively, the call scenario may be determined by analyzing keywords in the voice content and background noise.

[0033] Optionally, the method further comprises the steps of:

[0034] If the current users in the first target tags overlap, all first target tags of the same current user are summarized into one first target tag;

[0035] If the current users in the second target tags overlap, all second target tags of the same current user are summarized into one second target tag. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flowchart of an AI intelligent calling method according to the first embodiment of the present invention;

[0037] Figure 2 This is a flow chart of an AI intelligent calling method according to the second embodiment of the present invention. DETAILED DESCRIPTION

[0038] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. Instead, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0039] Example 1

[0040] This embodiment is applied to scenarios where AI voice robots are used for sales promotion. Existing technologies still cannot improve the low call completion rate and poor accuracy. This embodiment uses dynamic scene recognition and user profiling to construct voice tags, achieving not only precise user matching but also improving call completion rate. For details, please refer to the following description of this embodiment.

[0041] Among them, the communication rate refers to the ratio of users' willingness to communicate. Compared with the connection rate, it can better reflect the user's emotional tendency towards telephone sales, which is the basis for accepting telephone sales. Therefore, improving the call communication rate can ultimately improve the conversion rate of product sales.

[0042] Please refer to Figure 1 , an AI intelligent calling method, comprising the steps of:

[0043] S1. Acquire the voice content when promoting the current product to the current user, and obtain the call time period, purchase intention, and call scenario of the current user based on the voice content.

[0044] The call time period refers to the start and end time of the call, such as 08:37-08:39 on January 24, 2025.

[0045] In this embodiment, purchase intention includes product purchase propensity and user communication willingness. Product purchase propensity is assessed based on whether the user inquires about features, price, or whether the user actually purchases the product, including but not limited to questions like "It's suitable for oily skin," "How much does it cost?", etc. Product purchase propensity can be categorized into the following five levels: clearly expressing purchase intention or ultimately purchasing the product, inquiring in detail about the product's price and features, simply responding or asking irrelevant questions, essentially no response or a perfunctory response, and clearly expressing refusal. User communication willingness is assessed based on whether the user is willing to communicate, the quality of the user's response, or the user's emotional response, including but not limited to: hanging up the phone as soon as they hear a sales pitch, using interjections like "oh" or "hmm," and responding with a negative emotional tone. User communication willingness can be categorized into the following five levels: positive emotional response, neutral emotional response with a simple response, neutral emotional response but perfunctory tone, negative emotional response, and no response or quick hanging up. These levels are assigned a score, for example, from 5 to 1, from high to low.

[0046] At this point, the two can be combined to calculate the final purchase intention through weighted score calculation. Considering that the purpose of the call is to promote the product, the weight value of the product purchase intention is higher than the user's communication intention. The weighted scores in this embodiment are: product purchase intention is 70%, and user communication intention is 30%. The corresponding levels are also five levels, namely very high, high, normal, low, and very low, with corresponding scores of [4.5, 5], [3.5-4.5), [2.5-3.5), [1.5-2.5), and [0-1.5]. At this time, assuming that a "car charger" is being promoted, and the voice content includes "Does it support fast charging?" and "Can the price be cheaper?", the product purchase intention score is 4 points. At this time, the user's response is positive, so the user's communication intention is 5 points. The combined score of the two is 4.3 points, which means the purchase intention level is high.

[0047] In other embodiments, the purchase intention may include a tendency to purchase a product or a user's communication intention.

[0048] In this embodiment, the identification of the call scene includes at least one of the following methods:

[0049] Determine the call scenario by analyzing keywords in the voice content;

[0050] Or determine the call scenario by analyzing the background noise in the voice content;

[0051] Or you can determine the call scenario by analyzing keywords in the voice content and background noise.

[0052] Among them, keywords can be preset in the scene keyword library, such as "driving" is a driving scene, "meeting" is an office scene, "cooking" is a home scene, and so on. Background noise can be spectrally analyzed through MFCC feature extraction. For example, the sound of an operating lever, a loud driving sound, etc. are driving sounds, the sound of a keyboard is an office scene, and the sound of a range hood is a home scene. It should be noted that the description of the above-mentioned call scene is only an example. The existing scene recognition algorithm will comprehensively judge all the sound features to distinguish the scenes in which the user is making a call. For example, the sound of the scene where the chef is at work and the user is in the kitchen is not just the sound of the range hood, but also based on whether there are children's voices, other chefs' voices or the overall stove sound in the ambient sound.

[0053] S2. Obtain the user profile of the current user, associate the current product, call time period, purchase intention, call scenario, and user profile with the current user, and obtain a voice tag.

[0054] User profile tags include age, gender, occupation, location, consumer preferences, and more. Voice tags store the user ID for the current user, and the product ID and category for the current product. Call scenarios are differentiated not only based on the user's environment, but also on weather and ambient noise.

[0055] Here's an example of a speech tag:

[0056] {

[0057] "User ID": "U000001",

[0058] "Product ID": "S000001",

[0059] "Product Category": {"Automotive Supplies","Electronics"},

[0060] "Call Time": "08:37-08:39",

[0061] "Purchase intention": "High",

[0062] "Call Scenario": {"Driving Scenario","Rainy Scenario"},

[0063] “User portrait”: {“Age”: “20-30”, “Gender”: “Female”, “Occupation”: “Office worker”, “Geographic location”: “Fuzhou City, Fujian Province”, “Consumer preferences”: “Cosmetic products”, “Travel products”}.

[0064] S3. When it is necessary to promote the target product, a voice tag whose category of the current product is the same as that of the target product and whose purchase intention reaches the purchase recommendation level is matched from all voice tags as the first target tag. When judging whether the number of current users in the first target tag reaches the target number, if so, the current user in the first target tag is used as the target user. Otherwise, after removing the voice tags that are the same as the current user in the first target tag from all voice tags, a second target tag is matched from the remaining behavior tags based on the user portrait and the call scenario, and the current user in the second target tag and the current user in the first target tag are used together as target users.

[0065] In this embodiment, the steps are further included:

[0066] If there is overlap in the current users in the first target tags, all first target tags of the same current user are summarized into one first target tag;

[0067] If there is overlap in the current users in the second target tags, all the second target tags of the same current user are summarized into one second target tag.

[0068] Among them, the voice tag is saved based on the voice call. Therefore, there are multiple first target tags or second target tags with the same user. In this case, deduplication is required. In this embodiment, it is summarized into one target tag, and the most recent target tag can be retained.

[0069] In this embodiment, matching the second target tag based on the user portrait and the call scenario includes the following steps:

[0070] The call scenes that appear at a preset frequency ratio in all first target tags are selected as candidate scenes, and the voice tags that match the user portrait and the target product in the remaining behavior tags, whose call scenes belong to the candidate scenes and whose purchase intention reaches the purchase recommendation level are selected as the second target tags.

[0071] Among them, the preset frequency ratio is the percentage of the occurrence frequency of candidate scenes to the total number of scenes. In this embodiment, it is not 30%. Of course, when the occurrence frequency is more dispersed, the ratio can be appropriately lowered. Similarly, the ratio can be appropriately increased. The purpose is to count high-frequency scenes.

[0072] Among them, the user profile has a label, and the target product has a product category. Therefore, in this embodiment, the user profile and the target product correspond to each other in that there is a corresponding relationship between the user profile label and the target product category. For example, if the user profile label shows the occupation as driver, and the product category of "car charger" is "automotive supplies", then the two have a corresponding relationship; if the user profile label shows the occupation as mother, and the product category of "milk powder" is "maternal and child products", then the two have a corresponding relationship, and so on. Therefore, this embodiment can screen users whose profiles match the products in the same scenario, ensuring that the expanded target users receive sales calls in scenarios where they have purchasing intentions, thereby ensuring the accuracy and communication rate of voice calls.

[0073] Among them, there are historical voice tags stored in the database. The purchase recommendation level of this embodiment is normal. At this time, when the target product is "car fragrance", the category of the current product is "automotive supplies". Therefore, first match the voice tag whose current product category is "automotive supplies" and whose purchase intention reaches normal from all voice tags as the first target tag. If the target number of requirements is 12,000, and there are only 7,500 first target tags, then the call scene with a frequency of 30% among all first target tags is "driving scene", and the candidate scene is "driving scene". Finally, match the voice tag whose user profile corresponds to "automotive supplies", whose call scene belongs to "driving scene", and whose purchase intention reaches normal from the remaining behavior tags as the second target tag.

[0074] In this embodiment, step S3 further includes the following steps:

[0075] When the target product has an associated scenario, the voice tag that contains the associated scenario and the purchase intention reaches the purchase recommendation level will also be used as the first target tag.

[0076] For example, when the target product is "car fragrance," the associated scene is "driving scene." Therefore, a voice tag containing "driving scene" and indicating normal purchase intent is also selected as the first target tag. Other related relationships include "keyboard" and "office scene." Therefore, scene association recognition can enhance the relevance of the target product and scene, improving conversion rates.

[0077] Specifically, this embodiment of the method for associating products and scenarios is different from the traditional rule engine matching solution or machine learning model recognition solution. This embodiment specifically includes the following steps:

[0078] Add the score for calculating purchase intention to the voice tag;

[0079] Among all the voice tags, the average score of the purchase intention of the first product category in the first call scenario and the average score of the communication intention in scenarios other than the first call scenario are calculated, and the difference between the two average scores is used as the scenario causal score. If the scenario causal score is greater than the first preset causal score, an association relationship is established between the first product category and the first call scenario.

[0080] Among them, the average score of purchase intention of the first product category in the first call scenario refers to the average of the purchase intention scores in the voice tags that include the first call scenario and the first product category, and the average score of communication intention in scenarios other than the first call scenario refers to the average of the purchase intention scores in the voice tags that do not include the first call scenario but include the first product category.

[0081] In this embodiment, the first preset causal score is 0.8. For example, the first product category, "Rain Gear," has an average score of 3.68 in the first call scenario, "Rainy Day," and an average score of 2.45 in other scenarios, a difference of 1.2 points. Therefore, an association between "Rain Gear" and "Rainy Day" is established. It should be noted that this embodiment illustrates the association between products and scenarios for ease of understanding, and does not imply that all associations derived from the causal dependency relationships herein are directly accessible to users.

[0082] Therefore, the rainy day scenario used as an example in this embodiment is typically less common, and this type of data occurs less frequently. The second target tag, however, is designed to match target users in high-frequency scenarios. Therefore, this embodiment quantifies the influence of different scenarios on user purchase decisions through causal dependencies, thereby distinguishing true causal relationships from false correlations and improving recommendation robustness. Furthermore, in low-frequency but high-causal dependency scenarios, the matching scheme for the second target tag can complement each other, improving user matching accuracy and ensuring call completion rates.

[0083] In this embodiment, the steps are further included:

[0084] If after the call time node of the first target tag, the voice tag identical to the current user of the first target tag appears for a preset number of consecutive times and the purchase intention reaches the minimum communication level, the priority level of the first target tag will be adjusted to the second target tag or lower than the second target tag.

[0085] Among them, the preset number of times in this embodiment is three times, and the lowest communication level is low, then the priority level of the first target tag will be lowered. Whether it is adjusted to the second target tag or lower than the second target tag can be determined according to actual conditions, so as to give priority to allocating resources to high-willing users, improve the overall conversion efficiency, and ensure the accuracy and communication rate of voice calls.

[0086] S4. Determine the target promotion time of the target user according to the call time period in the target tag corresponding to the target user, and call the target user during the target promotion time to promote the target product.

[0087] The target tag corresponding to the target user includes the first target tag and the second target tag in step S3, which depends on whether the target user in step S3 is obtained based on only the first target tag or based on a combination of the first target tag and the second target tag.

[0088] At this time, the call time period is saved in the voice tag, such as the call time period of 08:37-08:39 in the previous example. Then, a time point can be selected within the interval of 08:37-08:39 to make a call to promote the "car aromatherapy".

[0089] This embodiment ensures that the promotion period matches the user's actual status through dynamic scene recognition and voice tag construction, takes into account the coverage and accuracy of voice calls, and reduces invalid calls, thereby significantly improving the accuracy and communication rate of intelligent calls.

[0090] Example 2

[0091] Please refer to Figure 2 , an AI intelligent calling method, based on the above embodiment 1, further includes the steps of:

[0092] S5. When there is a new user, obtain the user profile of the new user, match the voice tag corresponding to the user profile of the new user from all voice tags, the category of the current product is the same as the category of the target product, and the purchase intention reaches the purchase recommendation level as the third target tag, filter out the call scene with the highest frequency from all third target tags as the alternative scene, obtain the call time period of the current user with the same user profile as the new user in the alternative scene, use the time with the highest frequency among all the obtained call time periods as the new promotion time for the new user, and call the new user at the new promotion time to promote the target product.

[0093] The user profile of the new user may not have any tags, in which case the call can be made directly. Alternatively, the user profile of the new user may include some tags, in which case the following steps are included in step S5 corresponding to the user profile of the new user:

[0094] If the tag set in the user portrait of the new user is included in the tag set of the user portrait in the voice tag, the user portrait in the voice tag corresponds to the user portrait of the new user.

[0095] That is, if the tags of the new user's user profile are included in the user profile tag set in the voice tag, it is considered a corresponding relationship. For example, if the new user's user profile tags only include age 20-30, gender female, and occupation office worker, then the user profile tag set in the voice tag includes 20-30, female, and office worker, and it is considered a corresponding relationship. In this case, if the target product is "car fragrance", the third target tag is "the voice tag corresponding to the user profile of age 20-30, gender female, and occupation office worker, the current product category is "automotive supplies", and the purchase intention reaches normal."

[0096] In this embodiment, the steps of using the time with the highest frequency among all acquired call time periods as the new promotion time for new users include:

[0097] Divide the entire day into multiple time windows, calculate the frequency of occurrence of all acquired call time periods in each time window, and use the time window with the highest frequency as the new promotion time for new users.

[0098] Among them, the time windows of this embodiment are 48, and half an hour is a time window. At this time, if the call time period overlaps in two time windows at the same time, such as 09:27-09:32, the occurrence frequency of the two time windows is increased by 1. Finally, the time with the highest occurrence frequency among all call time periods is counted, such as 08:30-09:00, and the new user can be promoted in this time period.

[0099] In this embodiment, taking into full consideration the association between products and scenarios in the original embodiment, step S5 filters out the most frequently appearing call scenarios from all third target tags as candidate scenarios, obtains the call time periods of the current user with the same user profile as the new user in the candidate scenarios, and uses the most frequently appearing time period among all the obtained call time periods as the new promotion time for the new user, including the following steps:

[0100] Filter out the call scenarios with the highest combined scores of frequency and scenario causal score from all third target tags as candidate scenarios, and obtain the call time periods of the current user with the same user profile as the new user in the candidate scenarios;

[0101] The whole day is divided into multiple time windows, and the frequency of occurrence of all acquired call time periods in each time window is calculated. The time window with the highest combined score of the frequency of occurrence and the time window causal score is used as the new promotion time for new users.

[0102] The time window causal score and the scenario causal score are similar. The average score of the purchase intention of the first product category in the first time window and the average score of the communication intention in other time windows except the first time window are calculated, and the difference between the two average scores is used as the time window causal score.

[0103] In this embodiment, the combined score of occurrence frequency and scenario causal score is the final score obtained by multiplying the scenario causal score by the occurrence frequency. If the purchase intention score is high, such as 100, the scenario causal score is high, and the square root of the occurrence frequency and scenario causal score can be multiplied together. In other scenarios where causal effects are more emphasized, the scenario causal score can be raised to a power and then multiplied together. The same applies to the occurrence frequency and time window causal score.

[0104] At this time, if the alternative scenario is the "driving scenario", the call time period of the current user with the same user profile as the new user in the "driving scenario" is obtained, such as the time windows 8:30-9:00 and 18:00-18:30. At this time, the former has an occurrence frequency of 236 times and a time window causal score of 0.56, and the latter has an occurrence frequency of 178 times and a time window causal score of 0.82. The scores of the two are 132.16 and 145.96 respectively. Therefore, 18:00-18:30 is selected as the new promotion time for the new user, thereby further improving the accuracy and communication rate of cold start recommendations.

[0105] This embodiment addresses the cold start challenge for new users by matching high-frequency scenarios, profile similarity, and scenario-time distribution. This overcomes the limitations of traditional new user cold start methods, which rely on random calls or broad tags. It also reduces bias caused by data sparsity and provides new users with personalized marketing strategies that are strongly linked to the scenario, significantly improving new user engagement rates.

[0106] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0107] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions.

[0108] It should be noted that, in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims enumerating several means, several of these means may be embodied by one and the same hardware. The use of the words first, second, third etc. is for convenience only and does not indicate any order. These words may be understood as part of the component name.

[0109] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0110] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments after learning the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0111] Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if such modifications and variations fall within the scope of the claims and their equivalents, the present invention shall also include such modifications and variations.

Claims

1. An AI intelligent calling method, characterized in that: Including steps: Acquire the voice content when promoting the current product to the current user, and obtain the call time period, purchase intention, and call scenario of the current user based on the voice content; Obtaining a user profile of the current user, and associating the current product, the call time period, the purchase intention, the call scenario, and the user profile with the current user to obtain a voice tag; When it is necessary to promote a target product, a voice tag whose category of the current product is the same as that of the target product and whose purchase intention reaches the purchase recommendation level is matched from all voice tags as a first target tag. When it is determined whether the number of current users in the first target tag reaches the target number, if so, the current user in the first target tag is used as the target user. Otherwise, after removing the voice tags that are the same as the current user in the first target tag from all voice tags, a second target tag is matched based on the user portrait and the call scenario from the remaining behavior tags. The current user in the second target tag and the current user in the first target tag are used together as target users. Determining a target promotion time for the target user according to the call time period in the target tag corresponding to the target user, and calling the target user during the target promotion time to promote the target product; Also includes the steps: If the target product has an associated scenario, the voice tag containing the associated scenario and whose purchase intention reaches the purchase recommendation level will also be used as the first target tag, specifically including: Add the score for calculating purchase intention to the voice tag; Among all the voice tags, the average score of the purchase intention of the first product category in the first call scenario and the average score of the purchase intention in other scenarios except the first call scenario are counted, and the difference between the two average scores is used as the scenario causal score. If the scenario causal score is greater than the first preset causal score, an association relationship is established between the first product category and the first call scenario.

2. The AI ​​intelligent calling method according to claim 1, characterized in that: The step of matching the second target tag based on the user portrait and the call scenario includes the following steps: The call scenes that appear at a preset frequency ratio in all first target tags are selected as candidate scenes, and the voice tags that match the user portrait and the target product in the remaining behavior tags, whose call scenes belong to the candidate scenes and whose purchase intention reaches the purchase recommendation level are selected as second target tags.

3. The AI ​​intelligent calling method according to claim 1, characterized in that: Also includes the steps: When there is a new user, obtain the user portrait of the new user, match the voice tag corresponding to the user portrait of the new user from all voice tags, the category of the current product is the same as the category of the target product, and the purchase intention reaches the purchase recommendation level as the third target tag, filter out the call scene with the highest frequency from all third target tags as the alternative scene, obtain the call time period of the current user with the same user portrait as the new user in the alternative scene, use the time with the highest frequency in all the obtained call time periods as the new promotion time for the new user, and call the new user at the new promotion time to promote the target product.

4. The AI ​​intelligent calling method according to claim 3, characterized in that: The method of using the time with the highest frequency among all acquired call time periods as the new promotion time for the new user comprises the steps of: The whole day is divided into multiple time windows, and the occurrence frequency of all acquired call time periods in each time window is calculated. The time window with the highest occurrence frequency is used as the new promotion time for the new user.

5. The AI ​​intelligent calling method according to claim 3, characterized in that: The steps corresponding to the user profile of the new user include: If the tag set in the user portrait of the new user is included in the tag set of the user portrait in the voice tag, then the user portrait in the voice tag corresponds to the user portrait of the new user.

6. The AI ​​intelligent calling method according to claim 1, characterized in that: Also includes the steps: If, after the call time node of the first target tag, the same voice tag as the current user of the first target tag appears for a preset number of consecutive times and the purchase intention reaches the minimum communication level, the priority level of the first target tag will be adjusted to the second target tag or lower than the second target tag.

7. The AI ​​intelligent calling method according to claim 1, characterized in that: The purchase intention includes the product purchase tendency or the user communication willingness or all of the above.

8. The AI ​​intelligent calling method according to claim 1, characterized in that: The identification of the call scenario includes at least one of the following methods: Determine the call scenario by analyzing keywords in the voice content; Alternatively, the call scene is determined by analyzing the background noise in the voice content; Alternatively, the call scenario may be determined by analyzing keywords in the voice content and background noise.

9. An AI intelligent calling method according to any one of claims 1 to 8, characterized in that: Also includes the steps: If the current users in the first target tags overlap, all first target tags of the same current user are summarized into one first target tag; If the current users in the second target tags overlap, all second target tags of the same current user are summarized into one second target tag.

Images