An internet-based AI outbound call data extraction method and system

Abstract

The scheme discloses a data extraction method based on internet AI outbound call, comprising: obtaining required conversation data from all conversation information and data generated in the internet AI outbound call process; comparing and screening the obtained conversation data with data in a database to determine the matching degree of the conversation data in the internet AI outbound call system running process and the data in the database; extracting all data with the required matching degree from the database to form a final returned data set, and returning the data set to the conversation system of the internet AI outbound call system as reference data for the next conversation for the internet AI outbound call system. The application improves the data processing efficiency and data extraction accuracy of the internet AI outbound call system through reasonable and effective data processing and data extraction method.

Description

Technical Field

[0001] This invention belongs to the field of computer data processing and artificial intelligence technology, and specifically relates to a data extraction method and system based on Internet AI outbound calling. Background Technology

[0002] With my country's rapid economic development, the costs of labor, facilities, and equipment are constantly rising, leading to increasing competition in traditional outbound calling services. A large amount of repetitive and mechanical work consumes the majority of customer service staff's time, leaving little time for complex cases. Internet-based AI (Artificial Intelligence) outbound calling systems have emerged to address this need. They can assist in completing repetitive tasks, improve outbound calling efficiency, and thus save on significant human and material costs.

[0003] Internet-based AI outbound calling systems, as a platform-based AI technology, serve as a crucial AI entry point. They enable access to underlying AI computing power and can be integrated with various application scenarios. They have been widely adopted in the personal user sector, as well as in industries such as finance, healthcare, education, and smart home, handling functions such as online customer service, electronic medical record transcription, and personal assistants, thus promoting human-computer interaction and significantly improving work efficiency. However, despite their widespread application, internet-based AI outbound calling systems still face challenges, including low data extraction efficiency, high error rates, and the inability to obtain highly accurate data during actual outbound calls.

[0004] Therefore, improving the data processing efficiency and data extraction accuracy of the Internet AI outbound calling system through reasonable and effective technical means is a problem that needs to be solved. Summary of the Invention

[0005] In view of the above problems, this application provides a data extraction method, system and device based on Internet AI outbound calling to solve the above technical problems.

[0006] This application provides the following technical solutions:

[0007] Firstly, this application provides a data extraction method based on Internet AI outbound calling, the method comprising:

[0008] Extract the required dialogue data from all dialogue information and data generated during the Internet AI outbound calling process;

[0009] The acquired dialogue data is compared and filtered with data in the database to determine the matching degree between the dialogue data during the operation of the Internet AI outbound calling system and the data in the database.

[0010] All data that meet the matching requirements are extracted from the database and compiled into a final return dataset. This dataset is then returned to the dialogue system of the Internet AI outbound calling system as reference data for the next step of the dialogue.

[0011] All dialogue information and data generated during the operation of the Internet AI outbound calling system refers to all dialogue information and data generated when the AI ​​interacts with the user in the Internet AI outbound calling system.

[0012] The matching degree is used to characterize the degree of matching between the data obtained after comparison and filtering in the database and the dialogue data, i.e., the degree of relevance; the database is a graph database, and includes at least the local data of the Internet AI outbound calling system and searchable Internet data.

[0013] The data extraction process in this application involves filtering the data in the database before extracting all data that meets the matching requirements. Then, based on the filtering, the matching degree of the data to be extracted is determined. Finally, the data that meets the matching degree requirements is extracted. This method of filtering and extracting data layer by layer can improve the accuracy and efficiency of data extraction and reduce the load on data stream processing.

[0014] Furthermore, obtaining the required dialogue data from all dialogue information and data generated during the Internet AI outbound calling process includes:

[0015] From all the dialogue information and data generated by the Internet AI outbound calling system, select the first dialogue set, and then extract the first dataset from the first dialogue set;

[0016] The first dialogue set consists of all data including at least the most recent dialogue between the user and the Internet AI outbound calling system;

[0017] The first dataset consists of core data that fully reflects the semantics after filtering out data in the first dialogue set that is irrelevant to the core semantics. The data types include nodes and relationships.

[0018] Furthermore, the process of comparing and filtering the acquired dialogue data with data in the database to determine the matching degree between the dialogue data during the operation of the Internet AI outbound calling system and the data in the database includes:

[0019] The data in the first dataset is compared with the data in the database, and the first data filtering is performed. All the filtered data groups are combined to form the first returned dataset.

[0020] The first data filtering involves comparing the data in the first dataset with the data in the database. If the similarity between the node attributes in the database and the node attributes in the first dataset is greater than or equal to a first similarity threshold, then all data groups related to that node are extracted as the filtering result of the first data filtering process. The first returned dataset includes all data groups filtered out by the first data filtering process.

[0021] Determine the matching degree between each data group in the first returned dataset and the data in the first dataset. That is, each data group corresponds to a first matching degree relative to the first dataset, and sort the matching degrees.

[0022] The first matching degree is used to represent the degree of matching between each data group in the first returned dataset and the data in the first dataset, that is, the degree of correlation. The higher the first matching degree, the higher the correlation.

[0023] Furthermore, determining the matching degree between each data group in the first returned dataset and the data in the first dataset, i.e., determining the first matching degree, is to obtain the first matching degree corresponding to each data group by comprehensively calculating each data parameter in each data group in the first returned dataset, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes.

[0024] Furthermore, the step of extracting all data from the database that meets the matching requirements to form the final returned dataset includes:

[0025] Extract all data groups in the first returned dataset whose first matching degree is greater than or equal to the first extraction threshold, and combine all the extracted data groups to form the final returned dataset.

[0026] The first extraction threshold is the threshold that satisfies the first matching degree requirement of the final returned dataset;

[0027] Furthermore, the step of extracting all data groups that meet the first matching degree requirement from the first returned dataset and assembling all the extracted data groups into the final returned dataset also includes:

[0028] If at least one data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then all data groups in the first returned dataset with a first matching degree greater than or equal to the first extraction threshold will be extracted and combined to form the final returned dataset.

[0029] If no data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then further determination is made based on whether the first matching degree of the data group in the first returned dataset is greater than or equal to the second extraction threshold, as follows:

[0030] a) If the first matching degree of at least one data group in the first returned dataset is greater than or equal to the second extraction threshold, then all data groups in the first returned dataset with the first matching degree greater than or equal to the second extraction threshold are extracted and combined to form the final returned dataset;

[0031] b) If no data group in the first returned dataset has a first matching degree greater than or equal to the second extraction threshold, then extract the second dataset from the second dialogue set, compare each data group in the first returned dataset with the data in the second dataset, perform second data filtering, and form the second returned dataset from the filtered data groups; then determine the second matching degree of each data group in the second returned dataset.

[0032] Finally, extract all data groups in the second returned dataset whose second matching degree is greater than or equal to the first extraction threshold, and combine them to form the final returned dataset;

[0033] The second matching degree is used to represent the degree of matching between each data group in the second returned dataset and the data in the first and second datasets, i.e., the degree of correlation. The higher the second matching degree, the higher the correlation. The second matching degree is obtained by multiplying the first matching degree and the second matching degree correction coefficient. The second matching degree correction coefficient is obtained by comprehensively calculating the parameters of each data group in the second returned dataset, which means that the factors affecting the first matching degree of the data in the second dataset are added and the final matching degree result is corrected.

[0034] The second data filtering involves comparing each data group in the first returned dataset with the data in the second dataset. If the similarity between any node attribute of each data group in the first returned dataset and the node attribute in the first dataset is greater than or equal to a second similarity threshold, then that data group is extracted as the filtering result of the second data filtering process. The second returned dataset includes all data groups filtered out by the second data filtering process.

[0035] The second dialogue set is a set of dialogues whose dialogue time is before the first dialogue set and whose information and data do not overlap with those of the first dialogue set; the second dataset is the core data that can fully reflect the semantics after filtering out data in the second dialogue set that is irrelevant to the core semantics, and includes data types of nodes and relationships.

[0036] Both the second extraction threshold and the first extraction threshold are preset thresholds that meet the matching degree requirements of the final returned dataset, and the second extraction threshold is less than the first extraction threshold;

[0037] Furthermore, the second matching degree is calculated using the second matching degree relationship model;

[0038] The second matching degree relationship model includes:

[0039]

[0040] in, This represents the second degree of matching for the data set;

[0041] This is the second matching degree correction coefficient for the data set;

[0042] The first match degree of the data set;

[0043] The data groups mentioned in the above parameters are all the same data groups in the second returned dataset.

[0044] In summary, this application improves the data processing efficiency and data extraction accuracy of the Internet AI outbound calling system through reasonable and effective data processing and extraction methods.

[0045] Secondly, this application provides a data extraction system based on Internet AI outbound calling, the system comprising:

[0046] The acquisition module is used to obtain the required dialogue data from all the dialogue information and data generated during the Internet AI outbound calling process;

[0047] The processing module is used to compare and filter the dialogue information and data obtained during the Internet AI outbound call process with the data in the database to determine the matching degree between the dialogue information and data during the operation of the Internet AI outbound call system and the data in the database.

[0048] The output module is used to extract all data that meets the matching requirements from the database, form the final return dataset, and return it to the dialogue system of the Internet AI outbound calling system as reference data for the next dialogue.

[0049] All dialogue information and data generated during the operation of the Internet AI outbound calling system refers to all dialogue information and data generated when the AI ​​interacts with the user in the Internet AI outbound calling system.

[0050] The matching degree is used to characterize the degree of matching between the data obtained after comparison and filtering in the database and the dialogue data, i.e., the degree of relevance; the database is a graph database, and includes at least the local data of the Internet AI outbound calling system and searchable Internet data.

[0051] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0052] Fourthly, this application provides a computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to implement the method as described in the first aspect.

[0053] Compared with existing technologies, this application improves the data processing efficiency and data extraction accuracy of the Internet AI outbound calling system through reasonable and effective data processing and data extraction methods. Attached Figure Description

[0054] For ease of explanation, this application is described in detail below with reference to specific embodiments and accompanying drawings.

[0055] Figure 1 This is a schematic diagram of the method flow of this application;

[0056] Figure 2 This is a schematic diagram of another method flow of this application;

[0057] Figure 3 This is a schematic diagram of the system structure of this application;

[0058] Figure 4 This is a schematic diagram of a computer-readable storage medium according to this application;

[0059] Figure 5 This is a schematic diagram of the computer device described in this application. Detailed Implementation

[0060] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the figures. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0061] This application provides a data extraction method and system that improves the data processing efficiency and data extraction accuracy of Internet AI outbound calling systems in actual operation. It solves the problems of low data extraction efficiency, high error rate, and inability to obtain highly accurate data when talking to users during actual outbound calling processes in Internet AI outbound calling systems.

[0062] The overall concept of the technical solution in this application is as follows:

[0063] In the process of internet AI outbound calling, the acquired dialogue data is compared and filtered in real time with data in the database. Then, the matching degree between the dialogue data and the database data is further determined from the filtered data. Finally, all data with the required matching degree are extracted from the database to form the final returned dataset, which serves as reference data for the next dialogue step in the internet AI outbound calling system. This improves the data processing efficiency and data extraction accuracy of the internet AI outbound calling system, solving problems such as low data extraction efficiency, high error rate, and inability to obtain highly accurate data during actual outbound calls.

[0064] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0065] Example 1

[0066] like Figure 1-2 As shown, this application provides a data extraction method, system, and apparatus based on Internet AI outbound calling to solve the above-mentioned technical problems.

[0067] This application provides the following technical solutions:

[0068] Firstly, this application provides a data extraction method based on Internet AI outbound calling, the method comprising:

[0069] Extract the required dialogue data from all dialogue information and data generated during the Internet AI outbound calling process;

[0070] The acquired dialogue data is compared and filtered with data in the database to determine the matching degree between the dialogue data during the operation of the Internet AI outbound calling system and the data in the database.

[0071] All data that meet the matching requirements are extracted from the database and compiled into a final return dataset. This dataset is then returned to the dialogue system of the Internet AI outbound calling system as reference data for the next step of the dialogue.

[0072] All dialogue information and data generated during the operation of the Internet AI outbound calling system refers to all dialogue information and data generated when the AI ​​interacts with the user in the Internet AI outbound calling system.

[0073] The matching degree is used to characterize the degree of matching between the data obtained after comparison and filtering in the database and the dialogue data, i.e., the degree of relevance; the database is a graph database, and includes at least the local data of the Internet AI outbound calling system and searchable Internet data.

[0074] Specifically, the local data refers to the matched and optimized data stored internally within the Internet AI outbound calling system; the final returned dataset is returned to the dialogue system of the Internet AI outbound calling system, which involves converting the final returned dataset into different forms of data such as text, voice, and images, and outputting it to the dialogue platform of the Internet AI outbound calling system for interaction with customers.

[0075] Relational databases flourished in the 1980s. Many commercial companies (namely Oracle, Ingres, and IBM) supported relational models (table-based organization) for data management. At that time, the primary data management need was report generation. Until recent years, frequent schema changes, managing explosive data volumes, real-time query response times, and the demand for smarter data activation have made the advantages of graph models apparent.

[0076] Graph databases are databases that use nodes and edges as basic storage units. They are efficient databases that store and query data using a graph structure, providing an intuitive visualization of relationships. They are an excellent way to store, query, and analyze highly interconnected data. The key concepts of graph databases are nodes (the points in the graph) and relationships (the edges in the graph). Relationships connect nodes together, enabling fast graph retrieval operations.

[0077] Graph databases belong to the category of NoSQL databases. Graph databases differ significantly from relational databases in their data storage, querying, and data structure. Graph data structures directly store the dependencies between nodes, while relational databases and other types of NoSQL databases represent these relationships in a less direct way. Graph databases store the relationships between data as part of the data structure, and labels, directions, and attributes can be added to these relationships. In contrast, queries on relationships in other databases must be performed at runtime, which is why graph databases have a significant performance advantage in relational queries compared to other types of databases.

[0078] Graph databases have many advantages, such as:

[0079] (1) Using graphs to express relationships in the real world is direct, natural, and easy to model.

[0080] (2) Graph databases can efficiently insert large amounts of data. The application areas targeted by graph databases may have relatively large data volumes, such as knowledge graphs, social relationships, and risk control relationships. The total data volume is generally in the hundreds of millions or billions, and some even reach tens of billions of edges. Without table partitioning and database sharding, the data processing efficiency of MySQL drops significantly after inserting millions of data, while graph databases can basically handle data of hundreds of millions or more. For example, graph databases such as Neo4j, Titan (Janus), and Hugegraph can still maintain high data processing efficiency even when continuously inserting billions of data.

[0081] (3) Graph databases can efficiently query related data. Traditional relational databases are not good at performing relational queries, especially multi-level relations, because they generally require table joins, which are cumbersome operations involving a lot of I / O operations and memory consumption. Graph databases generally perform targeted optimizations for relational queries, such as in storage models, data structures, and query algorithms, to prevent queries of local data from triggering the reading of all data.

[0082] (4) Graph databases provide query languages ​​for graph retrieval, such as Gremlin and Cypher. Graph query languages ​​greatly facilitate the continuous development of association analysis business. Traditional solutions often require modification of data storage models and complex query scripts when requirements change, while graph databases have already optimized the abstract expression of business.

[0083] (5) Graph databases provide professional analysis algorithms and tools. For example, ShortestPath, PageRank, PersonalRank, Louvain, etc. Some graph databases also provide batch data import tools and visual graph display interfaces, making the data analysis results more intuitive.

[0084] This application, in the process of internet AI outbound calling, compares and filters the acquired dialogue data with data in a database in real time. Then, it further determines the matching degree between the dialogue data and the database data from the filtered data. Finally, it extracts all data from the database that meets the matching degree requirements to form the final returned dataset, which serves as reference data for the next dialogue step for the internet AI outbound calling system. The data extraction process of this application involves filtering the data in the database before extracting all data that meets the matching degree requirements, determining the matching degree of the data to be extracted based on the filtering, and finally extracting the data that meets the matching degree requirements. This layered filtering and extraction method improves the accuracy and efficiency of data extraction while reducing the load on data stream processing.

[0085] Furthermore, obtaining the required dialogue data from all dialogue information and data generated during the Internet AI outbound calling process includes:

[0086] From all the dialogue information and data generated by the Internet AI outbound calling system, select the first dialogue set, and then extract the first dataset from the first dialogue set;

[0087] The first dialogue set consists of all data including at least the most recent dialogue between the user and the Internet AI outbound calling system;

[0088] Specifically, the data in the first dialogue set includes, but is not limited to, data in various forms such as text, voice, and images;

[0089] The first dataset consists of core data that fully reflects the semantics after filtering out data in the first dialogue set that is irrelevant to the core semantics. The data types include nodes and relationships.

[0090] Specifically, the process of extracting the first dataset from the first dialogue set can be achieved using existing semantic recognition technology; while the nodes and relationships in the first dataset are basic data types in graph databases; nodes include node labels (i.e., categories) and node attributes (i.e., the actual data in the nodes), and relationships include relationship direction and relationship attributes (i.e., the actual data in the relationships).

[0091] Since this application selects the first dialogue set from all dialogue information and data generated in the Internet AI outbound calling process, and then extracts the first dataset after filtering from the first dialogue set, the data in the final first dataset is the effective data after removing useless noise data. This not only reduces the complexity of subsequent data processing, but also greatly reduces the error rate of subsequent data processing and improves the accuracy of subsequent data processing.

[0092] Furthermore, the process of comparing and filtering the acquired dialogue data with data in the database to determine the matching degree between the dialogue data during the operation of the Internet AI outbound calling system and the data in the database includes:

[0093] The data in the first dataset is compared with the data in the database, and the first data filtering is performed. All the filtered data groups are combined to form the first returned dataset.

[0094] The first data filtering involves comparing the data in the first dataset with the data in the database. If the similarity between the node attributes in the database and the node attributes in the first dataset is greater than or equal to a first similarity threshold, then all data groups related to that node are extracted as the filtering result of the first data filtering process. The first returned dataset includes all data groups filtered out by the first data filtering process.

[0095] Specifically, since the basic structure of a graph database consists of nodes and relations, the data filtered after data comparison in the database is output in the form of data groups (or data chains) composed of multiple nodes and relations. Therefore, the specific data form of the first returned dataset is one or more data groups; the similarity of the node attributes is determined by existing similarity algorithms or better similarity algorithms, such as text similarity algorithms.

[0096] Determine the matching degree between each data group in the first returned dataset and the data in the first dataset. That is, each data group corresponds to a first matching degree relative to the first dataset, and sort the matching degrees.

[0097] The first matching degree is used to represent the degree of matching between each data group in the first returned dataset and the data in the first dataset, that is, the degree of correlation. The higher the first matching degree, the higher the correlation.

[0098] Since the first returned dataset is extracted from the first dataset after the first data filtering and first matching degree filtering process, the extraction range is continuously narrowed through layers of filtering. Therefore, the final extracted first returned dataset is a collection of all data groups in the database that have a high degree of matching or high correlation with the first dataset. In this way, for the subsequent data processing process, not only is the difficulty of subsequent data processing simplified, but the matching degree of data is also improved, ensuring the accuracy of subsequent data.

[0099] Furthermore, determining the matching degree between each data group in the first returned dataset and the data in the first dataset, i.e., determining the first matching degree, is to obtain the first matching degree corresponding to each data group by comprehensively calculating each data parameter in each data group in the first returned dataset, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes.

[0100] Furthermore, the step of extracting all data from the database that meets the matching requirements to form the final returned dataset includes:

[0101] Extract all data groups in the first returned dataset whose first matching degree is greater than or equal to the first extraction threshold, and combine all the extracted data groups to form the final returned dataset.

[0102] The first extraction threshold is the threshold that satisfies the first matching degree requirement of the final returned dataset;

[0103] Furthermore, the step of extracting all data groups that meet the first matching degree requirement from the first returned dataset and assembling all the extracted data groups into the final returned dataset also includes:

[0104] If at least one data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then all data groups in the first returned dataset with a first matching degree greater than or equal to the first extraction threshold will be extracted and combined to form the final returned dataset.

[0105] If no data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then further determination is made based on whether the first matching degree of the data group in the first returned dataset is greater than or equal to the second extraction threshold, as follows:

[0106] a) If the first matching degree of at least one data group in the first returned dataset is greater than or equal to the second extraction threshold, then all data groups in the first returned dataset with the first matching degree greater than or equal to the second extraction threshold are extracted and combined to form the final returned dataset;

[0107] b) If no data group in the first returned dataset has a first matching degree greater than or equal to the second extraction threshold, then extract the second dataset from the second dialogue set, compare each data group in the first returned dataset with the data in the second dataset, perform second data filtering, and form the second returned dataset from the filtered data groups; then determine the second matching degree of each data group in the second returned dataset.

[0108] Finally, extract all data groups in the second returned dataset whose second matching degree is greater than or equal to the first extraction threshold, and combine them to form the final returned dataset;

[0109] Specifically, in actual implementation, the final returned dataset is not always obtained; there are cases where the final returned dataset cannot be obtained, as detailed below:

[0110] (1) If the first matching degree of any data group in the first returned dataset is greater than or equal to the first extraction threshold, it will be considered that the usable final returned dataset cannot be obtained, that is, the data returned to the Internet AI outbound calling system is empty;

[0111] (2) If the second matching degree of any data group in the second returned dataset is greater than or equal to the first extraction threshold, it will be considered that the usable final returned dataset cannot be obtained, that is, the data returned to the Internet AI outbound calling system is empty;

[0112] The second matching degree is used to represent the degree of matching between each data group in the second returned dataset and the data in the first and second datasets, i.e., the degree of correlation. The higher the second matching degree, the higher the correlation. The second matching degree is obtained by multiplying the first matching degree and the second matching degree correction coefficient. The second matching degree correction coefficient is obtained by comprehensively calculating the parameters of each data group in the second returned dataset, which means that the factors affecting the first matching degree of the data in the second dataset are added and the final matching degree result is corrected.

[0113] The second data filtering involves comparing each data group in the first returned dataset with the data in the second dataset. If the similarity between any node attribute of each data group in the first returned dataset and the node attribute in the first dataset is greater than or equal to a second similarity threshold, then that data group is extracted as the filtering result of the second data filtering process. The second returned dataset includes all data groups filtered out by the second data filtering process.

[0114] The second dialogue set is a set of dialogues whose dialogue time is before the first dialogue set and whose information and data do not overlap with those of the first dialogue set; the second dataset is the core data that can fully reflect the semantics after filtering out data in the second dialogue set that is irrelevant to the core semantics, and includes data types of nodes and relationships.

[0115] Both the second extraction threshold and the first extraction threshold are preset thresholds that meet the matching degree requirements of the final returned dataset, and the second extraction threshold is less than the first extraction threshold;

[0116] This application considers different scenarios regarding whether data groups in the first returned dataset meet the first extraction threshold. If the first extraction threshold is not met, but the second extraction threshold is, a comprehensive comparison with data from the second dataset is required. Therefore, based on matching the first dataset, factors for matching the second dataset are introduced. Specifically, the matching degree between the first and second datasets is used as a comprehensive factor, and the first matching degree is corrected with a correction coefficient to obtain the second matching degree. This second matching degree is then compared with the first extraction threshold to re-select data groups that meet the extraction requirements of the final returned dataset. In this way, based on the actual matching degree, through reasonable data extraction methods and accurate matching degree relationship models, the application can comprehensively consider various different situations in the data extraction process and develop reasonable and effective data extraction strategies for different situations, greatly improving the efficiency and accuracy of data extraction.

[0117] Furthermore, the second matching degree is calculated using the second matching degree relationship model;

[0118] The second matching degree relationship model includes:

[0119]

[0120] in, This represents the second degree of matching for the data set;

[0121] This is the second matching degree correction coefficient for the data set;

[0122] The first match degree of the data set;

[0123] The data groups mentioned in the above parameters are all the same data groups in the second returned dataset;

[0124] Furthermore, the second matching degree correction coefficient in the second matching degree relationship model is calculated using the correction formula.

[0125] Specifically, the data in the second dataset falls into two categories:

[0126] (1) Single data case, that is, the data in the second dataset can be a single node, or a "node-relationship" group composed of a single node and a single relation;

[0127] (2) Multiple data cases, that is, the data in the second dataset can be multiple single nodes, multiple "node-relationship" groups, or data composed of at least one single node and at least one "node-relationship" group;

[0128] Therefore, in the actual implementation process, there are two situations when determining the second matching degree:

[0129] (1) For single data cases, when determining the second matching degree of each data group, after the second data filtering, the parameters of a single node or a "node-relationship" group (i.e., the number of nodes, labels, attributes, and the number, direction, attributes, etc. of the relationship) will be used as the basis for determining the second matching degree of each data group, and the specific second matching degree will be calculated using the second matching degree relationship model.

[0130] (2) For the case of multiple data, when determining the second matching degree of each data group, the second data filtering and the second matching degree calculation will be performed independently for each single data (i.e., a single node or a "node-relationship" group) in the case of multiple data. In this way, if it is necessary to filter and sort a single data by second matching degree, the required result can be obtained quickly by calculating the second matching degree separately. If it is necessary to perform comprehensive second matching degree filtering and sorting on several single data or all single data in the second dataset, the second matching degree of several single data or all single data can be comprehensively calculated (such as summation, weighted summation, etc.) to quickly obtain the required result.

[0131] In summary, this application improves the data processing efficiency and data extraction accuracy of the Internet AI outbound calling system through reasonable and effective data processing and extraction methods.

[0132] Example 2

[0133] like Figure 3 As shown, this application provides a data extraction system for internet AI-based outbound calling, the system comprising:

[0134] The acquisition module is used to obtain the required dialogue data from all the dialogue information and data generated during the Internet AI outbound calling process;

[0135] The processing module is used to compare and filter the dialogue information and data obtained during the Internet AI outbound call process with the data in the database to determine the matching degree between the dialogue information and data during the operation of the Internet AI outbound call system and the data in the database.

[0136] The output module is used to extract all data that meets the matching requirements from the database, form the final return dataset, and return it to the dialogue system of the Internet AI outbound calling system as reference data for the next dialogue.

[0137] All dialogue information and data generated during the operation of the Internet AI outbound calling system refers to all dialogue information and data generated when the AI ​​interacts with the user in the Internet AI outbound calling system.

[0138] The matching degree is used to characterize the degree of matching between the data obtained after comparison and filtering in the database and the dialogue data, i.e., the degree of relevance; the database is a graph database, and includes at least the local data of the Internet AI outbound calling system and searchable Internet data.

[0139] Specifically, the local data refers to the matched and optimized data stored internally within the Internet AI outbound calling system; the final returned dataset is returned to the dialogue system of the Internet AI outbound calling system, which involves converting the final returned dataset into different forms of data such as text, voice, and images, and outputting it to the dialogue platform of the Internet AI outbound calling system for interaction with customers.

[0140] Furthermore, obtaining the required dialogue data from all dialogue information and data generated during the Internet AI outbound calling process includes:

[0141] From all the dialogue information and data generated by the Internet AI outbound calling system, select the first dialogue set, and then extract the first dataset from the first dialogue set;

[0142] The first dialogue set consists of all data including at least the most recent dialogue between the user and the Internet AI outbound calling system;

[0143] The first dataset consists of core data that fully reflects the semantics after filtering out data in the first dialogue set that is irrelevant to the core semantics. The data types include nodes and relationships.

[0144] Furthermore, the process of comparing and filtering the acquired dialogue data with data in the database to determine the matching degree between the dialogue data during the operation of the Internet AI outbound calling system and the data in the database includes:

[0145] The data in the first dataset is compared with the data in the database, and the first data filtering is performed. All the filtered data groups are combined to form the first returned dataset.

[0146] The first data filtering involves comparing the data in the first dataset with the data in the database. If the similarity between the node attributes in the database and the node attributes in the first dataset is greater than or equal to a first similarity threshold, then all data groups related to that node are extracted as the filtering result of the first data filtering process. The first returned dataset includes all data groups filtered out by the first data filtering process.

[0147] Determine the matching degree between each data group in the first returned dataset and the data in the first dataset. That is, each data group corresponds to a first matching degree relative to the first dataset, and sort the matching degrees.

[0148] The first matching degree is used to represent the degree of matching between each data group in the first returned dataset and the data in the first dataset, that is, the degree of correlation. The higher the first matching degree, the higher the correlation.

[0149] Furthermore, determining the matching degree between each data group in the first returned dataset and the data in the first dataset, i.e., determining the first matching degree, is to obtain the first matching degree corresponding to each data group by comprehensively calculating each data parameter in each data group in the first returned dataset, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes.

[0150] Furthermore, the step of extracting all data from the database that meets the matching requirements to form the final returned dataset includes:

[0151] Extract all data groups in the first returned dataset whose first matching degree is greater than or equal to the first extraction threshold, and combine all the extracted data groups to form the final returned dataset.

[0152] The first extraction threshold is the threshold that satisfies the first matching degree requirement of the final returned dataset;

[0153] Furthermore, the step of extracting all data groups that meet the first matching degree requirement from the first returned dataset and assembling all the extracted data groups into the final returned dataset also includes:

[0154] If at least one data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then all data groups in the first returned dataset with a first matching degree greater than or equal to the first extraction threshold will be extracted and combined to form the final returned dataset.

[0155] If no data group in the first returned dataset has a first matching degree greater than or equal to the first extraction threshold, then further determination is made based on whether the first matching degree of the data group in the first returned dataset is greater than or equal to the second extraction threshold, as follows:

[0156] a) If the first matching degree of at least one data group in the first returned dataset is greater than or equal to the second extraction threshold, then all data groups in the first returned dataset with the first matching degree greater than or equal to the second extraction threshold are extracted and combined to form the final returned dataset;

[0157] b) If no data group in the first returned dataset has a first matching degree greater than or equal to the second extraction threshold, then extract the second dataset from the second dialogue set, compare each data group in the first returned dataset with the data in the second dataset, perform second data filtering, and form the second returned dataset from the filtered data groups; then determine the second matching degree of each data group in the second returned dataset.

[0158] Finally, extract all data groups in the second returned dataset whose second matching degree is greater than or equal to the first extraction threshold, and combine them to form the final returned dataset;

[0159] The second matching degree is used to represent the degree of matching between each data group in the second returned dataset and the data in the first and second datasets, i.e., the degree of correlation. The higher the second matching degree, the higher the correlation. The second matching degree is obtained by multiplying the first matching degree and the second matching degree correction coefficient. The second matching degree correction coefficient is obtained by comprehensively calculating the parameters of each data group in the second returned dataset, which means that the factors affecting the first matching degree of the data in the second dataset are added and the final matching degree result is corrected.

[0160] The second data filtering involves comparing each data group in the first returned dataset with the data in the second dataset. If the similarity between any node attribute of each data group in the first returned dataset and the node attribute in the first dataset is greater than or equal to a second similarity threshold, then that data group is extracted as the filtering result of the second data filtering process. The second returned dataset includes all data groups filtered out by the second data filtering process.

[0161] The second dialogue set is a set of dialogues whose dialogue time is before the first dialogue set and whose information and data do not overlap with those of the first dialogue set; the second dataset is the core data that can fully reflect the semantics after filtering out data in the second dialogue set that is irrelevant to the core semantics, and includes data types of nodes and relationships.

[0162] Both the second extraction threshold and the first extraction threshold are preset thresholds that meet the matching degree requirements of the final returned dataset, and the second extraction threshold is less than the first extraction threshold;

[0163] Furthermore, the second matching degree is calculated using the second matching degree relationship model;

[0164] The second matching degree relationship model includes:

[0165]

[0166] in, This represents the second degree of matching for the data set;

[0167] This is the second matching degree correction coefficient for the data set;

[0168] The first match degree of the data set;

[0169] The data groups mentioned in the above parameters are all the same data groups in the second returned dataset.

[0170] Example 3

[0171] like Figure 4 As shown, this application provides a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the method described in Embodiment 1 above.

[0172] Example 4

[0173] like Figure 5 As shown, this application provides a computer device, characterized in that it includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to implement the method described in Embodiment 1 above.

[0174] Compared with existing technologies, this application improves the data processing efficiency and data extraction accuracy of the Internet AI outbound calling system through reasonable and effective data processing and data extraction methods.

[0175] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, media, devices, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0176] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0177] The modules or units described as separate components may or may not be physically separate. The components shown as modules or units may or may not be physical modules or units; that is, they may be located in one place or distributed across multiple network modules or units. Some or all of the modules or units can be selected to achieve the purpose of this embodiment according to actual needs.

[0178] Furthermore, the functional modules or units in the various embodiments of this application can be integrated into one processing module or unit, or each module or unit can exist physically separately, or two or more modules or units can be integrated into one module or unit. The integrated modules or units described above can be implemented in hardware or as software functional units.

[0179] The integrated systems, modules, and units, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them; although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1.A data extraction method based on Internet AI outbound call, characterized in that, The method comprises: From all the conversation information and data generated in the Internet AI outbound process, obtain the required conversation data; including: selecting a first conversation set, and then extracting a first data set from the first conversation set; the first conversation set is all the data including at least the last conversation between the user and the Internet AI outbound system; the first data set is the core data that can fully reflect the semantics after filtering the data irrelevant to the determined core semantics in the first conversation set, including data types of nodes and relationships; Compare and filter the obtained conversation data with the data in the database to determine the matching degree of the conversation data in the Internet AI outbound system with the data in the database; extract all the data with a matching degree meeting the requirements from the database to form a final return data set, including: comparing the data in the first data set with the data in the database, and performing first data filtering to form a first return data set; determine the first matching degree between each data group in the first return data set and the data in the first data set, and sort the matching degrees; extract all the data groups in the first return data set with a first matching degree greater than or equal to a first extraction threshold to form a final return data set; If none of the data groups in the first return data set has a first matching degree greater than or equal to the first extraction threshold, further determine whether the first matching degree of each data group in the first return data set is greater than or equal to a second extraction threshold, as follows: If none of the data groups in the first return data set has a first matching degree greater than or equal to the second extraction threshold, extract a second data set from a second conversation set, compare each data group in the first return data set with the data in the second data set, perform second data filtering, and then form a second return data set from the data groups in the filtering result; and then determine the second matching degree of each data group in the second return data set; Finally, extract all the data groups in the second return data set with a second matching degree greater than or equal to the first extraction threshold, and form a final return data set; The second matching degree is obtained by multiplying the first matching degree and a second matching degree correction coefficient; the second matching degree correction coefficient is obtained by comprehensive operation of the parameters of each data group in the second return data set, indicating that the influencing factors of the data in the second data set on the first matching degree are added and the final matching degree result is corrected; The second conversation set is a conversation set in all the conversation information and data, with a conversation time before the first conversation set and information and data not repeated with the first conversation set; the second data set is the core data that can fully reflect the semantics after filtering the data irrelevant to the determined core semantics in the second conversation set, including data types of nodes and relationships; The return data set is used as reference data for the next conversation in the conversation system of the Internet AI outbound system; The database is a graph database, and at least includes local data of the Internet AI outbound system and searchable Internet data. 2.The data extraction method based on Internet AI outbound call according to claim 1, characterized in that, The acquired conversation data is subjected to data comparison and screening with data in the database to determine a matching degree of the conversation data in the operation of the Internet AI outbound call system and the data in the database, comprising: The first data screening is to compare the data in the first data set with the data in the database, and if the node attribute of the data in the database is similar to the node attribute in the first data set with a similarity greater than or equal to a first similarity threshold, all data groups related to the node are extracted as the screening result of the first data screening process; the first returned data set comprises all data groups screened by the first data screening process; The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, each data group corresponds to a first matching degree relative to the first data set, and the matching degrees are sorted; The first matching degree is used to represent the matching degree, that is, the relevance, between each data group in the first returned data set and the data in the first data set, and the higher the first matching degree, the higher the relevance. 3.The data extraction method based on Internet AI outbound call according to claim 2, characterized in that, The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, the first matching degree is determined by comprehensively calculating each data parameter in each data group in the first returned data set, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes. 4.The data extraction method based on Internet AI outbound call according to claim 2, characterized in that, The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, the first matching degree is determined by comprehensively calculating each data parameter in each data group in the first returned data set, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes. The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, the first matching degree is determined by comprehensively calculating each data parameter in each data group in the first returned data set, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes. The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, the first matching degree is determined by comprehensively calculating each data parameter in each data group in the first returned data set, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes. 5.The data extraction method based on Internet AI outbound call according to claim 4, characterized in that, The matching degree between each data group in the first returned data set and the data in the first data set is determined, that is, the first matching degree is determined by comprehensively calculating each data parameter in each data group in the first returned data set, including the number of nodes, the similarity of node attributes, the number of relationships, and the similarity of relationship attributes. The second extraction threshold and the first extraction threshold are both preset thresholds that meet the requirement of the matching degree of the final returned data set, and the second extraction threshold is smaller than the first extraction threshold. ​ ​ ​ 6. A data extraction system based on Internet AI outbound call, characterized in that, The system is used for executing the data extraction method based on the Internet AI outbound call as claimed in any one of claims 1-5, and comprises: an acquisition module, configured to acquire required conversation data from all conversation information and data generated in the Internet AI outbound call process, including: selecting a first conversation set, and extracting a first data set from the first conversation set; the first conversation set is all data including at least the latest conversation between a user and the Internet AI outbound call system; the first data set is core data capable of completely reflecting semantics after filtering data irrelevant to the determined core semantics in the first conversation set, and includes data types of nodes and relationships; a processing module, configured to perform data comparison and screening between the acquired conversation data and data in a database, and determine a matching degree between conversation information and data in a running process of the Internet AI outbound call system and the data in the database; an output module, configured to extract all data with a matching degree reaching a requirement from the database to form a final return data set, including: comparing data in the first data set with data in the database, performing first data screening, and forming a first return data set; extracting all data groups with a first matching degree greater than or equal to a first extraction threshold value in the first return data set to form the final return data set; if the first matching degree of any data group in the first return data set is not greater than or equal to the first extraction threshold value, further determination is made according to whether the first matching degree of the data group in the first return data set is greater than or equal to a second extraction threshold value, and the determination is as follows: if the first matching degree of any data group in the first return data set is not greater than or equal to the second extraction threshold value, a second data set is extracted from a second conversation set, data in each data group in the first return data set is compared with data in the second data set, second data screening is performed, and data groups in the screening result are formed into a second return data set; and a second matching degree of each data group in the second return data set is determined; finally, all data groups with a second matching degree greater than or equal to the first extraction threshold value in the second return data set are extracted to form the final return data set; the second matching degree is obtained by multiplying the first matching degree and a second matching degree correction coefficient; the second matching degree correction coefficient is obtained by comprehensive operation of parameters of each data group in the second return data set, and represents that an influencing factor of data in the second data set on the first matching degree is added and the final matching degree result is corrected; the second conversation set is a conversation set with a conversation time before the first conversation set and without repeated information and data with the first conversation set in all conversation information and data; the second data set is core data capable of completely reflecting semantics after filtering data irrelevant to the determined core semantics in the second conversation set, and includes data types of nodes and relationships; the conversation system returns the reference data for the next conversation to the Internet AI outbound call system for use by the Internet AI outbound call system; the database is a graph database, and at least includes local data of the Internet AI outbound call system and searchable Internet data.

Images