CN114363466B - Intelligent cloud calling system based on AI - Google Patents
Intelligent cloud calling system based on AI Download PDFInfo
- Publication number
- CN114363466B CN114363466B CN202210279643.9A CN202210279643A CN114363466B CN 114363466 B CN114363466 B CN 114363466B CN 202210279643 A CN202210279643 A CN 202210279643A CN 114363466 B CN114363466 B CN 114363466B
- Authority
- CN
- China
- Prior art keywords
- standard
- length
- user
- data volume
- correction coefficient
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Telephonic Communication Services (AREA)
Abstract
The invention relates to an intelligent calling cloud system based on AI, the system includes a receiving module for receiving the calling request of the user; the analysis module is connected with the receiving module and used for analyzing the call request to obtain an analysis result; the storage module analyzes the corresponding relation between the result and the solution; and the display module acquires the solution according to the analysis result and the corresponding relation and displays the solution. The user portrait can be determined through the user request, the matching performance of the solution and the call request is improved according to the user portrait and the analysis result after the user request is analyzed, the solution displayed by the display module can be accurately matched with the call request, the actual request of the user is solved, the corresponding relation between the analysis result and the solution is established in advance, the process of retrieval and search is reduced, the extraction time of the solution is greatly shortened, and the response time length based on the user call request is effectively reduced.
Description
Technical Field
The invention relates to the technical field of call response data processing, in particular to an intelligent call cloud system based on AI.
Background
With the continuous development of society, more and more time processing begins to incline towards artificial intelligence, firstly, the increase of labor cost, secondly, the continuous development of artificial intelligence, so that the event processed by the artificial intelligence has high processing efficiency, and the coping processing is suitable.
Document publication No. CN113596267A discloses an artificial intelligence call system, which includes: the call control center is used for generating a call connection task, managing and controlling and distributing the call connection task, generating and sending a connection task distribution signal, receiving a call transfer task, managing and controlling and distributing the call transfer task, and generating and sending a transfer task distribution signal, wherein the call connection task comprises a call access task and a call calling task; the AI calling module is used for receiving and executing the connection task distribution signal, acquiring and receiving a sound signal, carrying out intelligent voice recognition according to the sound signal to generate a voice recognition result, searching a response library according to the recognition result, generating and outputting a response result, judging a manual calling requirement according to the voice recognition result, and generating and sending the call forwarding task; the manual calling module is used for receiving the switching task distribution signal, generating an execution request of the switching task distribution signal, receiving an execution request feedback signal and executing the switching task distribution signal according to the execution request feedback signal; and the post-processing learning module is used for generating a response feedback result according to the AI calling module and the artificial calling module, traversing, comparing and updating a response library according to the response feedback result, wherein the response feedback result is used for representing the satisfaction degree of a connection object in a call connection task executed by the artificial intelligent calling system to the response.
In the prior art, after a user request is obtained, the user needs to search in a response library according to the user requirement to generate and output a response result, and as the content in the response library is continuously increased, the time for searching in the response library to generate the response result is continuously increased, so that the fluency of the response obtained by the user is influenced.
Disclosure of Invention
Therefore, the invention provides an AI-based intelligent cloud calling system, which can solve the problem that the time spent for a user to answer in the prior art is long.
In order to achieve the above object, the present invention provides an AI-based intelligent calling cloud system, comprising:
the receiving module receives a call request of a user;
the analysis module is connected with the receiving module and used for analyzing the call request to obtain an analysis result;
the storage module analyzes the corresponding relation between the result and the solution;
the display module is used for acquiring a solution according to the analysis result and the corresponding relation and displaying the solution;
the analysis module comprises a data extraction unit, a data storage unit and a data processing unit;
the data extraction unit is used for extracting the call request, removing impurity information in the call request, acquiring request information and detecting the actual data volume of the request information;
a standard data volume D0 and an actual data volume D of the request information are preset in the data storage unit;
if the actual data volume D is larger than or equal to the standard data volume D0, performing secondary extraction on the call request;
if the actual data volume D is smaller than the standard data volume D0, the data volume is normal, and the request information is stored;
and the data processing unit is used for determining the length of a processing window for the request information according to the actual data quantity when secondary extraction is carried out.
Further, a first standard length l10, a second standard length l20 and a third standard length l30 of a processing window are preset in the data processing unit, and l10> l20> l30> 0;
if the request information is voice information, acquiring voiceprint information in the voice information, determining the actual age of the user according to the voiceprint information, and selecting a first standard length l10, a second standard length l20 or a third standard length l30 according to the actual age of the user;
a standard age range A12 is preset in the data processing unit, and if the actual age of the user is smaller than the minimum value A1 in the standard age range A12, a first standard length l10 is selected as the length of the processing window;
selecting a second standard length l20 as the length of the processing window if the actual age of the user belongs to the standard age group range a 12;
if the actual age of the user is greater than the maximum value a2 in the standard age group range a12, the third standard length l30 is selected as the length of the processing window.
Further, when the actual data volume D is larger than or equal to the standard data volume D0, correcting the standard length according to the amplitude of the deviation of the actual data volume from the standard data volume;
when the ith standard length li0 is selected as the length of the processing window, if the actual data volume D is more than or equal to the standard data volume D0 by 2 × D0>, the ith standard length li0 is corrected by adopting a first correction coefficient k 1;
if the actual data volume D is more than or equal to 5 multiplied by D0 and more than or equal to 2 multiplied by the standard data volume D0, the ith standard length li0 is corrected by adopting a second correction coefficient k 2;
if the actual data amount D is equal to or larger than 5 × D0, the i-th standard length li0 is corrected by using the third correction coefficient k3, and the first correction coefficient k1< the second correction coefficient k2< the third correction coefficient k 3.
Further, when the i-th standard length li0 is corrected by using the first correction coefficient k1, the corrected standard length is li 0' = li0 × (1+ k 1);
when the i-th standard length li0 is corrected by using the second correction coefficient k2, the corrected standard length li 0' = li0 × (1+ k 2);
when the i-th standard length li0 is corrected by using the third correction coefficient k3, the corrected standard length li 0' = li0 × (1+ k 3), i =1, 2, and 3.
Further, the first correction coefficient k1= l10/(l10+ l20+ l30);
the second correction coefficient k2= l20/(l10+ l20+ l30);
the third correction coefficient k3= l30/(l10+ l20+ l 30).
Further, the system also comprises a reply module, wherein the reply module is connected with the display module and is used for replying to the user within a reply time length according to the content displayed by the display module, and the reply time length is set to T0.
Further, a plurality of keywords are preset in the reply module, and the keywords are used for representing the professionalism in the display content;
the reply module is used for adjusting the response time length according to the number of the keywords.
Further, a first adjusting parameter α and a second adjusting parameter β are set in the reply module, and a standard number n0 is also set in advance;
if the number of the keywords in the display content is less than or equal to the standard number n0, indicating that the professional degree of the display content is not high, and adjusting the response time length by adopting a first adjustment parameter alpha;
if the number of keywords in the display content is greater than the standard number n0, which indicates that the professional degree of the display content is high, the response time length is adjusted by using a second adjustment parameter β.
Further, after the response time length is adjusted by using the first adjustment parameter α, a new response time length is used as the response time length in the next call response period, and the new response time length is T1' = T0 × (1+ α);
and after the response time length is adjusted by adopting a second adjustment parameter beta, taking a new response time length as the response time length in the next call response period, wherein the new response time length is T2' = T0 x (1+ beta), and 0< alpha < beta < 1.
Compared with the prior art, the method and the device have the advantages that the user portrait can be determined through the user request, the solution with higher precision based on the user request can be determined according to the user portrait and the analysis result after the user request is analyzed, the matching performance of the solution and the call request is improved, the solution displayed by the display module can be accurately matched with the call request, the actual request of the user is solved, the corresponding relation between the analysis result and the solution is established in advance, the process of searching is reduced, the extraction time of the solution is greatly shortened, and the response time length based on the user call request is effectively reduced.
Particularly, whether impurity information exists in the time data volume is determined by judging the size of the actual data volume, in practical application, when a user initiates a call request, for example, when a call is made, whether noise exists in the environment and is carried in the actual data, so that the actual data volume is increased, a standard data volume D0 is set, when the actual data volume D is less than the standard data volume D0, the data volume is normal, the request information is stored and analyzed, an analysis result is obtained, if the actual data volume D is more than or equal to the standard data volume D0, the actual data needs to be processed, the impurity information in the actual data is decoupled, namely the impurity information is extracted from the actual data, the original data information is retained and analyzed, and when the impurity information is extracted, the processing window length is adopted for window division processing, the processing efficiency is improved, the processing is not easy to leak, the accuracy of impurity information processing is improved, and the processing efficiency is improved.
Particularly, the actual age of the user is judged to determine whether the user belongs to the standard age range A12, if the actual age is smaller than the minimum value of the standard age range A12, the actual user is younger and less tolerant, therefore, when the user request is processed, the first standard length l10 is selected as the length of the processing window, the first standard length is bigger, the processing speed for the actual user request is fast, the waiting time of the user is reduced, the analysis result of the request information is conveniently and rapidly acquired, the actual waiting time of the user is reduced, the user calling experience is improved, if the actual age of the user is older, the tolerance is better, the third standard length l30 can be adopted as the length of the processing window, the processing after dividing the majority of the user request information is realized, the processing granularity for the user request information is improved by sacrificing the variable of the processing speed, the speed and the processing precision of the user call request are balanced, and the user requirements are met.
Particularly, the standard length is corrected by the range of the actual data volume deviating from the standard data volume through the correction coefficient, if the actual data volume D is larger than or equal to 2 × D0 and larger than or equal to the standard data volume D0, the ith standard length li0 is corrected by the first correction coefficient k1, if the actual data volume D is larger than or equal to 5 × D0 and larger than or equal to 2 × standard data volume D0, the ith standard length li0 is corrected by the second correction coefficient k2, if the actual data volume D is larger than or equal to 5 × D0, the ith standard length li0 is corrected by the third correction coefficient k3, the first correction coefficient k1 is smaller than the second correction coefficient k2 and smaller than the third correction coefficient k3, it can be seen that different correction coefficients are adopted for different deviation ranges, so that the actual data volume of the standard length is more in line with the actual data volume requirement, and the data processing efficiency is improved.
Particularly, the call information is cut in the next call request by using the corrected standard length, so that the analysis process of the request information is more efficient, the processing efficiency of the request information is improved, the balance between the analysis precision and the analysis speed is realized, the processing of the call request is in an intelligent adjustment state, dynamic intelligent adjustment can be performed according to actual data, the efficient processing of the actual request data is improved, the corresponding correction coefficients are represented by using the preset first standard length l10, the preset second standard length l20 and the preset third standard length l30, the forward relation between the standard length and the correction coefficients is realized, and when the standard length is large, the correction coefficients are also large, so that the request data can be effectively processed, and the data processing efficiency is improved.
Particularly, by setting the response module, the content displayed by the display module is responded to the user within the response time length, the response based on the user request information is completed, the matching degree of the user calling and the response is improved, and the response efficiency of the user calling is improved. The response time length T0 is adopted under the universality, the response time length is used as the reference time for responding to the user, the response time length is limited, the output speed of the display content of the display module is improved, the accurate control on the call response is improved, and the response efficiency is improved.
Particularly, a plurality of keywords are preset in the reply module, if more keywords are in the display content, the display content is high in professional degree, if the number of the keywords is small, the display content is not high in professional degree, different professional contents are processed in different response time lengths, and when the display content with high professional degree is responded, a user needs to respond in a longer time, so that the user can fully explain the display content and receive the display content conveniently; if the displayed content is not high in professional degree, the content can be output in a short time, so that the difficulty degree of the displayed content is effectively adjusted, the response time is effectively and reasonably distributed, and the response content output efficiency is improved.
Particularly, by setting the standard number n0, a first adjustment parameter α or a second adjustment parameter β is selected for the relationship between the number of keywords in the display content and the standard number n0, so that the length of the response time more meets the actual content requirement.
Drawings
Fig. 1 is a schematic structural diagram of an AI-based smart calling cloud system according to an embodiment of the present invention.
Detailed Description
In order that the objects and advantages of the invention will be more clearly understood, the invention is further described below with reference to examples; it should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only for explaining the technical principles of the present invention, and do not limit the scope of the present invention.
It should be noted that in the description of the present invention, the terms of direction or positional relationship indicated by the terms "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings, which are only for convenience of description, and do not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
Furthermore, it should be noted that, in the description of the present invention, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
Referring to fig. 1, an embodiment of the present invention provides an AI-based intelligent calling cloud system, which includes:
a receiving module 10 for receiving a call request of a user;
the analysis module 20 is connected with the receiving module and used for analyzing the call request to obtain an analysis result;
the storage module 30 analyzes the corresponding relation between the result and the solution;
and the display module 40 acquires a solution according to the analysis result and the corresponding relation, and displays the solution.
Specifically, the application scenario of the intelligent cloud calling system based on the AI in the embodiment of the invention can be based on a recommendation engine and a machine learning technology, the potential needs of the user are known by analyzing the user data and clustering the user characteristics, the feasibility and the success rate of the needs are analyzed, and the intelligent pushing of 'thousands of people and thousands of faces' is realized. When a user registers, related information of user development history such as user registration login information, log data information, user form information and the like is acquired to make a user portrait for a data source, labels and dimensionality division is performed after characteristics of data information such as brand merchants, intention agency dealers and the like are extracted, a corresponding label system is constructed, a corresponding matching rule and a weight calculation method are designed according to the label system, and a preliminary result of the user portrait label is obtained after calculation. And finally, dividing group users through a clustering algorithm to form a group portrait, analyzing the individual user portrait and the group user portrait respectively through visual user portrait display, displaying a formed result on a platform to be presented to a platform manager, and feeding back a support result to the users, so that high-quality experience and push are brought to the users, and the users can accurately obtain information.
The embodiment of the invention can determine the user portrait through the user request, and can determine the solution with higher precision based on the user request according to the user portrait and the analysis result after analyzing the user request, thereby improving the matching between the solution and the call request, enabling the solution displayed by the display module to be accurately matched with the call request, and solving the actual request of the user.
Specifically, the analysis module 20 includes a data extraction unit 21, a data storage unit 22, and a data processing unit 23;
the data extraction unit is used for extracting the call request, removing impurity information in the call request, acquiring request information and detecting the actual data volume of the request information;
a standard data volume D0 and an actual data volume D of the request information are preset in the data storage unit;
if the actual data volume D is larger than or equal to the standard data volume D0, performing secondary extraction on the call request;
if the actual data volume D is less than the standard data volume D0, indicating that the data volume is normal, and storing the request information;
and the data processing unit is used for determining the length of a processing window for the request information according to the actual data quantity when secondary extraction is carried out.
Specifically, in the embodiment of the present invention, whether there is impurity information in the time data volume is determined by determining the size of the actual data volume, in practical applications, when a user makes a call request, for example, when a call is made, there is noise in the environment, and the noise is carried in the actual data, so that the actual data volume is increased, a standard data volume D0 is set, when the actual data volume D is less than the standard data volume D0, which indicates that the data volume is normal, the request information is stored and analyzed, so as to obtain an analysis result, if the actual data volume D is greater than or equal to the standard data volume D0, the actual data needs to be processed, so as to decouple the impurity information in the actual data, that is, the impurity information is extracted from the actual data, and then the original data information is retained and analyzed, when the impurity information is extracted, the processing window length is used for performing windowing processing, the processing efficiency is improved, the processing is not easy to leak, the accuracy of impurity information processing is improved, and the processing efficiency is improved.
Specifically, a first standard length l10, a second standard length l20 and a third standard length l30 of a processing window are preset in the data processing unit, and l10> l20> l30> 0;
if the request information is voice information, acquiring voiceprint information in the voice information, determining the actual age of the user according to the voiceprint information, and selecting a first standard length l10, a second standard length l20 or a third standard length l30 according to the actual age of the user;
a standard age range A12 is preset in the data processing unit, and if the actual age of the user is smaller than the minimum value A1 in the standard age range A12, a first standard length l10 is selected as the length of the processing window;
selecting a second standard length l20 as the length of the processing window if the actual age of the user belongs to the standard age group range a 12;
if the actual age of the user is greater than the maximum value a2 in the standard age group range a12, the third standard length l30 is selected as the length of the processing window.
Specifically, the embodiment of the present invention determines whether the actual age of the user belongs to the standard age range a12, and if the actual age is smaller than the minimum value of the standard age range a12, the actual age of the user is small and the endurance is poor, so when processing the user request, the first standard length l10 is selected as the length of the processing window, and the first standard length is large, so the processing speed for the actual user request is fast, the waiting time of the user is reduced, the analysis result of the request information is conveniently and quickly obtained, the actual waiting time of the user is reduced, the user call experience is improved, if the actual age of the user is large, the endurance is good, the third standard length l30 can be used as the length of the processing window, the processing after dividing the majority of the user request information is realized, the processing granularity for the user request information is improved, and the processing speed is sacrificed, the speed and the processing precision of the user call request are balanced, and the user requirements are met.
Specifically, when the actual data volume D is larger than or equal to the standard data volume D0, the standard length is corrected according to the deviation of the actual data volume from the standard data volume;
when the ith standard length li0 is selected as the length of the processing window, if the actual data volume D is more than or equal to the standard data volume D0 by 2 multiplied by D0, the ith standard length li0 is corrected by adopting a first correction coefficient k 1;
if the actual data volume D is more than or equal to 5 multiplied by D0 and is more than or equal to 2 multiplied by the standard data volume D0, correcting the ith standard length li0 by adopting a second correction coefficient k 2;
if the actual data amount D is equal to or larger than 5 × D0, the i-th standard length li0 is corrected by using the third correction coefficient k3, and the first correction coefficient k1< the second correction coefficient k2< the third correction coefficient k 3.
Specifically, in the embodiment of the present invention, the standard length is corrected by the range of the actual data amount deviating from the standard data amount through the correction coefficient, if 2 × D0> the actual data amount D is not less than the standard data amount D0, the ith standard length li0 is corrected by using the first correction coefficient k1, if 5 × D0> the actual data amount D is not less than 2 × the standard data amount D0, the ith standard length li0 is corrected by using the second correction coefficient k2, if the actual data amount D is not less than 5 × D0, the ith standard length li0 is corrected by using the third correction coefficient k3, and the first correction coefficient k1< the second correction coefficient k2< the third correction coefficient k3, it can be seen that different correction coefficients are used for different deviation ranges, so that the actual value of the standard length better meets the actual data amount requirement, and the data processing efficiency is improved.
Specifically, when the i-th standard length li0 is corrected by using the first correction coefficient k1, the corrected standard length is li 0' = li0 × (1+ k 1);
when the i-th standard length li0 is corrected by using the second correction coefficient k2, the corrected standard length li 0' = li0 × (1+ k 2);
when the i-th standard length li0 is corrected by using the third correction coefficient k3, the corrected standard length li 0' = li0 × (1+ k 3), i =1, 2, and 3.
Specifically, the first correction coefficient k1= l10/(l10+ l20+ l30);
the second correction coefficient k2= l20/(l10+ l20+ l30);
the third correction coefficient k3= l30/(l10+ l20+ l 30).
Specifically, the embodiment of the present invention cuts the call information in the next call request by using the corrected standard length, so that the analysis process of the request information is more efficient, the processing efficiency of the request information is improved, the balance between the analysis accuracy and the analysis speed is realized, the processing of the call request is in an intelligent adjustment state, dynamic intelligent adjustment can be performed according to actual data, the efficient processing of the actual request data is improved, the corresponding correction coefficients are represented by using the preset first standard length l10, the preset second standard length l20, and the preset third standard length l30, the forward relationship between the standard length and the correction coefficients is realized, and when the standard length is large, the correction coefficients are also large, which facilitates effective processing of the request data, and improves the data processing efficiency.
Specifically, with reference to fig. 1, the AI-based intelligent call cloud system according to the embodiment of the present invention further includes a reply module 50, where the reply module is connected to the display module, and is configured to reply to the user within a reply duration according to the content displayed by the display module, and the reply duration is set to T0.
Specifically, the embodiment of the invention answers the content displayed by the display module to the user within the answering time by arranging the answering module, completes the answering based on the user request information, improves the matching degree of the user calling and answering, and improves the answering efficiency of the user calling. The response time length T0 is adopted under the universality, the response time length is used as the reference time for responding to the user, the response time length is limited, the output speed of the display content of the display module is improved, the accurate control on the call response is improved, and the response efficiency is improved.
Specifically, a plurality of keywords are preset in the reply module, and the keywords are used for representing the professionalism in the display content;
the reply module is used for adjusting the response time length according to the number of the keywords.
Specifically, the embodiment of the invention presets a plurality of keywords in the reply module, if more keywords are in the display content, the display content has higher professional degree, if the number of the keywords is less, the display content has low professional degree, different professional contents are processed by adopting different response time lengths, and when the display content with high professional degree needs to be responded by adopting longer time, so that a user can fully explain the display content and receive the display content conveniently; if the displayed content is not high in professional degree, the content can be output in a short time, so that the difficulty degree of the displayed content is effectively adjusted, the response time is effectively and reasonably distributed, and the response content output efficiency is improved.
Specifically, a first adjustment parameter α and a second adjustment parameter β are set in the reply module, and a standard number n0 is also set in advance;
if the number of the keywords in the display content is less than or equal to the standard number n0, indicating that the professional degree of the display content is not high, and adjusting the response time length by adopting a first adjustment parameter alpha;
if the number of keywords in the display content is greater than the standard number n0, which indicates that the professional degree of the display content is high, the response time length is adjusted by using a second adjustment parameter β.
Specifically, after the response time length is adjusted by using the first adjustment parameter α, a new response time length is used as the response time length in the next call response period, and the new response time length is T1' = T0 × (1+ α);
and after the response time length is adjusted by adopting a second adjustment parameter beta, taking a new response time length as the response time length in the next call response period, wherein the new response time length is T2' = T0 x (1+ beta), and 0< alpha < beta < 1.
Specifically, the standard number n0 is set, and the first adjustment parameter α or the second adjustment parameter β is selected for the relationship between the number of the keywords in the display content and the standard number n0, so that the length of the response time better meets the actual content requirement.
Specifically, the intelligent cleaning and calling center based on the telephone traffic robot comprises a communication network, an artificial intelligent server, an AI seat for executing incoming and outgoing tasks, the artificial seat, an intelligent quality inspection management system for executing quality inspection tasks and an intelligent management background CRM for executing data mining and analysis, wherein the communication network is connected with a user terminal and the artificial intelligent server, the artificial intelligent server is used for accessing data in the communication network, the AI seat is used for executing outgoing and incoming telephone tasks, the artificial seat is used for executing a call request which cannot be processed by the AI seat, the intelligent quality inspection management system comprises a quality inspection rule management module and a quality inspection result background management module, and the intelligent management CRM is used as a carrier and a processing analyzer of call data.
A traffic robot processing module is built in an original customer service system to replace the original manual processing function based on the intelligent cleaning and calling technology of the traffic robot, and the traffic robot processing module is introduced into the overall framework of the customer service system. Corresponding technical modules are used in the overall customer service architecture to replace the original functions in the traditional architecture, and the traditional process is nested. In a new system architecture, a voice recognition and voice synthesis module is used for replacing the original manual communication mode with a user; and a knowledge management module is used for replacing the original knowledge base construction mode. The artificial intelligence technology module is used for replacing the original work which depends on manual processing and has certain repeatability and regularity, the original labor cost is reduced, and the efficiency and the effect are improved.
The intelligent cleaning technology based on the telephone traffic robot is used for cleaning mass data in a first round, manual calling and screening are replaced, parts of blank numbers, shutdown and incapability of contact are screened out, calls are automatically dialed, a sales and customer service system for screening the intended customers is carried out, an outbound task is established by one key, customer numbers are led in, timed outbound is called, the tasks are completed in batches, the optimized telephone operation data are highly customized, one-customer-one-operation is carried out, service data are fed back quickly according to the telephone data, manual interruption is carried out in the supporting process, feedback information is collected, quick feedback is carried out through the client question and answer process through semantic recognition, and the intentions of the customers can be recognized quickly according to the communication conditions with the customers.
The artificial intelligence technology is adopted to solve the problem of repeated and tedious customer service business with high labor cost of a company, certain manpower can be replaced in scenes such as online consultation, seat assistance, marketing promotion, internal management and the like through AI customer service calling, related businesses are matched with high precision by utilizing different engines through intelligent analysis of user query, and the accuracy of dialogue reply is ensured through intelligent session central control. On one hand, 80% of manual work is borne, on the other hand, customer demand analysis, demand matching, automatic statistics, classification and priority arrangement are automatically completed, and language identification, context and emotion analysis, reporting and comparison capabilities and the like are improved by combining big data and deep learning. Ensuring that a service scene 7 x 24 serves, and simultaneously controlling the manual work rate to be below 20%; the system realizes the butt joint with a business system, and truly forms a bridge for communication consultation and business handling between a company and a client aiming at the output of a mobile terminal or a PC terminal.
The natural language understanding and dialogue management combined model based on deep reinforcement learning consists of three cascaded deep neural networks, two cyclic neural networks at the bottom respectively model a current dialogue sentence and dialogue expression until the current moment from bottom to top, and the deep neural network at the upper layer makes a dialogue action decision; the natural language text is directly mapped to the dialogue action, and natural language processing and dialogue management are fused. The model adopts a deep reinforcement learning technology for integrated training, natural language utilizes the deep learning technology, and text is directly used as the input of the model, so that a large number of states are avoided being defined manually, and the problem of state space index increase is solved; the method has the advantages that the method carries out combined modeling and learning by using a deep reinforcement learning mode, solves the problem of independence between modules, enables information between the two modules to be shared, enables errors between the two modules to be transmitted, and obtains better performance.
Aiming at different conversation contents of a user, named entities in the conversation contents are extracted by using keywords or key sentence patterns, matched into a conversational model, and knowledge information is acquired by using a deep learning neural network to generate corresponding response contents, so that the man-machine conversation interactivity is realized, a machine can interact with the user more naturally, and the conversation interactivity and the user experience are improved.
So far, the technical solutions of the present invention have been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of the present invention is obviously not limited to these specific embodiments. Equivalent changes or substitutions of related technical features can be made by those skilled in the art without departing from the principle of the invention, and the technical scheme after the changes or substitutions can fall into the protection scope of the invention.
The intelligent matching algorithm matches with the user requirements and scenes through personalized content. In order to realize an intelligent accurate matching function based on the user personalized demand information, a user personalized demand information data model is constructed; and (4) establishing an intelligent accurate matching model base, and establishing an intelligent matching model between the user personalized demand information and the model base. The intelligent matching process is divided into four stages of model library retrieval based on DFS, matching based on semantic similarity, parameter matching and optimization and adjustment. And realizing accurate retrieval of the subclass model library according to the requirement information of the user, and returning the retrieved matched node information of the subclass model library to the spectacle frame type matching link. The algorithm has the advantages that the complexity of the organization structure of the glasses model library is not required to be considered, and the accurate and efficient retrieval of the sub-class model library nodes is realized; the method and the device are convenient for recommending proper push content for the user, improve the accuracy and satisfaction of user screening, reduce the searching time and energy and have the characteristic of intelligence.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention; various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. An AI-based intelligent calling cloud system, comprising:
the receiving module receives a call request of a user;
the analysis module is connected with the receiving module and used for analyzing the call request to obtain an analysis result;
the storage module analyzes the corresponding relation between the result and the solution;
the display module is used for acquiring a solution according to the analysis result and the corresponding relation and displaying the solution;
the analysis module comprises a data extraction unit, a data storage unit and a data processing unit;
the data extraction unit is used for extracting the call request, removing impurity information in the call request, acquiring request information and detecting the actual data volume of the request information;
a standard data volume D0 and an actual data volume D of the request information are preset in the data storage unit;
if the actual data volume D is larger than or equal to the standard data volume D0, performing secondary extraction on the call request;
if the actual data volume D is less than the standard data volume D0, indicating that the data volume is normal, and storing the request information;
the data processing unit is used for determining the length of a processing window for the request information according to the actual data volume when secondary extraction is carried out;
the data processing unit is internally preset with a first standard length l10, a second standard length l20 and a third standard length l30 of a processing window, and l10> l20> l30> 0;
if the request information is voice information, acquiring voiceprint information in the voice information, determining the actual age of the user according to the voiceprint information, and selecting a first standard length l10, a second standard length l20 or a third standard length l30 according to the actual age of the user;
a standard age range A12 is preset in the data processing unit, and if the actual age of the user is smaller than the minimum value A1 in the standard age range A12, a first standard length l10 is selected as the length of the processing window;
selecting a second standard length l20 as the length of the processing window if the actual age of the user belongs to the standard age group range a 12;
if the actual age of the user is greater than the maximum value a2 in the standard age group range a12, then the third standard length l30 is selected as the length of the processing window;
when the actual data volume D is larger than or equal to the standard data volume D0, correcting the standard length according to the amplitude of the deviation of the actual data volume from the standard data volume;
when the ith standard length li0 is selected as the length of the processing window, if the actual data volume D is more than or equal to the standard data volume D0 by 2 × D0>, the ith standard length li0 is corrected by adopting a first correction coefficient k 1;
if the actual data volume D is more than or equal to 5 multiplied by D0 and is more than or equal to 2 multiplied by the standard data volume D0, correcting the ith standard length li0 by adopting a second correction coefficient k 2;
if the actual data amount D is equal to or larger than 5 × D0, the i-th standard length li0 is corrected by using the third correction coefficient k3, and the first correction coefficient k1< the second correction coefficient k2< the third correction coefficient k 3.
2. The AI-based smart call cloud system of claim 1,
when the i-th standard length li0 is corrected by using the first correction coefficient k1, the corrected standard length li 0' = li0 × (1+ k 1);
when the i-th standard length li0 is corrected by using the second correction coefficient k2, the corrected standard length li 0' = li0 × (1+ k 2);
when the i-th standard length li0 is corrected by using the third correction coefficient k3, the corrected standard length li 0' = li0 × (1+ k 3), i =1, 2, and 3.
3. The AI-based smart call cloud system of claim 2,
the first correction coefficient k1= l10/(l10+ l20+ l30);
the second correction coefficient k2= l20/(l10+ l20+ l30);
the third correction coefficient k3= l30/(l10+ l20+ l 30).
4. The AI-based smart call cloud system of claim 3, further comprising a reply module, connected to the presentation module, for replying to the user within a reply duration according to the content presented by the presentation module, the reply duration being set to T0.
5. The AI-based smart call cloud system of claim 4, wherein a plurality of keywords are preset in the reply module, the keywords being used to characterize the expertise in the displayed content;
the reply module is used for adjusting the response time length according to the number of the keywords.
6. The AI-based smart call cloud system of claim 5,
the reply module is internally provided with a first adjusting parameter alpha and a second adjusting parameter beta, and is also pre-provided with a standard number n 0;
if the number of the keywords in the display content is less than or equal to the standard number n0, indicating that the professional degree of the display content is not high, and adjusting the response time length by adopting a first adjustment parameter alpha;
if the number of keywords in the display content is greater than the standard number n0, which indicates that the professional degree of the display content is high, the response time length is adjusted by using a second adjustment parameter β.
7. The AI-based smart call cloud system of claim 6, wherein after the first adjustment parameter α is adopted to adjust the response duration, a new response duration is taken as a response duration in a next call response period, and the new response duration is T1' = T0 × (1+ α);
and after the response time length is adjusted by adopting a second adjustment parameter beta, taking a new response time length as the response time length in the next call response period, wherein the new response time length is T2' = T0 x (1+ beta), and 0< alpha < beta < 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210279643.9A CN114363466B (en) | 2022-03-22 | 2022-03-22 | Intelligent cloud calling system based on AI |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210279643.9A CN114363466B (en) | 2022-03-22 | 2022-03-22 | Intelligent cloud calling system based on AI |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114363466A CN114363466A (en) | 2022-04-15 |
| CN114363466B true CN114363466B (en) | 2022-06-10 |
Family
ID=81094464
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210279643.9A Active CN114363466B (en) | 2022-03-22 | 2022-03-22 | Intelligent cloud calling system based on AI |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114363466B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111639484A (en) * | 2020-05-15 | 2020-09-08 | 北京青牛技术股份有限公司 | Method for analyzing seat call content |
| CN112417885A (en) * | 2020-11-17 | 2021-02-26 | 平安科技（深圳）有限公司 | Answer generation method and device based on artificial intelligence, computer equipment and medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101944359B (en) * | 2010-07-23 | 2012-04-25 | 杭州网豆数字技术有限公司 | Voice recognition method facing specific crowd |
| US9521258B2 (en) * | 2012-11-21 | 2016-12-13 | Castel Communications, LLC | Real-time call center call monitoring and analysis |
| CN107832291B (en) * | 2017-10-26 | 2020-03-31 | 平安科技（深圳）有限公司 | Man-machine cooperation customer service method, electronic device and storage medium |
| CN111324392A (en) * | 2018-12-13 | 2020-06-23 | 北京京东尚科信息技术有限公司 | Method and device for automatically adjusting data processing time window |
| CN112560501B (en) * | 2020-12-25 | 2022-02-25 | 北京百度网讯科技有限公司 | Semantic feature generation method, model training method, device, equipment and medium |
| CN113138982B (en) * | 2021-05-25 | 2022-09-27 | 深圳市元宇宙科技有限公司 | Big data cleaning method |
| CN113851111A (en) * | 2021-09-13 | 2021-12-28 | 联想(北京)有限公司 | Voice recognition method and voice recognition device |
-
2022
- 2022-03-22 CN CN202210279643.9A patent/CN114363466B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111639484A (en) * | 2020-05-15 | 2020-09-08 | 北京青牛技术股份有限公司 | Method for analyzing seat call content |
| CN112417885A (en) * | 2020-11-17 | 2021-02-26 | 平安科技（深圳）有限公司 | Answer generation method and device based on artificial intelligence, computer equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114363466A (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10536579B2 (en) | System, method and marketplace for real-time interactive video/voice services using artificial intelligence | |
| US11553055B2 (en) | Automated communication-based intelligence engine | |
| US7865457B2 (en) | Knowledge management system automatically allocating expert resources | |
| US10055501B2 (en) | Web-based customer service interface | |
| US11676093B2 (en) | Inferring missing customer data in assigning a ticket to a customer, and preventing reopening of the ticket in response of determining trivial data | |
| US6434549B1 (en) | Network-based, human-mediated exchange of information | |
| US9118763B1 (en) | Real time feedback proxy | |
| CA3143020A1 (en) | Systems and methods for communication system intent analysis | |
| US20210350384A1 (en) | Assistance for customer service agents | |
| US20160217500A1 (en) | Systems and methods for management of automated dynamic messaging | |
| EP3899819B1 (en) | System and method of real-time wiki knowledge resources | |
| KR20170137419A (en) | Method, system and computer-readable recording medium for providing customer counseling service using real-time response message generation | |
| KR102163081B1 (en) | Interactive voice bot server and unmanned counsel system | |
| TWI690811B (en) | Intelligent Online Customer Service Convergence Core System | |
| US11778093B1 (en) | Apparatuses and methods involving an integrated contact center | |
| KR20110048675A (en) | Call center counsel method and counsel system using voice recognition and tagging | |
| CN114363466B (en) | Intelligent cloud calling system based on AI | |
| US20230342864A1 (en) | System and method for automatically responding to negative content | |
| CN116208709A (en) | Voice outbound method, device, electronic equipment and storage medium | |
| JP2009295061A (en) | Priority control device | |
| CN116431779B (en) | FAQ question-answering matching method and device in legal field, storage medium and electronic device | |
| US20230186897A1 (en) | Searching calls based on contextual similarity among calls | |
| CN113627560A (en) | Feature analysis method, device and medium based on continuous interaction | |
| Thawani et al. | Web-based context aware information retrieval in contact centers | |
| CA2973596A1 (en) | Systems and methods for management of automated dynamic messaging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |