Real-time conversation based product component recommendation method, device, equipment and medium

Abstract

The present application relates to the field of intelligent financial decision-making, especially suitable for the recommendation scene of financial management and investment products, and discloses a product component recommendation method based on real-time conversation, comprising: acquiring real-time conversation and business scene to query the product card node tree corresponding to the real-time conversation; identifying the dialogue intention of the real-time conversation, the dialogue intention including consultation intention and reply intention; respectively querying the node card of the consultation intention and the reply intention in the product card node tree to obtain the consultation card and the reply card, and screening the first recommended card of the real-time conversation from the consultation card and the reply card; extracting the conversation keywords of the real-time conversation, querying the conversation recommended card of the real-time conversation from the product card node tree according to the conversation keywords, screening the conversation recommended card, and obtaining the second recommended card; and determining the final recommended card of the real-time conversation according to the first recommended card and the second recommended card. The present application can improve the accuracy of the component recommendation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent financial decision-making, and in particular to a product component recommendation method and device based on real-time conversation, an electronic device and a computer readable storage medium. BACKGROUND

[0002] Real-time conversation refers to an uninterrupted request-response sequence between a current customer and a server. A component is a simple encapsulation of data and methods. A component product is a product designed for a specific occasion that can be assembled from components to quickly implement system changes and adapt to new requirements, such as financial planning products, insurance products, and e-government affairs. Product component recommendation is the dynamic provision of associated knowledge of products, activities, and benefits as the conversation content progresses. Component products will show the recommendation results to customers in the form of component cards, and are based on a unified component framework to empower the conversation process with more rich and clear multimedia information.

[0003] Currently, traditional business components are mainly matched based on the intent in the customer's question, but as the types of customer intent continue to increase, the corresponding intent matching node may not be found in the intent matching node tree of the component product, resulting in insufficient recommendation ability for product components and low accuracy of component recommendation results. SUMMARY

[0004] The present application provides a product component recommendation method and device based on real-time conversation, an electronic device and a computer readable storage medium, which mainly aims to improve the accuracy of component recommendation results.

[0005] To achieve the above-mentioned purpose, the present application provides a product component recommendation method based on real-time conversation, comprising:

[0006] Obtaining a real-time conversation and its corresponding business scenario, querying a product card node tree corresponding to the real-time conversation according to the business scenario;

[0007] Identifying the conversation intent of the real-time conversation using a preset intent recognition model, the conversation intent including a consultation intent and a reply intent;

[0008] Respectively querying the node card of the consultation intent and the reply intent in the product card node tree to obtain a consultation card and a reply card, respectively calculating the card weight of the consultation card and the reply card, and taking the card with a card weight greater than a preset weight as the first recommended card of the real-time conversation;

[0009] According to the dialogue intention, a conversation keyword of the real-time conversation is extracted, a conversation recommendation card of the real-time conversation is queried from the product card node tree according to the conversation keyword, and the conversation recommendation card is filtered to obtain a second recommendation card;

[0010] According to the first recommendation card and the second recommendation card, a final recommendation card of the real-time conversation is determined.

[0011] Optionally, the creating of the product card node tree corresponding to the real-time conversation comprises:

[0012] According to the business scenario, a sales script of the real-time conversation under the business scenario is formulated;

[0013] According to the sales script, a flow node tree of the sales script is drawn, and the flow node tree comprises a start node, a consultation intention node, a reply intention node, and a judgment node.

[0014] The associated intention keywords and component cards corresponding to the consultation intention node and the reply intention node are configured to generate the product card node tree corresponding to the real-time conversation.

[0015] Optionally, the identifying of the dialogue intention of the real-time conversation by using the preset intention recognition model, the dialogue intention comprising a consultation intention and a reply intention, comprises:

[0016] The consultation conversation and the reply conversation in the real-time conversation are obtained.

[0017] Text feature words of the consultation conversation and the reply conversation are extracted, and a preset intention recognition model is used to calculate an intention category matched by the text feature words in the preset intention recognition model.

[0018] The matched intention category is taken as the dialogue intention of the real-time conversation, and the dialogue intention comprises a consultation intention and a reply intention.

[0019] Optionally, the calculating of the intention category matched by the text feature words in the preset intention recognition model by using the preset intention recognition model comprises:

[0020] An intention category text in the preset intention recognition model is identified, and a word segmentation tool is used to perform word segmentation on the intention category text to obtain an intention category segmentation;

[0021] A vector conversion model that has been trained is used to calculate word vectors of the text feature words and the intention category segmentation;

[0022] Calculate the cosine similarity matrix of the text feature words and the word vector of the intent classification segmentation, and take the intent classification corresponding to the maximum element in the cosine similarity matrix as the matched intent classification.

[0023] Optionally, the respective calculation of the consultation card and the reply card card weight includes:

[0024] Respectively query the child nodes of the consultation card and the reply card in the product card node tree to obtain the consultation candidate card and the reply candidate card corresponding to the child nodes;

[0025] Obtain the consultation session and the reply session corresponding to the consultation intent and the reply intent in the real-time session, calculate the similarity of the consultation candidate card and the consultation session and the similarity of the reply candidate card and the reply session, and take the similarity as the card weight of the consultation card and the reply card.

[0026] Optionally, the extraction of the session keyword of the real-time session according to the dialogue intent includes:

[0027] Obtain the session text of the real-time session, and extract the feature words of the session text;

[0028] According to the dialogue intent, the keywords with dialogue intent are screened out from the feature words.

[0029] Optionally, the query of the session recommendation card of the real-time session from the product card node tree according to the session keyword includes:

[0030] Obtain the intent keywords of the consultation intent node and the reply intent node in the product card node tree;

[0031] According to the session keyword, the intent keywords and the session keywords are respectively converted into intent word vectors and session word vectors;

[0032] Calculate the similarity of the intent word vector and the session word vector, and take the component card corresponding to the intent word vector with the similarity greater than a preset threshold as the session recommendation card of the real-time session and mark the corresponding session recommendation card similarity.

[0033] In order to solve the above problems, the application also provides a product component recommendation device based on real-time session, the device comprises:

[0034] A product card node tree query module is used to obtain a real-time session and a corresponding business scenario thereof, and query a product card node tree corresponding to the real-time session according to the business scenario;

[0035] The dialogue intention recognition module is configured to recognize a dialogue intention of the real-time conversation by using a preset intention recognition model, the dialogue intention including a consultation intention and a reply intention.

[0036] The first recommended card generation module is configured to query a node card of the consultation intention and the reply intention in the product card node tree respectively to obtain a consultation card and a reply card, calculate a card weight of the consultation card and the reply card respectively, and take a card with a card weight greater than a preset weight as a first recommended card of the real-time conversation.

[0037] The second recommended card generation module is configured to extract a conversation keyword of the real-time conversation according to the dialogue intention, query a conversation recommended card of the real-time conversation from the product card node tree according to the conversation keyword, and screen the conversation recommended card to obtain a second recommended card.

[0038] The final recommended card generation module is configured to determine a final recommended card of the real-time conversation according to the first recommended card and the second recommended card.

[0039] To solve the above problems, the present application further provides an electronic device, which comprises:

[0040] at least one processor; and

[0041] a memory connected with the at least one processor in communication; wherein

[0042] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to implement the product component recommendation method based on the real-time conversation.

[0043] To solve the above problems, the present application further provides a computer readable storage medium, which stores at least one computer program, and the at least one computer program is executed by a processor in an electronic device to implement the product component recommendation method based on the real-time conversation.

[0044] As can be seen, this embodiment of the invention can obtain the operation objects and application environment of the application scenario by acquiring the real-time session and its corresponding business scenario. According to the business scenario, the product card node tree corresponding to the real-time session is queried to provide support for obtaining matching node cards and thus obtaining recommended cards for product components. A preset intent recognition model is used to identify the dialogue intent of the real-time session, which includes consultation intent and response intent. The dialogue intent of both parties in the real-time dialogue can be obtained to recommend corresponding product components based on the dialogue intent. Secondly, this embodiment of the invention obtains consultation cards and response cards by querying the node cards of the consultation intent and response intent in the product card node tree, which can provide candidate cards for obtaining the first recommended card. The card weights of the consultation cards and response cards are calculated respectively to ensure the final obtaining of the first recommended card. Cards with card weights greater than preset weights are selected as the real-time... The first recommendation card of the conversation determines the first recommendation result of the component, providing a basis for the subsequent combination of the second recommendation result to obtain the final product component recommendation result, thereby improving the accuracy of the recommendation result. Further, according to the dialogue intent, the present invention extracts the conversation keywords of the real-time conversation to provide another product component recommendation method for obtaining the second recommendation card. According to the conversation keywords, the conversation recommendation card of the real-time conversation is queried from the product card node tree. The component node and conversation recommendation card corresponding to the conversation keywords can be quickly retrieved in the product card node tree. The conversation recommendation card is also filtered to obtain the second recommendation card to obtain the component recommendation result of the keyword-associated recommendation method. Finally, based on the first recommendation card and the second recommendation card, the final recommendation card of the real-time conversation is determined. The final recommendation card can be determined based on the recommendation card obtained by the two recommendation methods to improve the accuracy of component recommendation. Therefore, the product component recommendation method, apparatus, electronic device, and storage medium based on real-time conversation proposed in this invention can improve the accuracy of component recommendation results, especially in the recommendation of financial products and e-government services. It can mix and match recommendation results from two dimensions: the dialogue intent of the real-time conversation and keyword association, which greatly improves the accuracy of product component recommendation, meets customers' personalized needs, and enhances customer experience, and has important practical significance for intelligent decision-making. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating a product component recommendation method based on real-time sessions provided in an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of a product component recommendation device based on real-time sessions provided in an embodiment of the present invention;

[0047] Figure 3An internal structure schematic diagram of an electronic device for implementing a product component recommendation method based on a real-time conversation is provided in an embodiment of the present application.

[0048] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0049] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0050] An embodiment of the present application provides a product component recommendation method based on a real-time conversation. The execution subject of the product component recommendation method based on a real-time conversation includes but is not limited to at least one of an electronic device capable of being configured to execute the method provided by the present application, such as a server and a terminal. In other words, the product component recommendation method based on a real-time conversation can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. basic cloud computing services.

[0051] Referring to Figure 1 FIG. 1 shows a flowchart of a product component recommendation method based on a real-time conversation provided by an embodiment of the present application. In the embodiment of the present application, the product component recommendation method based on a real-time conversation includes:

[0052] S1, obtaining a real-time conversation and a corresponding business scenario, and querying a product card node tree corresponding to the real-time conversation according to the business scenario.

[0053] In the embodiment of the present application, the operation object and the application environment of the application scenario can be obtained through the obtaining of the real-time conversation and the corresponding business scenario, which provides guarantee for the implementation of the subsequent component recommendation process. The business scenario is used to describe the application scenario of demand analysis on business items to achieve business goals, such as business scenarios suitable for insurance product recommendation for insurance objects, financial planning for customer asset income, and customer asset investment planning, etc.

[0054] Further, in an optional embodiment of the present application, the obtaining of the real-time conversation and the corresponding business scenario can be obtained through a data script, and the data script can be compiled through a JS script language.

[0055] In the embodiment of the present application, the product card node tree corresponding to the real-time conversation is queried according to the business scenario, which provides support for obtaining a matched node card and obtaining a recommended card of a product component. The product card node tree refers to a flowchart corresponding to a sales script formulated according to a business scenario.

[0056] Further, it should be understood that, in the embodiment of the present application, the product card node tree corresponding to the real-time conversation should be created before the product card node tree corresponding to the real-time conversation is queried, including: formulating a sales script of the real-time conversation under a business scenario according to the business scenario; drawing a flow node tree of the sales script according to the sales script, the flow node tree including a start node, a consultation intent node, a reply intent node, and a judgment node; configuring associated intent keywords and component cards corresponding to the consultation intent node and the reply intent node, and generating the product card node tree corresponding to the real-time conversation.

[0057] S2, identifying a conversation intent of the real-time conversation by using a preset intent recognition model, the conversation intent including a consultation intent and a reply intent.

[0058] In the embodiment of the present application, the conversation intent of the real-time conversation is identified by using the preset intent recognition model, and the conversation intent includes the consultation intent and the reply intent, so that the conversation intents of both parties in the real-time conversation can be obtained to recommend corresponding product components according to the conversation intents. The conversation intent is used to define a user's intention in a conversation to meet the user's demand and achieve the user's desired task, such as querying a financial project or an investment scheme. The consultation intent is an implementation target expressed in the form of a question, for example, a short-term 100,000-yuan high-interest-rate financial product. The reply intent refers to a customer service end identifying the consultation intent of the user's consultation and returning a corresponding reply strategy, for example, a certain financial product with an annual interest rate of 4.5% for a period of three years.

[0059] Further, in the optional embodiment of the present application, the conversation intent of the real-time conversation is identified by using the preset intent recognition model, and the conversation intent includes the consultation intent and the reply intent, including: obtaining a consultation conversation and a reply conversation in the real-time conversation; extracting text feature words of the consultation conversation and the reply conversation, calculating a matched intent classification of the text feature words in the preset intent recognition model by using the preset intent recognition model; and taking the matched intent classification as the conversation intent of the real-time conversation, the conversation intent including the consultation intent and the reply intent.

[0060] Further, in an optional embodiment of the present application, the extraction of the text feature words of the consultation session and the reply session can be achieved by calculating the consultation session and the reply session through a preset text analysis algorithm, the preset text analysis algorithm including information gain (IG), chi-square algorithm, etc., which is used for the selection of the representation of the text and its feature items to quantify the feature words extracted from the text to represent the text information.

[0061] Further, in an optional embodiment of the present application, the calculation of the matched intent classification of the text feature words in the preset intent recognition model using the preset intent recognition model includes: identifying the intent classification text in the preset intent recognition model, performing word segmentation on the intent classification text using a word segmentation tool to obtain intent classification word segmentation; calculating the word vectors of the text feature words and the intent classification word segmentation using a vector conversion model that has been trained; calculating the cosine similarity matrix of the word vectors of the text feature words and the intent classification word segmentation, and taking the intent classification corresponding to the maximum element in the cosine similarity matrix as the matched intent classification. The word segmentation tool is a tool for recombining continuous character sequences into word sequences according to certain specifications, such as jieba word segmentation, ltp word segmentation, and ir word segmentation, etc. The cosine similarity is used to evaluate the similarity of two vectors by calculating the cosine value of the included angle.

[0062] S3, respectively query the node card of the consultation intent and the reply intent in the product card node tree to obtain a consultation card and a reply card, calculate the card weight of the consultation card and the reply card respectively, and take the card with a card weight greater than a preset weight as the first recommended card of the real-time session.

[0063] In the embodiment of the present application, by respectively querying the node card of the consultation intent and the reply intent in the product card node tree to obtain a consultation card and a reply card, candidate cards can be provided for subsequent obtaining of the first recommended card.

[0064] Further, in an optional embodiment of the present application, the respective query of the node card of the consultation intent and the reply intent in the product card node tree to obtain a consultation card and a reply card can be achieved by querying the matching node card of the consultation intent and the reply intent in the product card node tree.

[0065] In the embodiment of the present application, by respectively calculating the card weight of the consultation card and the reply card, support can be provided for finally obtaining the first recommended card.

[0066] Further, the card weight of the consultation card and the reply card is calculated respectively in an optional embodiment of the application, comprising: the child nodes of the consultation card and the reply card in the product card node tree are inquired respectively to obtain the consultation candidate card and the reply candidate card corresponding to the child nodes; the consultation session and the reply session corresponding to the consultation intent and the reply intent in the real-time session are obtained, the similarity of the consultation candidate card and the consultation session and the similarity of the reply candidate card and the reply session are calculated, and the similarity is taken as the card weight of the consultation card and the reply card.

[0067] Further, the calculation of the similarity of the consultation candidate card and the consultation session can be realized by a similarity cosine function, a Pearson correlation coefficient, a Euclidean distance, etc.

[0068] In the embodiment of the application, the component first recommendation result is obtained by taking the card with the card weight greater than the preset weight as the first recommended card of the real-time session, which provides guarantee for obtaining the final product component recommendation result by combining the second recommendation result, so as to improve the accuracy of the recommendation result. The preset weight refers to the importance or contribution or similarity of a certain factor or index relative to a certain thing, which can be set to 0.8 in the application, or can be set according to the actual application scene.

[0069] S4, according to the dialogue intent, extracting the session keyword of the real-time session, according to the session keyword, querying the session recommendation card of the real-time session from the product card node tree, and screening the session recommendation card to obtain the second recommended card.

[0070] In the embodiment of the application, the session keyword of the real-time session is extracted according to the dialogue intent, which can provide another product component recommendation mode for obtaining the second recommended card.

[0071] Further, the session keyword of the real-time session is extracted according to the dialogue intent in an optional embodiment of the application, comprising: obtaining the session text of the real-time session, and extracting the feature word of the session text; according to the dialogue intent, the keyword with dialogue intent is screened out from the feature word. The feature word refers to the word with special meaning in the text, which is used to represent the concept identifier of the text information. The keyword refers to the important word that can express the core meaning in the text.

[0072] Further, in an optional embodiment of the present application, the extraction of the feature words of the conversation text can be achieved by calculating the conversation text through a preset text analysis algorithm, which includes information gain (IG), chi-square algorithm, etc., for the selection of the representation of the text and its feature items to quantify the feature words extracted from the text to represent the text information.

[0073] In the embodiment of the present application, by querying the conversation recommendation card of the real-time conversation from the product card node tree according to the conversation keyword, the component node and the conversation recommendation card corresponding to the conversation keyword can be quickly searched in the product card node tree.

[0074] Further, in an optional embodiment of the present application, the querying of the conversation recommendation card of the real-time conversation from the product card node tree according to the conversation keyword includes: obtaining the intent keywords of the consultation intent node and the reply intent node in the product card node tree; converting the intent keywords and the conversation keyword into intent word vectors and conversation word vectors, respectively, according to the conversation keyword; calculating the similarity of the intent word vectors and the conversation word vectors, and taking the component card corresponding to the intent word vector with a similarity greater than a preset threshold as the conversation recommendation card of the real-time conversation and marking the corresponding conversation recommendation card similarity.

[0075] Further, in an optional embodiment of the present application, the conversion of the intent word vectors and the conversation word vectors can be achieved by a trained vector conversion model, such as Skip-gram, CBOW, LBL, NNLM, C&W, and Glove vector conversion model.

[0076] In the embodiment of the present application, by screening the conversation recommendation card to obtain the second recommendation card, the component recommendation result of the keyword association recommendation mode can be obtained.

[0077] Further, in an optional embodiment of the present application, the screening of the conversation recommendation card to obtain the second recommendation card can be achieved by a preset screening rule of the conversation recommendation card. The preset screening rule of the conversation recommendation card can be to sort the conversation recommendation card similarity in descending order, screen the top several recommendation cards with the highest conversation recommendation card similarity as the second recommendation card.

[0078] S5, determining the final recommendation card of the real-time conversation according to the first recommendation card and the second recommendation card.

[0079] In the embodiment of the present application, the final recommended card of the real-time session can be determined according to the recommended card obtained by the two recommended modes, so that the accuracy of component recommendation is improved.

[0080] Further, in an optional embodiment of the present application, the final recommended card of the real-time session can be determined according to the first recommended card and the second recommended card by mixing and sorting the card weights corresponding to the first recommended card and the second recommended card, and the recommended card corresponding to the highest card weight is taken as the final recommended card.

[0081] It can be seen that the embodiment of the application can obtain the operation object and application environment of the application scenario by acquiring the real-time conversation and the corresponding business scenario, query the product card node tree corresponding to the real-time conversation according to the business scenario to support the subsequent obtaining of the matched node card and thus obtain the recommended card of the product component, and identify the dialogue intention of the real-time conversation by using a preset intention recognition model, the dialogue intention including a consultation intention and a reply intention, so as to obtain the dialogue intention of the real-time conversation to recommend the corresponding product component according to the dialogue intention. Secondly, the embodiment of the application can obtain the consultation card and the reply card by respectively querying the node card of the consultation intention and the reply intention in the product card node tree to provide the candidate card for the subsequent obtaining of the first recommended card, and calculate the card weight of the consultation card and the reply card to provide the guarantee for the final obtaining of the first recommended card. The card with a card weight greater than a preset weight is determined as the first recommended card of the real-time conversation to determine the component first recommended result, which provides the basis for the subsequent combination of the second recommended result to obtain the final product component recommended result, so as to improve the accuracy of the recommended result. Further, the embodiment of the application extracts the conversation keyword of the real-time conversation according to the dialogue intention to provide another product component recommendation mode for the subsequent obtaining of the second recommended card, and queries the conversation recommended card of the real-time conversation from the product card node tree according to the conversation keyword, so as to quickly search the component node and the conversation recommended card corresponding to the conversation keyword in the product card node tree, filter the conversation recommended card to obtain the second recommended card to obtain the component recommended result of the recommendation mode associated with the keyword, and finally determine the final recommended card of the real-time conversation according to the first recommended card and the second recommended card, so as to determine the final recommended card according to the recommended cards obtained by the two recommendation modes to improve the accuracy of the component recommendation. Therefore, the product component recommendation method, device, electronic equipment and storage medium based on the real-time conversation provided by the embodiment of the application can improve the accuracy of the component recommended result, especially in the aspects of the financial management product and the electronic government service recommendation, can mix the recommended results from the two dimensions of the dialogue intention and the keyword association of the real-time conversation, greatly improve the accuracy of the product component recommendation, meet the personalized needs of customers and improve the customer experience, and have important practical significance of intelligent decision-making.

[0082] As Figure 2 shown in the figure is a functional module diagram of the product component recommendation device based on the real-time conversation.

[0083] The product component recommendation device based on real-time conversation in the application can be installed in an electronic device. According to the functions implemented, the product component recommendation device based on real-time conversation can include a product card node tree query module 101, a conversation intent recognition module 102, a first recommendation card generation module 103, a second recommendation card generation module 104, and a final recommendation card generation module 105. The modules in the application can also be referred to as units, which refer to a series of computer program segments that can be executed by the processor of an electronic device and can complete a fixed function, which are stored in the memory of the electronic device.

[0084] In the embodiment, the functions of each module / unit are as follows:

[0085] The product card node tree query module 101 is configured to obtain a real-time conversation and a corresponding business scenario thereof, and query a product card node tree corresponding to the real-time conversation according to the business scenario.

[0086] The conversation intent recognition module 102 is configured to recognize a conversation intent of the real-time conversation by using a preset intent recognition model, wherein the conversation intent includes a consultation intent and a reply intent.

[0087] The first recommendation card generation module 103 is configured to query a node card of the consultation intent and the reply intent in the product card node tree respectively, obtain a consultation card and a reply card, calculate a card weight of the consultation card and the reply card respectively, and take a card with a card weight greater than a preset weight as a first recommendation card of the real-time conversation.

[0088] The second recommendation card generation module 104 is configured to extract a conversation keyword of the real-time conversation according to the conversation intent, query a conversation recommendation card of the real-time conversation from the product card node tree according to the conversation keyword, and screen the conversation recommendation card to obtain a second recommendation card.

[0089] The final recommendation card generation module 105 is configured to determine a final recommendation card of the real-time conversation according to the first recommendation card and the second recommendation card.

[0090] In detail, the modules in the product component recommendation device based on real-time conversation 100 in the embodiment of the application use the same technical means as the product component recommendation method based on real-time conversation in the above Figure 1 , and can produce the same technical effects, which will not be described here.

[0091] As shown in Figure 3 , it is a structural schematic diagram of an electronic device 1 for implementing the product component recommendation method based on real-time conversation in the application.

[0092] The electronic device 1 can include a processor 10, a memory 11, a communication bus 12, and a communication interface 13, and can further include a computer program, such as a real-time conversation-based product component recommendation program, stored in the memory 11 and executable on the processor 10.

[0093] In some embodiments, the processor 10 can be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more combinations of a central processing unit (CPU), a microprocessor, a digital processing chip, a graphics processor, and various control chips. The processor 10 is the control unit of the electronic device 1, which connects various components of the electronic device 1 through various interfaces and lines, executes programs or modules stored in the memory 11 (such as a real-time conversation-based product component recommendation program), and calls data stored in the memory 11 to perform various functions and process data of the electronic device 1.

[0094] The memory 11 includes at least one type of readable storage medium, such as a flash memory, a mobile hard disk, a multimedia card, a card-type memory (such as an SD or DX memory), a magnetic memory, a disk, an optical disk, and the like. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the memory 11 can include both an internal storage unit and an external storage device of the electronic device 1. The memory 11 can be used not only to store application software and various data installed in the electronic device 1, such as the code of a real-time conversation-based product component recommendation program, but also to temporarily store data that has been or will be output.

[0095] The communication bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize the connection and communication between the memory 11 and the at least one processor 10, etc.

[0096] The communication interface 13 is configured to realize the communication between the electronic device 1 and other devices, and includes a network interface and a staff interface. Optionally, the network interface can include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), and is usually used to establish a communication connection between the electronic device 1 and other electronic devices 1. The staff interface can be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the staff interface can also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. The display can also be appropriately referred to as a display screen or a display unit, and is used to display the information processed in the electronic device 1 and to display a visual staff interface.

[0097] Figure 3 Only the electronic device 1 with components is shown, and those skilled in the art can understand that, Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0098] For example, although not shown, the electronic device 1 can also include a power supply (such as a battery) for supplying power to each component. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, so as to realize functions such as charge management, discharge management, and power consumption management through the power management device. The power supply can also include one or more direct current or alternating current power supplies, a recharging device, a power supply fault detection circuit, a power supply converter or inverter, a power supply status indicator, etc. The electronic device 1 can also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which are not described here.

[0099] It should be understood that the embodiments are only for illustration, and the scope of the patent invention is not limited by the structure.

[0100] The product component recommendation program stored in the memory 11 in the electronic device 1 is a combination of multiple computer programs, which can realize the following functions when running in the processor 10:

[0101] Obtaining a real-time conversation and a corresponding business scenario, and querying a product card node tree corresponding to the real-time conversation according to the business scenario;

[0102] Identifying a dialogue intention of the real-time conversation by using a preset intention recognition model, the dialogue intention including a consultation intention and a reply intention;

[0103] Respectively querying a node card of the consultation intention and the reply intention in the product card node tree to obtain a consultation card and a reply card, respectively calculating card weights of the consultation card and the reply card, and taking a card with a card weight greater than a preset weight as a first recommended card of the real-time conversation;

[0104] According to the dialogue intention, extracting a conversation keyword of the real-time conversation, querying a conversation recommended card of the real-time conversation from the product card node tree according to the conversation keyword, and screening the conversation recommended card to obtain a second recommended card;

[0105] According to the first recommended card and the second recommended card, determining a final recommended card of the real-time conversation.

[0106] Specifically, the specific implementation method of the processor 10 to the above computer programs can refer to Figure 1 The description of related steps in the corresponding embodiments will not be repeated here.

[0107] Further, the modules / units integrated in the electronic device 1 are realized in the form of software function units and sold or used as independent products, which can be stored in a nonvolatile computer readable storage medium. The computer readable storage medium can be volatile or nonvolatile. For example, the computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM).

[0108] The application also provides a computer readable storage medium, which stores a computer program, and the computer program can realize the following functions when executed by the processor of the electronic device 1:

[0109] Obtaining a real-time conversation and a corresponding business scenario, and querying a product card node tree corresponding to the real-time conversation according to the business scenario;

[0110] An intent recognition model is used to identify a dialogue intent of the real-time conversation, and the dialogue intent includes a consultation intent and a reply intent.

[0111] The consultation intent and the reply intent are respectively used to query a node card in the product card node tree, to obtain a consultation card and a reply card, the card weights of the consultation card and the reply card are calculated respectively, and the card whose card weight is greater than a preset weight is taken as a first recommended card of the real-time conversation.

[0112] According to the dialogue intent, a conversation keyword of the real-time conversation is extracted, a conversation recommended card of the real-time conversation is queried from the product card node tree according to the conversation keyword, and the conversation recommended card is screened to obtain a second recommended card.

[0113] According to the first recommended card and the second recommended card, a final recommended card of the real-time conversation is determined.

[0114] In several embodiments provided in the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and the division of the modules is merely a logical function division, and there can be another division manner in actual implementation.

[0115] The modules illustrated as separated components can or can not be physically separated, and the components illustrated as modules can or can not be physical units, that is, can be located in one place or distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0116] In addition, each function module in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software function modules.

[0117] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0118] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any additional reference signs in the claims should not be considered as limiting the claims involved.

[0119] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. The artificial intelligence (AI) is to use a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results.

[0120] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The plurality of units or devices stated in the system claims can also be implemented by one unit or device through software or hardware. The second word is used to indicate the name, and does not indicate any specific order.

[0121] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application and are not limited. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A real-time session based product component recommendation method, characterized by, The method comprises: obtaining a real-time session and a corresponding business scenario thereof, and querying a product card node tree corresponding to the real-time session according to the business scenario; identifying conversation texts of both parties of the real-time session by using a preset intention recognition model to obtain an inquiry intention of a client and a reply intention of a customer service end; inquiring a node card of the inquiry intention and the reply intention in the product card node tree respectively to obtain an inquiry card and a reply card, calculating card weights of the inquiry card and the reply card respectively, and taking a card with a card weight greater than a preset weight as a first recommended card of the real-time session; extracting a conversation keyword of the real-time session according to the inquiry intention and the reply intention, inquiring a conversation recommended card of the real-time session from the product card node tree according to the conversation keyword, and screening the conversation recommended card to obtain a second recommended card; determining a final recommended card of the real-time session according to the first recommended card and the second recommended card.

2. The real-time conversational based product component recommendation method of claim 1, wherein, creating a product card node tree corresponding to the real-time session comprises: formulating a sales script of the real-time session under a business scenario according to the business scenario; drawing a flow node tree of the sales script according to the sales script, wherein the flow node tree comprises a start node, an inquiry intention node, a reply intention node, and a judgment node; configuring associated intention keywords and component cards corresponding to the inquiry intention node and the reply intention node to generate the product card node tree corresponding to the real-time session.

3. The real-time conversational based product component recommendation method of claim 1, wherein, The method comprises: obtaining an inquiry conversation of the client and a reply conversation of the customer service end in the real-time session; extracting text feature words of the inquiry conversation and the reply conversation, calculating an intention classification matched by the text feature words in the preset intention recognition model by using the preset intention recognition model; classifying the matched intention classification as the inquiry intention of the client and the reply intention of the customer service end.

4. The real-time conversational based product component recommendation method of claim 1, wherein, The method comprises: identifying intention classification texts in the preset intention recognition model, performing word segmentation on the intention classification texts by using a word segmentation tool to obtain intention classification word segmentation; calculating word vectors of the text feature words and the intention classification word segmentation by using a trained vector conversion model; calculating a cosine similarity matrix of the word vectors of the text feature words and the intention classification word segmentation, and taking an intention classification corresponding to a maximum element in the cosine similarity matrix as the matched intention classification.

5. The real-time conversational based product component recommendation method of claim 1, wherein, The method comprises: inquiring sub-nodes of the inquiry card and the reply card in the product card node tree respectively to obtain inquiry candidate cards and reply candidate cards corresponding to the sub-nodes; The consultation intent and the reply intent corresponding to the real-time conversation are obtained, similarity of the consultation candidate card and the consultation conversation and similarity of the reply candidate card and the reply conversation are calculated, and the similarity is taken as the card weight of the consultation card and the reply card.

6. The real-time conversational based product component recommendation method of claim 1, wherein, The conversation keywords of the real-time conversation are extracted according to the consultation intent and the reply intent, and the conversation keywords include: The conversation text of the real-time conversation is obtained, and the feature words of the conversation text are extracted; The keywords with conversation intent are screened from the feature words according to the consultation intent and the reply intent.

7. The real-time conversational based product component recommendation method of claim 1, wherein, The conversation recommendation card of the real-time conversation is queried from the product card node tree according to the conversation keywords, and the conversation recommendation card includes: The intent keywords of the consultation intent node and the reply intent node in the product card node tree are obtained; The intent keywords and the conversation keywords are respectively converted into intent word vectors and conversation word vectors according to the conversation keywords; The similarity of the intent word vectors and the conversation word vectors is calculated, and the component card corresponding to the intent word vector with the similarity greater than a preset threshold is taken as the conversation recommendation card of the real-time conversation and is marked with the corresponding conversation recommendation card similarity.

8. A real-time session based product component recommendation apparatus, characterized by, The device includes: A product card node tree query module is configured to obtain a real-time conversation and a corresponding business scenario of the real-time conversation, and query a product card node tree corresponding to the real-time conversation according to the business scenario; A conversation intent recognition module is configured to recognize conversation text of the real-time conversation by using a preset intent recognition model, and obtain a consultation intent of a client and a reply intent of a customer service end; A first recommendation card generation module is configured to query node cards of the consultation intent and the reply intent in the product card node tree respectively, obtain a consultation card and a reply card, calculate card weights of the consultation card and the reply card respectively, and take a card with a card weight greater than a preset weight as a first recommendation card of the real-time conversation; A second recommendation card generation module is configured to extract conversation keywords of the real-time conversation according to the consultation intent and the reply intent, query a conversation recommendation card of the real-time conversation from the product card node tree according to the conversation keywords, and screen the conversation recommendation card to obtain a second recommendation card; A final recommendation card generation module is configured to determine a final recommendation card of the real-time conversation according to the first recommendation card and the second recommendation card.

9. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected with the at least one processor in communication; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the product component recommendation method based on the real-time conversation as claimed in any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the product component recommendation method based on the real-time conversation as claimed in any one of claims 1 to 7.

Images