Hotline complaint-to-do collaborative application method and system

Abstract

The invention belongs to the technical field of online government affair service platforms, and relates to a hotline complaint-to-do collaborative application method and a hotline complaint-to-do collaborative application system. According to the method, the public appeal work order is obtained through multiple channels, the key information in the public appeal work order is extracted, and the key information is utilized to form the input work order; according to the TF-IDF algorithm, the input work orders are preliminarily classified, and through combination with the TF-IDF algorithm, accurate classification of the input work orders is facilitated; performing secondary classification on the preliminarily classified input work orders according to an appeal classification model; according to the mapping formula and the conditional probability extraction formula, the input work orders after secondary classification are automatically dispatched, the mapping formula and the conditional probability extraction formula can be automatically matched to the most appropriate processing department or personnel, and the order dispatching efficiency and accuracy are improved. According to the invention, accurate classification and automatic distribution of the crowd appeal work orders can be realized, and the solving efficiency of the crowd appeals is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of online service platforms, and relates to a hotline complaint-handling collaborative application method and system. Background Art

[0002] At present, hotline staff are usually used to classify work orders. Due to the huge number of work orders received every day, hotline staff may encounter many difficulties in classification. On the one hand, the content of work orders is varied and covers a wide range of fields, from urban management, environmental protection to social security, medical care, etc. Manual classification is affected by subjective factors, resulting in inaccurate classification of work orders. On the other hand, due to the huge workload, staff may become fatigued and negligent under high-pressure environment, resulting in deviations in the classification of work orders, and then inaccurate classification of work orders, resulting in some requests not being correctly assigned to the corresponding processing departments, increasing the difficulty and time of processing.

[0003] The purpose of the present invention is to provide a hotline complaint-handling collaborative application method and system to solve the technical problem of inaccurate work order classification. The present invention can achieve accurate classification and automatic distribution of work orders for public demands, thereby improving the efficiency of solving public demands. Summary of the invention

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a hotline complaint handling collaborative application method, characterized in that it includes the following steps: Obtain public appeal work orders through multiple channels, extract key information from the public appeal work orders, and use the key information to form input work orders; Perform preliminary classification of input work orders based on the TF-IDF algorithm; Perform secondary classification on the input work orders that have been initially classified according to the demand classification model; Automatically dispatch the input work orders after secondary classification.

[0007] Furthermore, the multi-channel acquisition of mass appeal work orders, extraction of key information in the mass appeal work orders, and use of the key information to form input work orders are as follows: For the public appeal work orders obtained through the telephone channel, the user appeal is responded to through the intelligent question and answer module. If the public appeal is not resolved by the question and answer module, key information is extracted from the appeal work order and used to form an input work order; For the appeal information obtained through online network channels and media channels, data verification is performed using the keywords, regular expressions, and URL security verification rules maintained by the system. If the requirements are met after verification, an entry work order is generated. If the requirements are not met after verification, key information in the appeal information is extracted and an entry work order is generated using the key information. The key information includes service channels, types of demands, public sentiment, location of the problem, time of incident, location of incident, subject of the demand and expectations of the demand.

[0008] Furthermore, for the mass appeal work orders obtained through the telephone channel, the user appeals are responded to through the intelligent question and answer module, as follows: For public appeal work orders obtained through telephone channels, the intelligent question-and-answer module is entered through intelligent voice to understand the public's intentions and automatically respond to user questions; If the intelligent question-and-answer module cannot solve the public's demands, the call will be answered manually, and the IVR navigation will be used to determine whether the call is busy. If the line is not busy, reduce the time of intelligent voice or IVR navigation; If the line is occupied, increase the navigation time and guide the customer to accept the work order through the intelligent navigation online channel.

[0009] Furthermore, the input work orders are initially classified according to the TF-IDF algorithm, as follows: Perform preprocessing operations on the input work order to remove noise information and interference data to obtain data documents; Split the data document into word units through Jieba word segmentation; The TF-IDF algorithm is used to extract the features of word units in the data document to obtain word features; Determine the nature and field of the work order to be entered based on word features; Perform preliminary classification of incoming work orders according to their nature, field and urgency; The nature of the work order includes: consultation, complaint, request for help, report, suggestion; The areas of the work orders include: real estate, heating, water supply, gas supply, environmental protection, transportation, education, civil affairs, urban management, demolition and market.

[0010] Furthermore, the TF-IDF algorithm is used to extract features from word units to obtain word features, which are as follows: Obtaining the frequency of occurrence of the word unit in the data document according to the number of times the word unit appears in the data document and the total number of word units in the data document; Obtain inverse document frequency based on the total number of data documents, word units, and data documents; The features of the word units in the document are extracted according to the frequency of occurrence of the word units in the data document and the inverse document frequency to obtain the word features.

[0011] Furthermore, the frequency of occurrence of the word unit in the data document is obtained according to the number of occurrences of the word unit in the data document and the total number of word units in the data document, as follows:

[0012] in Word unit In the data document The frequency of occurrence in Word unit In the data document The number of times it appears in Representation data document Total number of word units; The inverse document frequency is obtained according to the total number of data documents, word units and data documents, as follows:

[0013] in Representation data document Does it contain word units? , if included, it is 1, if not included, it is 0; Indicates the total number of data documents; Represents inverse document frequency.

[0014] Furthermore, the demand classification model is as follows: Obtain the completed historical public demand work order data, push the completed historical public demand work order data to the demand classification model, and generate a corresponding linked list based on the prediction results of the demand classification model; According to the linked list, classification information related to the linked list index is obtained; Merge the classified information with the completed historical public request work order data to generate enhanced category information; The demand classification model is trained using the information of the enhanced category to enhance the classification ability of the demand classification model.

[0015] Furthermore, the input work orders after secondary classification are automatically dispatched according to the mapping formula and the conditional probability extraction formula, as follows: Use the mapping formula to map each word in the input work order after secondary classification into a high-dimensional vector space to obtain the word vector; The geometric relationship between word vectors and the conditional probability extraction formula are used to extract and recommend the corresponding dispatch unit information, and the corresponding input work order is dispatched to the corresponding unit according to the dispatch unit information.

[0016] Furthermore, the mapping formula is as follows:

[0017] in, is a word vector, For the The word vector component of each word, is the dimension of the vector space, Take a positive integer, is the transpose operation, A technique for mapping each word in the input work order after secondary classification into a vector space; The conditional probability extraction formula is as follows:

[0018] in, is probability; For event characteristics; For preference settings; For distribution units , For distribution units ; is a known event feature; Characteristics of the dispatch unit.

[0019] In a second aspect, the present invention provides a hotline complaint handling collaborative application system, characterized in that it includes: Demand work order acquisition module: used to obtain public demand work orders through multiple channels, extract key information from public demand work orders, and use the key information to form input work orders; Preliminary classification module: used to perform preliminary classification of input work orders based on the TF-IDF algorithm; Secondary classification module: used to perform secondary classification on the input work orders that have been initially classified according to the demand classification model; Classification dispatching module: used to automatically dispatch input work orders after secondary classification based on mapping formula and conditional probability extraction formula.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention obtains the work orders of public demands through multiple channels, so as to cover more public demands, extract key information from the work orders of public demands, help to quickly understand the core content of demands, use key information to form input work orders, ensure the integrity and accuracy of information, and provide a basis for the classification and processing of demands; the input work orders are preliminarily classified according to the TF-IDF algorithm, and the combination of the TF-IDF algorithm is conducive to the accurate classification of the input work orders; the input work orders preliminarily classified are secondary classified according to the demand classification model, and the demand classification model can be updated and learned based on historical data, and can classify the input work orders more accurately and meticulously; the input work orders after secondary classification are automatically dispatched according to the mapping formula and the conditional probability extraction formula, and the mapping formula and the conditional probability extraction formula can automatically match the most suitable processing department or personnel, thereby improving the efficiency and accuracy of dispatching. The present invention can realize the accurate classification and automatic dispatch of the work orders of public demands, and improve the efficiency of solving public demands.

[0021] 2. The present invention performs preprocessing operations on the input work order, removes noise information and interference data, obtains data documents, and is conducive to the accuracy of work order classification. The data document is split into word units through Jieba word segmentation, which is conducive to refinement to the word level and makes feature extraction more accurate; the TF-IDF algorithm is used to extract the features of the word units in the data document to obtain word features; the nature and field of the input work order are judged according to the word features, which is conducive to adapting to the work order classification needs in different fields.

[0022] 3. The present invention obtains completed historical public demand work order data, pushes the completed historical public demand work order data to the demand classification model, and then trains and updates the demand classification model, which is beneficial to improving the accuracy and efficiency of demand classification.

[0023] 4. The system of the present invention includes a demand work order acquisition module, a preliminary classification module, a secondary classification module and a classification dispatching module. The demand work order acquisition module is used to obtain the public demand work orders through multiple channels, extract the key information in the public demand work orders, and use the key information to form input work orders; the preliminary classification module is used to perform preliminary classification of the input work orders according to the TF-IDF algorithm; the secondary classification module is used to perform secondary classification of the preliminary classified input work orders according to the demand classification model; the classification dispatching module is used to automatically dispatch the input work orders after secondary classification according to the mapping formula and the conditional probability extraction formula. The various modules cooperate with each other to realize the accurate classification and dispatch of the public demand work orders, avoid the problem of unclear responsibilities and shirking of responsibilities during the work order processing, and improve the efficiency of resolving public demands. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flowchart of a method according to an embodiment of the present invention; Figure 2A structural diagram of obtaining mass demands in an embodiment of the present invention; Figure 3 is a flow chart of the method of the present invention; Figure 4 It is a system connection diagram of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 3 The present invention discloses a hotline complaint handling collaborative application method, comprising the following steps: S1, obtaining public appeal work orders through multiple channels, so as to cover more public needs, extract key information from public appeal work orders, help to quickly understand the core content of the appeal, and use key information to form input work orders, ensuring the integrity and accuracy of information, and providing a basis for appeal classification and processing; S2, preliminarily classifies the input work orders according to the TF-IDF algorithm. By combining the TF-IDF algorithm, it is conducive to the accurate classification of the input work orders; S3, perform secondary classification on the input work orders that have been initially classified according to the demand classification model. The demand classification model can be updated and learned based on historical data, and can classify the input work orders more accurately and meticulously. By continuously training and optimizing the demand classification model, the accuracy and stability of classification can be continuously improved; S4, automatically dispatch the input work orders after secondary classification according to the mapping formula and conditional probability extraction formula. The mapping formula and conditional probability extraction formula can automatically match the most appropriate processing department or personnel to improve the efficiency and accuracy of dispatching.

[0028] The present invention can realize the accurate classification and automatic dispatching of the work orders for the public's demands, thus improving the efficiency of solving the public's demands.

[0029] Embodiment 1: See also Figure 3 This embodiment discloses a hotline complaint handling collaborative application method, including the following steps: S1, obtain the public appeal work order through multiple channels, extract the key information from the public appeal work order, and use the key information to form the input work order, as follows: For the public appeal work orders obtained through the telephone channel, the user appeal is responded to through the intelligent question and answer module. If the public appeal is not resolved by the question and answer module, key information is extracted from the appeal work order and used to form an input work order; For the appeal information obtained through online network channels and media channels, data verification is performed using the keywords, regular expressions, and URL security verification rules maintained by the system. If the requirements are met after verification, an entry work order is generated. If the requirements are not met after verification, key information in the appeal information is extracted and an entry work order is generated using the key information. The key information includes service channels, types of demands, public sentiment, location of the problem, time of incident, location of incident, subject of the demand and expectations of the demand.

[0030] Preferably, for the mass appeal work order obtained through the telephone channel, the user appeal is replied through the intelligent question and answer module, as follows: For public appeal work orders obtained through telephone channels, the intelligent question-and-answer module is entered through intelligent voice to understand the public's intentions and automatically respond to user questions; If the intelligent question-and-answer module cannot solve the public's demands, the call will be answered manually, and the IVR navigation will be used to determine whether the call is busy. If the line is not busy, reduce the time of intelligent voice or IVR navigation; If the line is occupied, increase the navigation time and guide the customer to accept the work order through the intelligent navigation online channel.

[0031] S2, preliminarily classifies the input work orders according to the TF-IDF algorithm, as follows: Perform preprocessing operations on the input work orders to remove noise information and interference data, which is beneficial to the accuracy of work order classification and obtain data documents; The data document is split into word units through Jieba word segmentation, which is conducive to refining to the word level and making feature extraction more accurate; The TF-IDF algorithm is used to extract the features of word units in the data document to obtain word features, as follows: Obtaining the frequency of occurrence of the word unit in the data document according to the number of times the word unit appears in the data document and the total number of word units in the data document; Obtain inverse document frequency based on the total number of data documents, word units, and data documents; The features of the word units in the document are extracted according to the frequency of occurrence of the word units in the data document and the inverse document frequency to obtain the word features.

[0032] Judging the nature and field of the work order based on word features is helpful to meet the needs of work order classification in different fields; Perform preliminary classification of incoming work orders according to their nature, field and urgency; The nature of the work order includes: consultation, complaint, request for help, report, suggestion; The areas of the work orders include: real estate, heating, water supply, gas supply, environmental protection, transportation, education, civil affairs, urban management, demolition and market.

[0033] Preferably, the frequency of occurrence of the word unit in the data document is obtained according to the number of occurrences of the word unit in the data document and the total number of word units in the data document, as follows:

[0034] in Word unit In the data document The frequency of occurrence in Word unit In the data document The number of times it appears in Representation data document Total number of word units; The inverse document frequency is obtained according to the total number of data documents, word units and data documents, as follows:

[0035] in Representation data document Does it contain word units? , if included, it is 1, if not included, it is 0; Indicates the total number of data documents; Represents inverse document frequency.

[0036] S3, secondary classification of the input work orders that have been initially classified according to the demand classification model; Preferably, the demand classification model is as follows: Obtain the completed historical public demand work order data, push the completed historical public demand work order data to the demand classification model, and generate a corresponding linked list based on the prediction results of the demand classification model; According to the linked list, classification information related to the linked list index is obtained; Merge the classified information with the completed historical public request work order data to generate enhanced category information; The demand classification model is trained using the information of the enhanced category to enhance the classification ability of the demand classification model.

[0037] S4, automatically dispatch the input work orders after secondary classification according to the mapping formula and conditional probability extraction formula, as follows: Use the mapping formula to map each word in the input work order after secondary classification into a high-dimensional vector space to obtain the word vector; Preferably, the mapping formula is as follows:

[0038] in, is a word vector, For the The word vector component of each word, is the dimension of the vector space, Take a positive integer, is the transpose operation, A technique for mapping each word in the input work order after secondary classification into a vector space; The geometric relationship between word vectors and the conditional probability extraction formula are used to extract and recommend the corresponding dispatch unit information, and the corresponding input work order is dispatched to the corresponding unit according to the dispatch unit information.

[0039] Preferably, the conditional probability extraction formula is as follows:

[0040] in, is probability; For event characteristics; For preference settings; For distribution units , For distribution units ; is a known event feature; is the characteristic of the dispatch unit; the dispatch unit is compared by comparing the event characteristics and dispatch units The probability of selecting the corresponding dispatch unit.

[0041] S5, set a reasonable processing time limit for the work order after dispatch and conduct full supervision and inspection, as follows: Set a processing time limit for the assigned work order, starting from the creation time of the work order. After the time starts, the verification procedure will be initiated within the set time. If the assigned work order does not fall within the scope of responsibility, the work order will be returned with the reason, basis and transfer suggestion for the return. If the assigned work order falls within the scope of responsibility, it will continue to be processed; Mark the expected overdue work orders in yellow and the overdue work orders in red, and automatically send out warnings and overtime reminders according to the set time; For urgent, difficult, dangerous, serious and abnormal work orders, automatic warning reminders will be issued and reported to the superior sub-center. According to the superior sub-center's supervision instructions on urgent, difficult, dangerous, serious and abnormal work orders, the work orders will be included in the supervision process and reassigned to other commissioned units for further processing.

[0042] Preferably, after the order is dispatched, if specific requirements for auxiliary processing are received from the person handling the work order, the auxiliary material information will be pushed to the processing department.

[0043] Embodiment 2: See also Figure 1 This embodiment forms a complete and efficient hotline complaint handling collaborative application system by clarifying each link and step. Through comprehensive acceptance, classified disposal, accurate dispatch, and time-limited handling, it ensures that the people's demands are handled in a timely and effective manner; at the same time, through supervision, inspection, follow-up evaluation and assessment mechanisms, it guarantees the quality and efficiency of the handling work. The establishment and implementation of this system will greatly improve the overall level and service quality of the hotline complaint handling work, which specifically includes the following steps: S1. See Figure 2 , comprehensively accept demands through multiple channels: integrate multiple channels such as telephone, Internet and media to achieve unified access and processing of people's demands.

[0044] Users can submit feedback through any channel and ultimately receive and process it in a unified manner. After entering the telephone channel, the agent will answer the call and register it in the work order database. Through intelligent voice or IVR navigation, it will be adjusted to determine whether the call is busy. If the line is not busy, the intelligent voice or IVR navigation time will be appropriately reduced. If the line is busy, the navigation time will be increased, guiding citizens to accept work orders through online channels such as intelligent navigation or WeChat. Online network and media channel users fill in their own demand information and form a demand work order based on the demand information. After the system obtains the demand information through a unified interface, it will enter the demand information into the rule verification for data verification. The verified data will be entered into the work order and written into the work order task pool; The system will automatically complete the hot spots, assigned departments, involved units and other information in the work order according to the agent's shift schedule and write it into the work order database.

[0045] S2. Classification and disposal: Based on the consultation, complaint, suggestion, help and report information and the degree of urgency in the work order information of the accepted request, the TF-IDF algorithm is used to identify the word features in the entered work order, and the nature and field of the entered work order are preliminarily determined. The entered work order is filtered to the corresponding function menu to achieve preliminary classification of the entered work order.

[0046] S3. Precision dispatch system: formulate classification and handling standards in the system, establish special management records for unreasonable demands, demands with unclear rights and responsibilities, demands for buck-passing, and corporate demands, and implement classification and handling together with regular demands to achieve refined classification of demands. Through the dispatch information of historical demands, the demand classification model is trained, and the input work orders that have been initially classified are reclassified according to the demand classification model; The TF-IDF algorithm, the nature of the input work orders and the fields of the input work orders are used to preliminarily classify the input work orders to realize the intelligent dispatching function.

[0047] A special management ledger will be established and dispatched for unreasonable demands, demands with unclear rights and responsibilities, demands that are based on buck-passing and corporate demands.

[0048] Each word in the input work order after secondary classification is mapped to a high-dimensional vector space using a mapping formula to obtain a word vector. The mapping formula is as follows:

[0049] in, is a word vector, For the The word vector component of each word, is the dimension of the vector space, Take a positive integer, is the transpose operation, A technique for mapping each word in the input work order after secondary classification into a vector space; The geometric relationship between word vectors and the conditional probability extraction formula are used to extract and recommend the corresponding dispatch unit information, and the corresponding input work order is dispatched to the corresponding unit according to the dispatch unit information.

[0050] Preferably, the conditional probability extraction formula is as follows:

[0051] in, is probability; For event characteristics; For preference settings; For distribution units , For distribution units ; is a known event feature; is the characteristic of the dispatch unit; the dispatch unit is compared by comparing the event characteristics and dispatch units The probability of selecting the corresponding dispatch unit.

[0052] For work orders with clear street offices and matters, they will be directly assigned to the streets and towns within the scope of responsibility for processing, so as to achieve the purpose of direct assignment and immediate processing.

[0053] S4. Clarify the hierarchical coordination mechanism: Based on the nature, scope of impact, urgency and other factors of the work order problem, formulate clear grading standards to analyze the data and assist in the allocation of functions.

[0054] For example, problems can be divided into urgent, important and general levels, where urgent means the impact is significant and needs to be resolved immediately; important means a quick response is required but not an immediate resolution; and general means it can be handled through normal procedures.

[0055] Direct dispatch to grassroots level for processing rules: For general issues, they will be handled directly by front-line customer service or relevant grassroots departments. Customer service will directly reply and handle them or assign them to grassroots level for direct processing to quickly respond to customer needs.

[0056] Feedback on middle-level coordination rules: For important issues or problems that cannot be solved at the grassroots level, the work order will be transferred to the middle-level management for cross-departmental coordination based on the keywords in the classification to ensure effective integration of resources and rapid processing.

[0057] Reporting rules for high-level decision-making: For urgent or major risk issues, the work order will be transferred to high-level leadership for intervention based on the keywords in the classification, and response strategies will be formulated and the handling opinions will be directly notified to the handling department for processing.

[0058] S5. Time-limited processing: Set a reasonable time limit for processing requests after the request is assigned. Set the time limit corresponding to the consultation, complaint, suggestion, help and report categories and the degree of urgency. The time limit is calculated based on the creation time of the request work order, and the processing unit or personnel are required to complete the processing within the specified time; Start the verification procedure within 1 hour. For work orders that are not within the scope of responsibility, apply for return according to the procedure, and clearly state the reasons, basis and transfer suggestions for return; S6. First-contact person and assistance requirements: Specific requirements for assistance can be put forward to the person handling the claim during the handling process, such as providing necessary information, assisting in investigation and evidence collection, etc. The system will push the information to the handling department, and the auxiliary handling department will be more clearly responsible for handling the claim.

[0059] S7. Full-process supervision and inspection: Set up full-process supervision and assessment rules, mark the expected and overdue work orders in yellow and red, and the system will automatically issue warnings and overtime reminders according to the set time; for urgent, difficult, dangerous and abnormal work orders, the system will automatically issue warnings and report them to the superior sub-center. The work order process remains unchanged and is still handled by the original commissioned unit. The superior sub-center can issue supervision instructions, include the work order in the supervision process, and reassign it to other commissioned units for continued processing.

[0060] In summary, the beneficial effects of the present invention are: 1. Speed ​​up the receiving process: Adjustments are made through intelligent voice or IVR navigation to determine whether the call is busy. If the line is not busy, the intelligent voice or IVR navigation time is appropriately reduced. If the line is busy, the navigation time is increased, and citizens are guided to accept work orders through online channels such as intelligent navigation or WeChat.

[0061] 2. Speed ​​up the dispatch process: Through the latest artificial intelligence technology, we can realize intelligent classification, dispatching, processing and data analysis of demands, automatically recommend the organizing unit and complete automatic dispatch, saving manual dispatch time.

[0062] 3. Speed ​​up collaborative reporting: Establish a cross-departmental and cross-level collaborative processing mechanism. The first-level collaborative reporting work model is that the first-level department initiates the co-organization at the same level, and another department cooperates to handle the work order. The responsibility for the work order is borne by the initiating and co-organizing unit.

[0063] It should be noted that the IVR navigation, intelligent voice and jieba word segmentation are all conventional technologies. IVR navigation, namely interactive voice response (Interactive Voice Response), is a telephone navigation technology that allows computers to interact with users through voice recognition and key input. Intelligent voice, namely intelligent voice technology, is a technology for realizing human-computer language communication, including speech recognition technology (ASR) and speech synthesis technology (TTS). Intelligent voice technology enables machines to understand, recognize and respond to human voice commands, thereby providing a more convenient and efficient interactive experience. Among them, jieba word segmentation is a Chinese text segmentation tool, which is implemented based on Python and focuses on efficient and accurate word segmentation processing of Chinese text. Word segmentation is a basic step in natural language processing. It divides continuous text into independent word units to facilitate subsequent tasks such as text analysis, information retrieval, sentiment analysis, and machine translation.

[0064] Based on the above method, the present invention also discloses a hotline complaint handling collaborative application system, see Figure 4 ,include: Demand work order acquisition module: used to obtain public demand work orders through multiple channels, extract key information from public demand work orders, and use the key information to form input work orders; Preliminary classification module: used to perform preliminary classification of input work orders based on the TF-IDF algorithm; Secondary classification module: used to perform secondary classification on the input work orders that have been initially classified according to the demand classification model; Classification dispatching module: used to automatically dispatch input work orders after secondary classification based on mapping formula and conditional probability extraction formula.

[0065] The various modules of the system of the present invention cooperate with each other, which can realize the accurate classification and distribution of work orders for public demands, and improve the efficiency of solving public demands.

[0066] In summary, the beneficial effects of the present invention are as follows: The present invention can realize the accurate classification and distribution of work orders for mass demands, which is conducive to improving the response speed of resolving mass demands: through intelligent processing and omni-channel integration, the response speed of mass demands can be significantly improved.

[0067] The present invention is conducive to enhancing public satisfaction: quickly responding to and effectively resolving demands, and improving public satisfaction with the service.

[0068] The present invention is conducive to promoting the transformation of hotline complaint handling collaborative applications towards digitalization and intelligence.

[0069] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A hotline complaint handling collaborative application method, characterized in that: The following steps are involved: Obtain public appeal work orders through multiple channels, extract key information from the public appeal work orders, and use the key information to form input work orders; Perform preliminary classification of input work orders based on the TF-IDF algorithm; Perform secondary classification on the input work orders that have been initially classified according to the demand classification model; The input work orders after secondary classification are automatically dispatched according to the mapping formula and conditional probability extraction formula.

2. The hotline complaint handling collaborative application method according to claim 1 is characterized in that: The method of obtaining the public appeal work order through multiple channels, extracting key information from the public appeal work order, and using the key information to form an input work order is as follows: For the public appeal work orders obtained through the telephone channel, the user appeal is responded to through the intelligent question and answer module. If the public appeal is not resolved by the question and answer module, key information is extracted from the appeal work order and used to form an input work order; For the appeal information obtained through online network channels and media channels, data verification is performed using the keywords, regular expressions, and URL security verification rules maintained by the system. If the requirements are met after verification, an entry work order is generated. If the requirements are not met after verification, key information in the appeal information is extracted and an entry work order is generated using the key information. The key information includes service channels, types of demands, public sentiment, location of the problem, time of incident, location of incident, subject of the demand and expectations of the demand.

3. The hotline complaint handling collaborative application method according to claim 2 is characterized in that: For the mass appeal work orders obtained through the telephone channel, the user appeals are responded to through the intelligent question and answer module, as follows: For public appeal work orders obtained through telephone channels, the intelligent question-and-answer module is entered through intelligent voice to understand the public's intentions and automatically respond to user questions; If the intelligent question-and-answer module cannot solve the public's demands, the call will be answered manually, and the IVR navigation will be used to determine whether the call is busy. If the line is not busy, reduce the time of intelligent voice or IVR navigation; If the line is occupied, increase the navigation time and guide the customer to accept the work order through the intelligent navigation online channel.

4. The hotline complaint handling collaborative application method according to claim 1 is characterized in that: The input work orders are initially classified according to the TF-IDF algorithm, as follows: Perform preprocessing operations on the input work order to remove noise information and interference data to obtain data documents; Split the data document into word units through Jieba word segmentation; The TF-IDF algorithm is used to extract the features of word units in the data document to obtain word features; Determine the nature and field of the work order to be entered based on word features; Perform preliminary classification of incoming work orders according to their nature, field and urgency; The nature of the work order includes: consultation, complaint, request for help, report, suggestion; The areas of the work orders include: real estate, heating, water supply, gas supply, environmental protection, transportation, education, civil affairs, urban management, demolition and market.

5. The hotline complaint handling collaborative application method according to claim 1 is characterized in that: The TF-IDF algorithm is used to extract features from word units to obtain word features, which are as follows: Obtaining the frequency of occurrence of the word unit in the data document according to the number of times the word unit appears in the data document and the total number of word units in the data document; Obtain inverse document frequency based on the total number of data documents, word units, and data documents; The features of the word units in the document are extracted according to the frequency of occurrence of the word units in the data document and the inverse document frequency to obtain the word features.

6. The hotline complaint handling collaborative application method according to claim 1, characterized in that: The frequency of occurrence of the word unit in the data document is obtained according to the number of occurrences of the word unit in the data document and the total number of word units in the data document, as follows: in Word unit In the data document The frequency of occurrence in Word unit In the data document The number of times it appears in Representation data document Total number of word units; The inverse document frequency is obtained according to the total number of data documents, word units and data documents, as follows: in Representation data document Does it contain word units? , if included, it is 1, if not included, it is 0; Indicates the total number of data documents; Represents inverse document frequency.

7. The hotline complaint handling collaborative application method according to claim 1 is characterized in that: The demand classification model is as follows: Obtain the completed historical public demand work order data, push the completed historical public demand work order data to the demand classification model, and generate a corresponding linked list based on the prediction results of the demand classification model; According to the linked list, classification information related to the linked list index is obtained; Merge the classified information with the completed historical public request work order data to generate enhanced category information; The demand classification model is trained using the information of the enhanced category to enhance the classification ability of the demand classification model.

8. The hotline complaint handling collaborative application method according to claim 1, characterized in that: The automatic dispatching of the input work orders after secondary classification according to the mapping formula and the conditional probability extraction formula is as follows: Use the mapping formula to map each word in the input work order after secondary classification into a high-dimensional vector space to obtain the word vector; The geometric relationship between word vectors and the conditional probability extraction formula are used to extract and recommend the corresponding dispatch unit information, and the corresponding input work order is dispatched to the corresponding unit according to the dispatch unit information.

9. The hotline complaint handling collaborative application method according to claim 8, characterized in that: The mapping formula is as follows: in, is a word vector, For the The word vector component of each word, is the dimension of the vector space, Take a positive integer, is the transpose operation, A technique for mapping each word in the input work order after secondary classification into a vector space; The conditional probability extraction formula is as follows: in, is probability; For event characteristics; For preference settings; For distribution units , For distribution units ; is a known event feature; Characteristics of the dispatch unit.

10. A hotline complaint handling collaborative application system, characterized in that: include: Demand work order acquisition module: used to obtain public demand work orders through multiple channels, extract key information from public demand work orders, and use the key information to form input work orders; Preliminary classification module: used to perform preliminary classification of input work orders based on the TF-IDF algorithm; Secondary classification module: used to perform secondary classification on the input work orders that have been initially classified according to the demand classification model; Classification dispatching module: used to automatically dispatch input work orders after secondary classification based on mapping formula and conditional probability extraction formula.

Images