Information collection method and device based on machine learning and readable storage medium

Abstract

The invention provides an information collection method based on machine learning, and the method comprises the steps: obtaining promotion collection information which comprises main body information and contact information, carrying out the contact based on the contact information through employing a robot customer service, and converting a call record into a text through employing an automatic voice recognition technology in a contact process; performing intention judgment and slot collection on the text by adopting a natural language processing technology; and according to the intention judgment and the slot collection result, generating a response audio for replying. According to the scheme, the popularization collection information is obtained through an automatic means, the information collection speed is greatly increased, and a large amount of time is saved for follow-up business follow-up. The robot customer service trained based on the machine learning technology in advance is adopted for contact, the communication efficiency is greatly improved, the natural language processing technology is used for intention judgment and slot position collection, and the accuracy and integrity of information collection are improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to an information collection method, device, and readable storage medium based on machine learning. Background Art

[0002] In the marketing and business development process of the interior design industry, efficiently and accurately collecting renovation-related information from potential customers is key to improving business conversion rates and optimizing service quality. Traditional information collection methods have many limitations and are unable to meet the needs of the modern market.

[0003] In the early days, renovation companies primarily relied on offline events and paper questionnaires to collect customer information. This approach was not only inefficient and consuming significant manpower, material resources, and time, but the information collected was often incomplete and inaccurate. For example, paper questionnaires distributed at renovation exhibitions could be filled out carelessly or with incomplete information, hindering subsequent business follow-up.

[0004] With the development of internet technology, online marketing has become increasingly mainstream, and interior design companies have begun collecting customer information through channels such as website forms and social media. However, while this method has improved information collection efficiency, it still suffers from uneven information quality. Many customers may only provide partial key information when filling out online forms, or even include false information, making it difficult for interior design companies to accurately analyze customer needs and provide service recommendations. Summary of the Invention

[0005] The present invention aims to provide a machine learning-based information collection method to address the problems of low efficiency and potential information loss or errors in existing information collection. The machine learning-based information collection method provided in this application includes:

[0006] Acquire promotional collection information, wherein the promotional collection information includes subject information and contact information, wherein the subject information is house area information;

[0007] Using a robot customer service to make contact based on the contact information, the robot customer service is pre-trained based on machine learning technology;

[0008] Automatic speech recognition technology is used to convert call recordings into text during the contact process;

[0009] Using natural language processing technology to determine the intent of the text and collect slots;

[0010] Based on the intention judgment and the results of the slot collection, a response audio is generated for reply.

[0011] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, before the robot customer service is used to contact the customer based on the contact information, the method further includes:

[0012] Assign different types of robots based on city information, lead time information and number of calls already made.

[0013] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, the allocation of different types of robots based on city information, clue time information, and the number of dial times already made includes:

[0014] The city information, clue time information and number of dialed times information have different weights respectively;

[0015] The total weight value corresponding to the subject information is calculated. If it exceeds a preset value, the robot type is determined to be the first robot. If it does not exceed the preset value, the robot type is determined to be the second robot, and the performance of the first robot is better than that of the second robot.

[0016] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, the slot includes any one or more of the house type, city, area, whether the house is delivered and the amount of work information.

[0017] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, generating a response audio for reply based on the intention judgment and the result of the slot collection includes:

[0018] Based on the intention judgment and the results of the slot collection, select and generate a response audio in the preset canvas for reply.

[0019] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, before using natural language processing technology to perform intent judgment and slot collection on the text, the method further includes:

[0020] The text is corrected using a text error correction model based on deep learning.

[0021] Based on the information collection method provided in the first aspect of the embodiment of the present application, optionally, the method further includes:

[0022] The decoration information is associated with the promotion collection information and stored.

[0023] A second aspect of the embodiments of the present application provides an information collection device based on machine learning, including:

[0024] An acquisition unit, configured to acquire promotional collection information, wherein the promotional collection information includes subject information and contact information, and the subject information is house area information;

[0025] A contact unit, configured to use a robot customer service to make contact based on the contact information, wherein the robot customer service is pre-trained based on machine learning technology;

[0026] A conversion unit, used to convert call recordings into text using automatic speech recognition technology during the contact process;

[0027] A collection unit, configured to use natural language processing technology to determine the intent of the text and collect slots;

[0028] The reply unit is used to generate a response audio for reply based on the intention judgment and the results of the slot collection.

[0029] Based on the machine learning-based information collection device provided in the second aspect of the embodiment of the present application, optionally, the contact unit is further configured to:

[0030] Assign different types of robots based on city information, lead time information and number of calls already made.

[0031] Based on the machine learning-based information acquisition device provided in the second aspect of the embodiment of the present application, optionally, the contact unit is specifically configured to:

[0032] The city information, clue time information and number of dialed times information have different weights respectively;

[0033] The total weight value corresponding to the subject information is calculated. If it exceeds a preset value, the robot type is determined to be the first robot. If it does not exceed the preset value, the robot type is determined to be the second robot, and the performance of the first robot is better than that of the second robot.

[0034] Based on the machine learning-based information collection device provided in the second aspect of the embodiment of the present application, optionally, the slot includes any one or more of the house type, city, area, whether the house is delivered, and project quantity information.

[0035] Based on the machine learning-based information collection device provided in the second aspect of the embodiment of the present application, optionally, the reply unit is specifically configured to:

[0036] Based on the intention judgment and the results of the slot collection, select and generate a response audio in the preset canvas for reply.

[0037] Based on the machine learning-based information collection device provided in the second aspect of the embodiment of the present application, optionally, the collection unit is further configured to:

[0038] The text is corrected using a text error correction model based on deep learning.

[0039] Based on the machine learning-based information collection device provided in the second aspect of the embodiment of the present application, optionally, the collection unit is also used to associate and store the decoration information with the promotion collection information.

[0040] A third aspect of the embodiments of the present application provides an information collection device based on machine learning, including:

[0041] CPU, memory, input and output interfaces, wired or wireless network interfaces, and power supply;

[0042] The memory is a transient storage memory or a persistent storage memory;

[0043] The central processing unit is configured to communicate with the memory and execute instruction operations in the memory on the device to perform the method as described in any one of the first aspects of the embodiments of the present application.

[0044] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a computer, enable the computer to execute the method described in any one of the first aspects of the embodiments of the present application.

[0045] A fifth aspect of the embodiments of the present application provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute any one of the methods described in the first aspect of the embodiments of the present application.

[0046] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: This application provides a machine learning-based information collection method, comprising: obtaining promotional information, including subject information and contact information, wherein the subject information is house area information; using a robot customer service representative pre-trained using machine learning technology to make contact based on the contact information; converting call recordings into text using automatic speech recognition technology; using natural language processing technology to determine intent and collect slots from the text; and generating a response audio based on the results of the determination and collection of slots. This solution obtains promotional information through automated means, avoiding the tedious process of manual information collection, significantly improving the speed of information collection and saving a significant amount of time for subsequent business follow-up. Using a robot customer service representative pre-trained using machine learning technology to make contact enables 24-hour uninterrupted service. The robot can quickly respond to customers and handle multiple customer inquiries simultaneously, significantly improving communication efficiency compared to human customer service representatives. By using natural language processing technology for intent determination and slot collection, it can deeply understand the true intentions and key information in customer speech, improving the accuracy and completeness of information collection. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] To more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. A person of ordinary skill in the art can also derive other drawings based on the provided drawings without inventive effort. It should be understood that the drawings provided in this section are only used to better understand the present solution and do not constitute a limitation of the present application.

[0048] Figure 1 A flowchart of an embodiment of the information collection method based on machine learning provided in this application.

[0049] Figure 2 This is another flowchart of an embodiment of the information collection method based on machine learning provided in this application.

[0050] Figure 3 A schematic diagram of the promotion information collection process provided by this application.

[0051] Figure 4 A schematic diagram of the slot configuration provided for this application.

[0052] Figure 5 A schematic diagram of the dialogue flow text provided for this application.

[0053] Figure 6A structural diagram of an embodiment of the information acquisition device based on machine learning provided in this application.

[0054] Figure 7 This is a structural diagram of an embodiment of the information acquisition device based on machine learning provided in this application. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application. At the same time, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0056] The terms "first," "second," "third," "fourth," and the like (if any) in the specification and claims of this application and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions, e.g., a process, method, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or apparatus.

[0057] In the marketing and business development process of the interior design industry, efficiently and accurately collecting renovation-related information from potential customers is key to improving business conversion rates and optimizing service quality. Traditional information collection methods have many limitations and are unable to meet the needs of the modern market.

[0058] In the early days, renovation companies primarily relied on offline events and paper questionnaires to collect customer information. This approach was not only inefficient and consuming significant manpower, material resources, and time, but the information collected was often incomplete and inaccurate. For example, paper questionnaires distributed at renovation exhibitions could be filled out carelessly or with incomplete information, hindering subsequent business follow-up.

[0059] With the development of internet technology, online marketing has become increasingly mainstream, and interior design companies have begun collecting customer information through channels such as website forms and social media. However, while this method has improved information collection efficiency, it still suffers from uneven information quality. Many customers may only provide partial key information when filling out online forms, or even include false information, making it difficult for interior design companies to accurately analyze customer needs and provide service recommendations.

[0060] To solve the above problems, the present invention provides an information collection method based on machine learning. Figure 1 , an embodiment of the information collection method based on machine learning provided in this application includes: steps 101 to 105.

[0061] 101. Obtain promotional collection information.

[0062] Acquire promotional collection information, the promotional collection information including subject information and contact information, the subject information being house area information;

[0063] Specifically, users fill out information on the promotion page, and the system collects this information into promotional data. This includes both subject information and contact information. Subject information is typically represented by the area of the property, a key indicator for measuring renovation workload, material usage, and costs. Contact information provides a channel for subsequent communication with users, facilitating further understanding of their needs and providing services.

[0064] It is understandable that the promotional information collected mainly comes from the user's filling in the web promotion page, but the acquisition method is not limited to this. In different business scenarios, it can also be collected by scanning codes, embedding forms into other platforms, and connecting to third-party data interfaces. Moreover, in terms of information content, although the house area information is the main information, other key information such as the purpose of the house (self-residence, investment, etc.) and the expected renovation time may also be added. At the same time, the contact information obtained is mainly mobile phone numbers, and the contact information is not limited to manual filling by users. When technology permits and compliance is met, user device IDs, WeChat numbers and other identifiers can be automatically obtained to facilitate the association of subsequent communication data with the user's permission.

[0065] 102. Use robot customer service to contact based on the contact information.

[0066] After obtaining a potential customer's contact information, a pre-trained robot customer service agent based on machine learning technology will contact the customer. Machine learning technology enables robot customer service agents to absorb vast amounts of conversational data and business knowledge, thus developing a certain level of intelligent interaction capabilities. Based on pre-set rules and algorithms, they can communicate naturally and smoothly with customers, understanding their needs and intentions. It's understandable that robot customer service agents aren't limited to training with a single machine learning model; they can also combine multiple machine learning algorithms or deep learning technologies to enhance their performance and intelligence.

[0067] 103. Automatic speech recognition technology is used to convert call recordings into text during the contact process.

[0068] During customer service calls, automated speech recognition (ASR) technology is used to convert call recordings into text for easier processing and analysis. This converts voice information into editable, searchable, and analyzable text data, providing a foundation for further intent assessment and information extraction. ASR technology offers a variety of implementations and models, each with varying accuracy, processing speed, and adaptability to different accents and language environments. In practice, the appropriate speech recognition technology can be selected based on specific business needs and scenarios.

[0069] 104. Use natural language processing technology to determine the intent of the text and collect slots

[0070] The converted text is analyzed in depth using natural language processing (NLP) technology, primarily focusing on intent assessment and slot collection. Intent assessment identifies the customer's true needs and objectives expressed during the conversation, such as whether they want to learn about decorating styles, inquire about renovation prices, express impatience, or indicate they are currently busy. Slot collection extracts key business-related information from the text, such as property type, whether the property has been delivered, and renovation budget, and populates this information into pre-defined slots.

[0071] Natural language processing technology can employ a variety of approaches, including rule-based methods, machine learning methods (such as support vector machines and decision trees), and deep learning methods (such as recurrent neural networks and Transformer models). In practical applications, appropriate natural language processing technologies and models can be selected based on specific business needs and data characteristics. The definition of slots is not fixed and can be adjusted and expanded based on business changes and expansion to meet diverse information collection needs.

[0072] 105. Generate a response audio based on the intention judgment and the results of the slot collection for reply.

[0073] After determining intent and collecting slots, an audio response is generated based on the analysis results and sent back to the customer via the robot customer service. The audio response is personalized based on the customer's needs and the information provided, providing accurate and useful information and solutions. There are various ways to generate audio responses, such as using text-to-speech (TTS) technology to convert text into audio. Alternatively, you can pre-record commonly used voice templates and combine and splice them to suit different situations.

[0074] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: This application provides a machine learning-based information collection method, comprising: obtaining promotional information, the promotional information including subject information and contact information, the subject information being house area information; using a robot customer service representative pre-trained based on machine learning technology to make contact based on the contact information; converting the call recording into text using automatic speech recognition technology during the contact process; using natural language processing technology to perform intent determination and slot collection on the text; and generating a response audio based on the results of the intent determination and slot collection. This solution obtains promotional information through automated means, avoiding the tedious process of manual information collection, greatly improving the speed of information collection, and saving a considerable amount of time for subsequent business follow-up. Using a robot customer service representative pre-trained based on machine learning technology to make contact greatly improves communication efficiency. By using natural language processing technology for intent determination and slot collection, it can deeply understand the true intention and key information in the customer's speech, improving the accuracy and completeness of information collection.

[0075] In order to facilitate the use of this method in actual implementation, this application also provides a more detailed embodiment that can be optionally implemented, please refer to Figure 2 , an embodiment of the information collection method based on machine learning of the present application includes: steps 201 to 105.

[0076] 201. Obtain promotion collection information.

[0077] Acquire promotional collection information, wherein the promotional collection information includes subject information and contact information, wherein the subject information is house area information;

[0078] An example of promoting the acquisition of information can be found in Figure 3 , the user fills in the room area information and contact information, which indicates that there is a desire to renovate. Figure 1Similar to step 101, it is understood that the promotional information can be obtained from multiple platforms, such as mobile applications and mini-programs. Furthermore, the information collected is not limited to the current settings. As the business develops, more dimensional information can be added, such as the user's initial intention for decoration style and special decoration requirements.

[0079] 202. Assign different types of robots based on city information, lead time information and number of calls

[0080] Specifically, the city information, lead time information, and number of calls made are each assigned different weights. The total weight corresponding to the subject information is calculated. If it exceeds a preset value, the robot type is determined to be the first robot; if it does not exceed the preset value, the robot type is determined to be the second robot, and the performance of the first robot is superior to that of the second robot. When the calculated total weight exceeds the preset value, it indicates that the lead has higher value and potential and should be assigned to the first robot with better performance for follow-up. If the total weight does not exceed the preset value, the second robot is assigned. The first robot has a richer knowledge base, more intelligent conversational strategies, and stronger problem-solving capabilities, which can better handle complex customer needs and improve business conversion rates.

[0081] Specifically, the first robot can use large-scale, diverse data, combined with complex deep learning algorithms (such as those based on the Transformer architecture) and reinforcement learning technology, and trained for a long time on a high-performance cluster, with professional personnel continuously monitoring and adjusting. The knowledge graph is complete and in-depth, covering information from many aspects of the industry, with a sound update and maintenance mechanism, and capable of complex reasoning and decision-making. The second robot can use simple machine learning algorithms (such as rule-based models), which have a short training time and can be completed on ordinary servers. The knowledge graph is simple and has limited coverage, and is mainly used for simple information query and matching.

[0082] In the actual implementation process, multiple types of robots can be set up to meet different needs. For example, for users in a specific city, a customer service robot trained for that city can be used, and a third robot can be used to handle some special situations or clues with specific needs.

[0083] 203. Use robot customer service to contact based on the contact information.

[0084] Communication methods are not limited to voice calls but can also be expanded to video calls. The training methods and speech templates of robot customer service can also be continuously updated and optimized according to actual communication situations.

[0085] 204. Automatic speech recognition technology is used to convert call recordings into text during the contact process.

[0086] Automatic speech recognition (ASR) technology is based on deep learning models, such as those based on the Transformer architecture. These models are trained on large amounts of speech data to learn the mapping between speech features and text. When a customer service robot speaks with a user, the recording is transmitted in real time to the ASR system for recognition and conversion to text.

[0087] 205. Using a text error correction model based on deep learning to perform error correction on the text

[0088] Utilize deep learning-based text correction models, such as those based on recurrent neural networks (RNNs) or Transformers. These models learn from a large number of correct and incorrect text pairs to identify and correct errors in text. When processing the converted speech text, the models detect and correct typos, grammatical errors, and other errors. Text correction models can continuously update training data to adapt to new language habits and error types. Furthermore, they can combine multiple correction techniques, such as rule-based correction, to improve the accuracy and efficiency of correction.

[0089] 206. Use natural language processing technology to determine the intent of the text and collect slots

[0090] Using natural language processing (NLP) technology, including lexical analysis, syntactic analysis, semantic understanding and other links. Through the trained NLP model, the text is analyzed to identify the user's intention, such as asking about the decoration price, understanding the decoration style, etc., and key information is extracted and filled into the corresponding slots, such as house type, budget, etc. The slots include any one or more of the house type, city, area, whether the house is delivered and the amount of work. In the actual implementation process, the slot settings can refer to Figure 4 , can be adjusted according to actual needs in actual conditions and is not limited here.

[0091] Condition: Housing Type - Number of Inquiries = 0: When the number of inquiries about housing type is 0, the system will execute the jump strategy to the "housing type" related process. This means that if the user has not been asked about the housing type, the system will proceed according to this condition and guide the robot to ask the user about the housing type, such as "Okay, I understand. What is your housing type?"

[0092] Condition: Slot - Property Type in (Unfinished, Renovated): When the user's response indicates the property type is "Unfinished" or "Renovated," the system will execute the "Next Slot" jump strategy. For example, if the owner's response indicates the property type is an old house, which fits the "Renovated" category, the system will jump to the next slot based on this condition and proceed with the subsequent process.

[0093] Response Condition: Current Query Slot = Property Type, Note: Reply to this Slot: When the system is currently querying the property type and the user has replied to that slot, this logic is used for processing. For example, if the owner indicates that the property type is an old house, this condition is met, and the system can perform subsequent actions based on this, such as updating the slot information and advancing the conversation process.

[0094] Condition 1: Slot - House Type: Not Empty, Condition 2: Slot - House Type in (Finely Decorated House, Simply Decorated House), Condition 4: Slot - House Type not in (Unfinished House, Renovated House, Finely Decorated House, Simply Decorated House), Remarks: Unanswered for this slot: When the slot house type is not empty and is a finely decorated house or a simply decorated house, and when the slot house type is not unfinished house, renovated house, finely decorated house, or simply decorated house, the system will regard it as unanswered for this slot (possibly because the reply type does not meet the expected common type) and may take corresponding measures, such as re-asking, guiding the user to confirm, etc.

[0095] 207. Select and generate a response audio in a preset canvas based on the intention judgment and the results of the slot collection to reply.

[0096] Specifically, a communication process can refer to Figure 5 Based on intent judgment and slot collection, the system matches the corresponding reply content in a preset speech template library (i.e., the preset canvas). It then uses text-to-speech (TTS) technology to convert the selected text into an audio format for the reply. TTS technology uses deep learning models to generate more natural and fluent speech.

[0097] 208. Associate the decoration information with the promotion collection information and store it

[0098] Using database management technology, we associate and store renovation information (such as intent determination results and slot collection information) with promotional information (such as user basic information and entry time). We can use a relational database (such as MySQL) or a non-relational database (such as MongoDB) to establish data tables and associated fields to ensure data consistency and queryability.

[0099] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: This application provides a machine learning-based information collection method, comprising: obtaining promotional information, the promotional information including subject information and contact information, the subject information being house area information; using a robot customer service representative pre-trained based on machine learning technology to make contact based on the contact information; converting the call recording into text using automatic speech recognition technology during the contact process; using natural language processing technology to perform intent determination and slot collection on the text; and generating a response audio based on the results of the intent determination and slot collection. This solution obtains promotional information through automated means, avoiding the tedious process of manual information collection, greatly improving the speed of information collection, and saving a considerable amount of time for subsequent business follow-up. Using a robot customer service representative pre-trained based on machine learning technology to make contact greatly improves communication efficiency. By using natural language processing technology for intent determination and slot collection, it can deeply understand the true intention and key information in the customer's speech, improving the accuracy and completeness of information collection.

[0100] The above content describes the information collection method based on machine learning provided by this application. To support the implementation of the above embodiments, this application also provides an information collection device based on machine learning. Figure 6 , an embodiment of the information acquisition device based on machine learning provided by this application includes:

[0101] An acquisition unit 601 is configured to acquire promotional collection information, wherein the promotional collection information includes subject information and contact information, wherein the subject information is house area information;

[0102] A contact unit 602 is configured to use a robot customer service to make contact based on the contact information, where the robot customer service is pre-trained based on machine learning technology;

[0103] A conversion unit 603 is used to convert the call recording into text using automatic speech recognition technology during the call process;

[0104] A collection unit 604 is configured to use natural language processing technology to determine the intent of the text and collect slots;

[0105] The reply unit 605 is used to generate a response audio according to the intention judgment and the results of the slot collection.

[0106] Optionally, the contact unit is further configured to:

[0107] Assign different types of robots based on city information, lead time information and number of calls already made.

[0108] Optionally, the contact unit is specifically configured to:

[0109] The city information, clue time information and number of dialed times information have different weights respectively;

[0110] The total weight value corresponding to the subject information is calculated. If it exceeds a preset value, the robot type is determined to be the first robot. If it does not exceed the preset value, the robot type is determined to be the second robot, and the performance of the first robot is better than that of the second robot.

[0111] Optionally, the slot includes any one or more of house type, city, area, whether the house has been delivered, and construction quantity information.

[0112] Optionally, the reply unit is specifically configured to:

[0113] Based on the intention judgment and the results of the slot collection, select and generate a response audio in the preset canvas for reply.

[0114] Optionally, the collecting unit is further used for:

[0115] The text is corrected using a text error correction model based on deep learning.

[0116] Optionally, the collection unit is further configured to associate and store the decoration information with the promotion collection information.

[0117] In this embodiment, the processes performed by each unit in the device are the same as those described above. Figure 1 The method processes described in the corresponding embodiments are similar and will not be repeated here.

[0118] Figure 7 It is a structural diagram of an information collection device provided in an embodiment of the present application. The information collection device 700 may include one or more central processing units (CPU) 701 and a memory 705. The memory 705 stores one or more applications or data.

[0119] In this embodiment, the specific functional module division in the central processing unit 701 can be the same as the above Figure 6 The functional module division method of each unit described in is similar and will not be repeated here.

[0120] Memory 705 can be volatile or persistent storage. The program stored in memory 705 can include one or more modules, each of which can include a series of instruction operations on the server. Furthermore, the central processing unit 701 can be configured to communicate with memory 705 and execute the series of instruction operations in memory 705 on the information collection device 700.

[0121] The information collection device 700 may also include one or more power supplies 702 , one or more wired or wireless network interfaces 703 , one or more input and output interfaces 704 , and the like.

[0122] The CPU 701 can execute the aforementioned Figure 1 The operations performed by the information collection method in the illustrated embodiment will not be described in detail here.

[0123] An embodiment of the present application also provides a computer storage medium for storing computer software instructions used for the above-mentioned information collection method, which includes a program designed for executing the information collection method.

[0124] The information collection method can be as described above Figure 1 The information collection method described in .

[0125] The present application also provides a computer program product, which includes computer software instructions that can be loaded by a processor to implement the above Figure 1 The process of any information collection method.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the equivalent transformation of circuits and the division of units are only a kind of logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0127] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0128] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for collecting information based on machine learning, characterized in that: include: Acquire promotional collection information, the promotional collection information including subject information and contact information, the subject information being house area information; Using a robot customer service to make contact based on the contact information, the robot customer service is pre-trained based on machine learning technology; Automatic speech recognition technology is used to convert call recordings into text during the contact process; Using natural language processing technology to determine the intent of the text and collect slots; Based on the intention judgment and the results of the slot collection, a response audio is generated for reply.

2. The information collection method according to claim 1, characterized in that: Before the robot customer service is used to contact based on the contact information, the method further includes: Assign different types of robots based on city information, lead time information and number of calls already made.

3. The information collection method according to claim 2, characterized in that: The allocation of different types of robots based on city information, clue time information and the number of dialed times includes: The city information, clue time information and number of dialed times information have different weights respectively; The total weight value corresponding to the subject information is calculated. If it exceeds a preset value, the robot type is determined to be the first robot. If it does not exceed the preset value, the robot type is determined to be the second robot, and the performance of the first robot is better than that of the second robot.

4. The information collection method according to claim 1, characterized in that: The slot includes any one or more of house type, city, area, whether the house is delivered and construction quantity information.

5. The information collection method according to claim 1, characterized in that: Generating a response audio for reply based on the intention judgment and the result of the slot collection includes: Based on the intention judgment and the results of the slot collection, select and generate a response audio in the preset canvas for reply.

6. The information collection method according to claim 1, characterized in that: Before using natural language processing technology to determine the intent of the text and collect slots, the method further includes: The text is corrected using a text error correction model based on deep learning.

7. The information collection method based on machine learning according to claim 1, characterized in that: The method further comprises: The slot collection result is associated with the promotion collection information and stored.

8. An information collection device based on machine learning, characterized in that: include: An acquisition unit, configured to acquire promotional collection information, wherein the promotional collection information includes subject information and contact information, and the subject information is house area information; A contact unit, configured to use a robot customer service to make contact based on the contact information, wherein the robot customer service is pre-trained based on machine learning technology; A conversion unit, used to convert call recordings into text using automatic speech recognition technology during the contact process; A collection unit, configured to use natural language processing technology to determine the intent of the text and collect slots; The reply unit is used to generate a response audio for reply based on the intention judgment and the results of the slot collection.

9. An information collection device based on machine learning, characterized in that: include: CPU, memory, input and output interfaces, wired or wireless network interfaces, and power supply; The memory is a transient storage memory or a persistent storage memory; The central processing unit is configured to communicate with the memory and execute instructions in the memory on the device to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The method comprises instructions, which, when executed on a computer, cause the computer to perform the method according to any one of claims 1 to 7.

Images