CN116304007A

CN116304007A - Information recommendation method and device, storage medium and electronic equipment

Info

Publication number: CN116304007A
Application number: CN202211411149.XA
Authority: CN
Inventors: 蔡天慧
Original assignee: Ant Fortune Shanghai Financial Information Service Co ltd
Current assignee: Ant Fortune Shanghai Financial Information Service Co ltd
Priority date: 2022-11-11
Filing date: 2022-11-11
Publication date: 2023-06-23

Abstract

The specification discloses an information recommendation method, an information recommendation device, a storage medium and electronic equipment, wherein the method comprises the following steps: determining user dialogue intention based on user dialogue sentences in an interactive dialogue scene, then carrying out information recommendation recall processing based on the user dialogue intention to obtain at least one type of reference transaction content information aiming at the user dialogue intention, carrying out transaction content screening based on the reference transaction content information corresponding to the user dialogue intention and the service characteristic information by acquiring service characteristic information aiming at a customer service end to obtain at least one type of recommended transaction content information aiming at the customer service end, and indicating the customer service end to carry out dialogue reply processing on the customer service end based on the recommended transaction content information in the interactive dialogue scene.

Description

Information recommendation method and device, storage medium and electronic equipment

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to an information recommendation method, an information recommendation device, a storage medium, and an electronic device.

Background

With the development of computer technology, electronic devices are rapidly popularized, and various application programs and web page end programs for providing life convenience services are layered, so as to provide business services (such as travel business services, takeaway business services, consumption financial business services, online shopping business services and the like) for eating and wearing rows of users. In the process of using these business services, users often involve interactive dialogue scenes in customer service systems, for example, users may initiate dialogue queries to system customer service to solve problems or communicate with customer service to solve service matters.

Disclosure of Invention

The specification provides an information recommendation method, an information recommendation device, a storage medium and electronic equipment, wherein the technical scheme is as follows:

in a first aspect, the present specification provides an information recommendation method, the method including:

acquiring user dialogue sentences in an interactive dialogue scene, wherein the interactive dialogue scene is a dialogue scene corresponding to a client and a customer service;

determining user dialogue intention based on the user dialogue statement, and carrying out information recommendation recall processing based on the user dialogue intention to obtain at least one type of reference transaction content information aiming at the user dialogue intention;

acquiring service characteristic information aiming at the customer service end, and performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service characteristic information to obtain at least one type of recommended transaction content information aiming at the customer service end;

and under the interactive dialogue scene, indicating the customer service side to perform dialogue reply processing on the user side based on at least one type of recommended transaction content information.

In a second aspect, the present specification provides an information recommendation apparatus, the apparatus comprising:

the sentence acquisition module is used for acquiring user dialogue sentences in an interactive dialogue scene, wherein the interactive dialogue scene is a dialogue scene corresponding to a client and a customer service;

The recommendation recall module is used for determining user dialogue intentions based on the user dialogue sentences, and carrying out information recommendation recall processing based on the user dialogue intentions to obtain at least one type of reference transaction content information aiming at the user dialogue intentions;

the content screening module is used for acquiring service characteristic information aiming at the customer service end, and carrying out transaction content screening on the basis of the at least one type of reference transaction content information corresponding to the user dialogue intention and the service characteristic information to obtain at least one type of recommended transaction content information aiming at the customer service end;

and the information recommending module is used for indicating the customer service side to perform dialogue reply processing on the user side based on at least one type of recommended transaction content information under the interactive dialogue scene.

In a third aspect, the present description provides a computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the above-described method steps.

In a fourth aspect, the present description provides an electronic device, which may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the above-mentioned method steps.

The technical scheme provided by some embodiments of the present specification has the following beneficial effects:

in one or more embodiments of the present disclosure, an electronic device determines a user dialogue intent based on a user dialogue statement in an interactive dialogue scene, and then performs information recommendation recall processing based on the user dialogue intent to obtain at least one type of reference transaction content information for the user dialogue intent, and screens a plurality of types of reference transaction content information transaction contents based on acquired service feature information for a customer service end to screen recommended transaction content information matching with the current customer service end self recommendation characteristic, so that the customer service end is instructed to perform dialogue reply processing on the user end based on the recommended transaction content information in the interactive dialogue scene, thereby avoiding the situation that the matching degree of the general information recommendation content and the customer service end self recommendation characteristic is low, saving the time of customer service end information recommendation, improving the accuracy of information recommendation and the information recommendation efficiency of the customer service end based on the customer service side information recommendation characteristic and the accurate content recommendation realized by the user side dialogue and the user behavior characteristic, and improving the information recommendation effect of the customer service end in the interactive dialogue scene.

Drawings

In order to more clearly illustrate the technical solutions of the present specification or the prior art, the following description will briefly introduce the drawings that are required to be used in the embodiments or the prior art descriptions, it is obvious that the drawings in the following description are only some embodiments of the present specification, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic view of an information recommendation system provided in the present specification;

FIG. 2 is a flow chart of another information recommendation method provided in the present specification;

FIG. 3 is a flow chart of another information recommendation method provided in the present specification;

FIG. 4 is a schematic diagram of a model process involved in the information recommendation method provided in the present specification;

fig. 5 is a schematic structural view of an information recommendation device provided in the present specification;

fig. 6 is a schematic structural view of an intention determining unit provided in the present specification;

FIG. 7 is a schematic diagram of a recommended recall unit provided herein;

FIG. 8 is a schematic diagram of a content screening module provided in the present specification;

fig. 9 is a schematic diagram of a structure of a content screening unit provided in the present specification;

Fig. 10 is a schematic structural view of an electronic device provided in the present specification;

FIG. 11 is a schematic diagram of the architecture of the operating system and user space provided herein;

FIG. 12 is an architecture diagram of the android operating system of FIG. 11;

FIG. 13 is an architecture diagram of the IOS operating system of FIG. 11.

Detailed Description

The following description of the embodiments of the present invention will be made apparent from, and elucidated with reference to, the drawings of the present specification, in which embodiments described are only some, but not all, embodiments of the present specification. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.

In the description of the present specification, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. In the description of the present specification, it should be noted that, unless expressly specified and limited otherwise, "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus. The specific meaning of the terms in this specification will be understood by those of ordinary skill in the art in the light of the specific circumstances. In addition, in the description of the present specification, unless otherwise indicated, "a plurality" means two or more. "and/or", describes an association relationship of an association object, and indicates that there may be three relationships, for example, a and/or B, and may indicate: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.

In the related art, in order to better provide a dialogue interactive experience for a user or customer service, it is often involved to perform intent recognition on dialogue sentences of the user side, and the system engine intelligence for integrating information recommendation based on dialogue intent of the user can generate one or more types of information recommendation content according to the dialogue intent of the user, where the one or more types of system content can assist the customer service of the customer service side to perform dialogue reply to the user of the user side. However, for the customer service side, the associated customer service side is usually massive, the customer service side is difficult to face massive user problems, even if one or more types of information recommendation contents are generated in an assisted manner according to the dialogue intention system of the user, due to the limitation of the information recommendation, the assisted generation of general information recommendation contents can only be suitable for general interactive dialogue scenes and can not help the customer service side to rapidly conduct personalized recommendation, and the customer service side needs to spend time and effort to screen point of interest (POI) contents to conduct dialogue reply to the customer side based on the assisted generation of information recommendation contents, which definitely does not meet the real-time requirement in the interactive dialogue scenes, greatly influences the information recommendation efficiency of the customer service side, and causes the situation that the whole information recommendation effect in the interactive dialogue scenes is poor.

The present specification is described in detail below with reference to specific examples.

Referring to fig. 1, a schematic view of a scenario of an information recommendation system provided in the present specification is provided. As shown in fig. 1, the information recommendation system may include at least a client cluster and a service platform 100.

The client cluster may include at least one client, where an associated object of the client may be a user or a customer service, and the customer service is associated with a service platform 100 that provides a corresponding transaction service (such as a consumer financial transaction service, an online shopping transaction, and a logistics express transaction service), as shown in fig. 1, specifically includes a client 1 corresponding to an associated object 1, clients 2 and … corresponding to an associated object 2, and a client n corresponding to an associated object n is an integer greater than 0.

Each client in the client cluster may be a communication-enabled electronic device including, but not limited to: wearable devices, handheld devices, personal computers, tablet computers, vehicle-mounted devices, smart phones, computing devices, or other processing devices connected to a wireless modem, etc. Electronic devices in different networks may be called different names, for example: a user equipment, an access terminal, a subscriber unit, a subscriber station, a mobile station, a remote terminal, a mobile device, a user terminal, a wireless communication device, a user agent or user equipment, a cellular telephone, a cordless telephone, a personal digital assistant (personal digital assistant, PDA), an electronic device in a 5G network or future evolution network, and the like.

The service platform 100 may be a separate server device, such as: rack-mounted, blade, tower-type or cabinet-type server equipment or hardware equipment with stronger computing capacity such as workstations, mainframe computers and the like is adopted; the server cluster may also be a server cluster formed by a plurality of servers, and each server in the server cluster may be formed in a symmetrical manner, wherein each server is functionally equivalent and functionally equivalent in a transaction link, and each server may independently provide services to the outside, and the independent provision of services may be understood as no assistance of another server is needed.

In one or more embodiments of the present disclosure, the service platform 100 may establish a communication connection with a client in a client cluster and a user side, and complete data interaction in an information recommendation process in an interaction dialogue scene based on the communication connection, for example, the service platform 100 may collect a user dialogue statement from the user side based on an information recommendation method of the present disclosure; for another example, the service platform 100 may execute the information recommendation method of the present disclosure to obtain a plurality of types of reference transaction content information corresponding to the user dialogue intention, thereby recalling the user dialogue intention, without pushing information content to the customer service end, and then perform transaction content screening on the plurality of types of reference transaction content information based on the service feature information of the customer service end, to obtain recommended transaction content information conforming to the customer service end. For another example, a user terminal in the client cluster may initiate a dialogue window to the service platform 100 to perform a dialogue, the service platform 100 may allocate a customer service terminal to the user terminal, and the customer service terminal may obtain, based on an information recommendation method of the service platform 100, a plurality of types of recommended transaction content information that fit the user terminal after receiving dialogue sentences of the user terminal, so as to perform dialogue reply processing to the user terminal, and so on.

It should be noted that, the service platform 100 establishes a communication connection with at least one client in the client cluster through a network for interactive communication, where the network may be a wireless network, or may be a wired network, where the wireless network includes, but is not limited to, a cellular network, a wireless local area network, an infrared network, or a bluetooth network, and the wired network includes, but is not limited to, an ethernet network, a universal serial bus (universal serial bus, USB), or a controller area network. In one or more embodiments of the specification, techniques and/or formats including HyperText Mark-up Language (HTML), extensible markup Language (Extensible Markup Language, XML), and the like are used to represent data exchanged over a network (e.g., target compression packages). All or some of the links may also be encrypted using conventional encryption techniques such as secure socket layer (Secure Socket Layer, SSL), transport layer security (Transport Layer Security, TLS), virtual private network (Virtual Private Network, VPN), internet protocol security (Internet Protocol Security, IPsec), and the like. In other embodiments, custom and/or dedicated data communication techniques may also be used in place of or in addition to the data communication techniques described above.

The embodiments of the dialog structure processing system provided in the present disclosure and the dialog structure processing methods in one or more embodiments belong to the same concept, where an execution body corresponding to the dialog structure processing methods related to one or more embodiments in the present disclosure is an electronic device, and the electronic device may be the service platform 100 described above; the execution subject corresponding to the dialog structure processing method in one or more embodiments of the present disclosure may also be a client, and specifically determined based on an actual application environment. Embodiments of the dialog structure processing system may be implemented by the following method embodiments, which are not described herein.

Based on the schematic view of the scenario shown in fig. 1, the information recommendation method provided in one or more embodiments of the present disclosure is described in detail below.

Referring to fig. 2, a flow diagram of an information recommendation method, which may be implemented in a computer program and may be executed on a von neumann system-based information recommendation device, is provided for one or more embodiments of the present disclosure. The computer program may be integrated in the application or may run as a stand-alone tool class application. The information recommending device may be a service platform.

Specifically, the information recommendation method comprises the following steps:

s102: acquiring user dialogue sentences in an interactive dialogue scene, wherein the interactive dialogue scene is a dialogue scene corresponding to a client and a customer service;

the interactive dialogue scene can be a dialogue scene of a user side and a customer service side involved in corresponding transactions, and the interactive dialogue mode used by the interactive dialogue scene can be various, for example, a telephone mode is adopted to carry out dialogue between the user side and the customer service side, for example, an instant messaging window mode is adopted to carry out dialogue between the user side and the customer service side, for example, a video/audio communication mode is adopted to carry out dialogue between the user side and the customer service side.

Schematically, in an interactive dialogue scene where a user terminal and a customer service terminal are located, the electronic device may collect user dialogue sentences in real time, obtain at least one type of recommended transaction content information by executing an information recommendation method according to one or more embodiments of the present disclosure, send the at least one type of recommended transaction content information to the customer service terminal, after the customer service terminal receives several types of recommended transaction content information, select target transaction content information based on the recommended transaction content information and perform dialogue quick reply processing on the user terminal based on the target transaction content information in the interactive dialogue scene, for example, send the target transaction content information to the user terminal quickly, or perform finishing information optimization on the target transaction content information by the customer service terminal and output the target transaction content information to the user terminal

It can be understood that, in the interactive dialogue scene, the user end and the customer service end as the two interactive parties can perform multiple rounds of dialogue based on corresponding transaction requirements to generate dialogue sentences, for example, the user end can interact with the customer service end based on the transaction consultation requirements on the service platform to generate multiple rounds of dialogue sentences, for example, the user of the user end can initiate multiple rounds of commodity inquiry sentences to the customer service of the customer service end based on commodity details, for example, the user of the user end can initiate multiple rounds of object inquiry sentences to the customer service of the customer service end based on the money elimination objects (such as a certain fund product object, a certain financial product object and the like).

It can be understood that the user dialogue sentence is at least one dialogue sentence initiated or sent by the user terminal to the customer service terminal in the interactive dialogue scene, and it can be understood that the user dialogue sentence can be a real-time dialogue sentence sent or initiated by the user terminal to the customer service terminal currently, or can be a real-time dialogue sentence and a historical dialogue sentence sent or initiated by the user terminal to the customer service terminal currently in the interactive dialogue scene, that is, the number of the user dialogue sentences can be one or more, and the user dialogue intention of the current user terminal can be accurately predicted by combining the user dialogue sentence in the interactive dialogue scene with one or more interactive dialogue sentences.

In one or more embodiments of the present disclosure, an interactive session scene is generated based on actual transaction requirements, where the interactive session scene may be a travel service session scene, a take-out service, an online shopping, a pay-off service session scene, or the like, and the interactive session scene generally includes at least one round of session between a user side and a server side, and by executing an information recommendation method related to the present disclosure, at least one type of recommended transaction content information for the customer service side may be obtained, so that the customer service of the customer service side is assisted to perform transaction content recommendation to the user side based on the determined plurality of recommended transaction content information in the interactive session scene.

In one or more embodiments of the present disclosure, after one or more user dialogue sentences are acquired, dialogue noise reduction processing may be performed on the user dialogue sentences, so as to filter non-key contents in the dialogue to retain key dialogue contents, for example, non-key contents such as call-in-call dialogue information, automatic response dialogue information, and salutation information in the dialogue, and the noise-reduced dialogue data only retains sentences in a role query dialogue form or a role reply dialogue form.

S104: determining user dialogue intention based on the user dialogue statement, and carrying out information recommendation recall processing based on the user dialogue intention to obtain at least one type of reference transaction content information aiming at the user dialogue intention;

In a possible implementation manner, the user dialogue intent may be determined only through the user dialogue sentence, which may be understood that the electronic device may perform intent semantic recognition based on the user dialogue sentence (such as the user dialogue query), so as to obtain the user dialogue intent corresponding to the user dialogue sentence.

Illustratively, the intent semantic recognition object may be a dialogue sentence for one or more user dialogue sentences, the dialogue full text semantic (full text semantic may be understood as the whole information semantic from the initiating dialogue to the current dialogue sentence) is referred to in the specification to extract the slot information which is not involved in the dialogue sentence but expressed by the user, the intent recognition model may be constructed based on a machine learning model, the intent recognition model may be used to take the user dialogue sentence as data input, the user dialogue sentence includes one or more dialogue sentences, the dialogue sentence and the full text semantic information of the dialogue sentence are recognized based on the intent recognition model to obtain the accurate dialogue semantic feature of the user dialogue sentence, the dialogue semantic feature fuses the semantic of the dialogue sentence and the full text semantic of the dialogue sentence, and the intent recognition model then performs the intent recognition based on the dialogue semantic feature of the user dialogue sentence, thereby outputting the user dialogue intent corresponding to the user dialogue sentence.

The dialogue semantic features are dialogue semantic attributes specific to unstructured data expressed in natural language, and comprise dialogue intention, dialogue topic description, bottom feature meaning, context semantic and other semantic elements.

Alternatively, the intent recognition model may be trained based on a machine learning model, including, but not limited to, fitting of one or more of a convolutional neural network (Convolutional Neural Network, CNN) model, a deep neural network (Deep Neural Network, DNN) model, a recurrent neural network (Recurrent Neural Networks, RNN) model, a pre-trained language model (Bidirectional Encoder Representation from Transformers, BERT), an embedded (emmbedding) model, a gradient-lifting decision tree (Gradient Boosting Decision Tree, GBDT) model, a logistic regression (Logistic Regression, LR) model, a BERT model, and the like.

Illustratively, a large amount of dialogue text sample data can be obtained in advance, the dialogue text sample data is used for training the intention recognition model, and the trained intention recognition model can be obtained after training is completed. In the actual application stage, in the interactive dialogue scene between the user side and the customer service side, according to the real-time user dialogue statement of the user side, the interactive dialogue statement is input into the intention recognition model, and the intention recognition model is used for recognizing the user dialogue intention corresponding to the user dialogue statement.

In a possible implementation manner, the user behavior information of the user terminal may be obtained, the electronic device determines the user dialogue intent through the user dialogue statement and the user behavior information, which may be understood that the electronic device may perform intent semantic recognition based on the user dialogue statement (such as the user dialogue query) and the user behavior information, so as to obtain the user dialogue intent corresponding to the user dialogue statement.

The user behavior information is behavior data generated by a user at the user end in the process of browsing or using a target transaction service (provided by a service platform), and the user behavior information can be fit of one or more of browsing behavior data, operation behavior data, data transfer amount data, user account data amount and other data types of corresponding transaction objects in the target transaction service.

It will be appreciated that user behavior information is typically data that has been generated relative to the time corresponding to the interactive dialog scenario.

Illustratively, the device providing the corresponding transaction service (e.g. consumption financial service, shopping service, logistics service, etc.) can collect user behavior information and user identification in a transaction service application or a transaction service web page end in the preset transaction service, wherein the user identification is used for carrying out unique identity identification on the corresponding user from a plurality of users, the user behavior information can be fit of one or more of browsing behavior data, operation behavior data, data transfer amount data, user account data amount and other data types of the corresponding transaction object in the target transaction service,

The user behavior information may be used to some extent to indicate that the user's interactive behavior intent is identified in the interactive dialog scenario, which is associated with the user's current intent of the user dialog sentence. For example, the user side can send the user behavior information and the user identification to the electronic device (such as a service platform) for storage in the daily use process of the user, and the electronic device uses the user identification data as the associated keywords to store in a preset database. Furthermore, in the actual application stage, the electronic device can comprehensively measure the intention of the user behavior dimension in addition to the intention of the reference dialog dimension in the recognition stage of the user dialog statement, and deep mine the potential user dialog intention of the current user dialog statement, so that the problem that the user intention is difficult to mine based on the dialog dimension in objective dialog forms such as unobvious dialog intention representation, fuzzy dialog information and the like in some interactive dialog scenes can be avoided, and the intention recognition accuracy can be improved.

In one or more embodiments of the present disclosure, a pre-trained intent recognition model may be used to input a user dialogue sentence (such as a user dialogue query) and user behavior information as models, where the intent recognition model performs intent semantic recognition based on the user dialogue sentence and the user behavior information, so as to obtain a user dialogue intent corresponding to the user dialogue sentence.

It will be appreciated that in one or more embodiments of the present description, an intent recognition model may be provided with model recognition capabilities to recognize user dialog intent from a dialog dimension based solely on user dialog statements, or to recognize user dialog intent from a combination of dialog and user behavior dimensions based on user dialog statements and user behavior information, through model training.

Illustratively, in the model training stage of the intent recognition model, a large amount of dialogue text sample data is adopted, the dialogue text sample data comprises sample user behavior information besides at least sample dialogue sentences of a sample user, the intent recognition model is trained by using the information dialogue text sample data at least comprising sample dialogue sentences of the sample user and sample user lines, and a trained intent recognition model can be obtained after training is completed.

Schematically, in the model training process, dialogue text sample data can be marked, user dialogue intention labels are marked, model training adjustment is performed by adopting a back propagation learning algorithm based on the output user intention of each round of intention recognition model and the user dialogue intention labels until the model meets the model training ending condition, and a trained intention recognition model is obtained.

In practical application, the electronic device is generally configured with a recommendation information matching policy, wherein the recommendation information matching policy can be characterized by integrating a personalized information recommendation system or a personalized information recommendation rule, and a plurality of types of reference transaction content information can be output by adopting the recommendation information matching policy based on intention information; the recommendation information matching policy is determined based on a platform information matching algorithm set in a development and operation stage, for example, for certain type of intention information, the recommendation information matching policy can be associated with type A reference transaction content, type B reference transaction content, type C reference transaction content and the like according to the set platform information recommendation algorithm. It can be appreciated that the several types of reference transaction content information output by the recommendation information matching policy are typically intended user recommendations.

In one or more embodiments of the present disclosure, the electronic device determines several types of reference transaction content information based on a configured personalized information recommendation system or a personalized information recommendation rule with a user dialogue intention as intention information, and does not directly output the several types of reference transaction content information at this time, but acquires the several types of reference transaction content information for information recommendation recall.

Illustratively, the configured personalized information recommendation system or personalized information recommendation rule is specifically set based on the actual application environment, and is not specifically limited herein.

In one possible implementation, the electronic device may determine a target intent type corresponding to the user dialog intent, and generate at least one type of reference transaction content information based on a recommendation information matching policy corresponding to the target intent type;

in one or more embodiments of the present description, a user dialog intention may be a subdivided multi-level intention type, the user dialog intention being characterized by an intention encoding format of the multi-level intention type, the user dialog intention may be composed of a number of intention major classes, each of which may contain at least one sub-class of intention subclass, such as, for example, a consumer financial transaction, such as: the intention is formed by three intention types, namely ' product information query|fund evaluation|search for red, intention ' market interpretation|market + industry|operation query| ', the intention can further focus on the intention subdivision field through ' intention coding format of multi-level intention type ', the meaning of deeply mined user intention can be represented, and information recommendation content can be accurately matched based on the representation form of user dialogue intention with higher fine granularity.

Alternatively, each target intention type may be preset with a recommendation information matching policy, that is, several associated reference transaction content items are set for different target intention types, and then reference transaction content information under the several reference transaction content items is generated.

Such as: the intention type of the consumption finance target can be set to be subjective type, objective type and boring type based on the actual application condition of the dialogue, and the recommendation information matching strategy corresponding to the subjective type is recommendation: transaction contents of category items such as a communication frame item, a transaction product object card item, a quotation view item and the like; the recommendation information matching strategy corresponding to the objective class is recommendation: transaction product object introduction items, platform/service introduction items, financial knowledge items and the like; the recommended information matching strategy corresponding to the gossip class is recommended: transaction content of category items such as calling dialogue information items, chat reference conversation items and the like;

illustratively, assuming that the target intent type is objective type intent, and specifically intent "product information query |fund evaluation|equity query," a recommendation information matching strategy such as entity matching, FAQ matching, etc. may be employed to generate or obtain product introduction, financial knowledge, etc. for a suitable fund object.

Further, after generating at least one type of reference transaction content information based on the recommended information matching policy corresponding to the target intention type, the electronic device may cancel pushing the reference transaction content information to the customer service side and perform a system recall process on the reference transaction content information, so as to obtain at least one type of reference transaction content information after the system recall process.

S106: acquiring service characteristic information aiming at the customer service end, and performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service characteristic information to obtain at least one type of recommended transaction content information aiming at the customer service end;

in one or more embodiments of the present disclosure, a target transaction service (such as a consumer financial service, a shopping service, or an internet of things service) corresponding to an electronic device is generally configured with a plurality of reference customer service ends, customer service personnel of different reference customer service ends can correspond to different recommendation favorites, different recommendation styles, and different customer service levels, and when the customer service ends provide services to users, general information recommendation contents or unified information recommendation contents determined based on a configured recommendation information matching policy are not considered on one hand, customer service personalized service recommendation characteristics of different customer service ends are not considered, personalized characteristics are not provided, and a plurality of types of information recommendation contents positioned or hit are not necessarily information recommendation characteristics conforming to customer service of the current customer service end; on the other hand, the customer service side recommends the content based on the unified hit or the auxiliary generated general information and needs to expend time and effort to screen the content of the point of interest (POI) to carry out dialogue reply to the user side, which definitely does not meet the real-time requirement in the interactive dialogue scene, greatly influences the information recommendation efficiency of the customer service side, and causes poor whole information recommendation effect in the interactive dialogue scene.

Further, after the electronic device carries out system recall to obtain at least one type of reference transaction content information, transaction content screening can be carried out on the at least one type of reference transaction content information corresponding to the user dialogue intention by combining with service characteristic information of the customer service end so as to screen recommended transaction content information which is fit for the personalized recommendation characteristic of the current customer service end.

The service feature information may be understood as a customer service feature characterizing when the customer service side performs information recommendation service on the user side under a corresponding transaction (such as a cash elimination transaction, a shopping transaction, and an internet of things transaction), such as content recommendation preference when the customer service side performs information recommendation, a commonly used customer service communication frame type, an affiliated service style, a good service class, service recommendation result index attributes (such as exposure, adoption amount, and adoption rate in service recommendation), customer service work index attributes (such as instant messaging amount, number of outbound calls, duration of outbound calls, and the like of current customer service), data transfer conversion features (such as data transfer conversion index of a user after the customer service information recommendation), and the like. It can be understood that the personalized recommendation characteristics and the service conversion characteristics of the customer service end can be represented through the service characteristic information, and the secondary personalized screening recommendation based on the customer service end for the general information recommendation content or the unified information recommendation content can be assisted.

In a possible implementation manner, transaction content screening is performed based on the at least one type of reference transaction content information and the service feature information corresponding to the user dialogue intention, which may be that recommendation sequence adjustment is performed on a plurality of types of reference transaction content information in combination with service feature information of a customer service end, and reference transaction content information conforming to personalized recommendation features of a current customer service end is promoted and sequenced with priority, so that at least one type of recommended transaction content information for the customer service end after recommendation sequence adjustment can be obtained; the business content information meeting the recommendation preference of the customer service side can be promoted and sequenced according to the customer service behavior characteristics of the customer service side, the business content information meeting the service style of the customer service side can be promoted and sequenced according to the customer service behavior characteristics of the customer service side, and the business content information meeting the recommendation preference of the customer service side can be promoted and sequenced according to the business content information with better transfer and conversion characteristics of the customer service side.

In a possible implementation manner, the transaction content screening may be performed based on the at least one type of reference transaction content information and the service feature information corresponding to the user dialogue intention, which may be that transaction content information that is not matched with the service feature information of the customer service end is filtered, so that at least one type of recommended transaction content information for the customer service end after filtering may be obtained; for example, the transaction content information with poor transfer and conversion characteristics with customer service side data can be deleted, etc.

Optionally, based on service feature information of the customer service end, personalized grading can be performed on each type of reference transaction content information to obtain grading scores of each type of reference transaction content information, and each type of reference transaction content information is adjusted according to the high-low order of the grading scores, for example, the display order of each type of reference transaction content information is adjusted, for example, the reference transaction content information with the grading scores smaller than a grading threshold value is deleted, so that at least one type of recommended transaction content information aiming at the customer service end can be obtained, at this time, the obtained recommended transaction content information can fully consider customer service personalized service recommendation characteristics of different customer service ends, the recommended transaction content has personalized characteristics, and the situation that a plurality of positioned or hit type of information recommended content is not necessarily matched with the information recommendation characteristics of the customer service end at present is avoided; on the other hand, the time for the customer service terminal to perform dialogue reply to the user terminal based on the unified hit information recommendation content screening point of interest (POI) is saved, the information recommendation efficiency of the customer service terminal is improved, and the whole information recommendation effect under an interactive dialogue scene is improved; on the other hand, the screening of the content recommended by the follow-up information is indicated by the data transfer and conversion condition of the user side after the information recommendation of the customer service side, and the accurate content recommendation can be realized.

S108: and under the interactive dialogue scene, indicating the customer service side to perform dialogue reply processing on the user side based on at least one type of recommended transaction content information.

The electronic device may send at least one type of recommended transaction content information to the customer service side, and after the customer service side receives the plurality of types of recommended transaction content information, the customer service side may select target transaction content information based on the recommended transaction content information in an interactive dialogue scene and perform dialogue reply processing on the user side based on the target transaction content information, for example, send the target transaction content information to the user side, and perform dialogue reply to the user side based on the target transaction content information, for example, so as to provide professional customer service transaction service.

Referring to fig. 3, fig. 3 is a flow chart illustrating another embodiment of an information recommendation method according to one or more embodiments of the present disclosure. Specific:

s202: acquiring user dialogue sentences in an interactive dialogue scene, wherein the interactive dialogue scene is a dialogue scene corresponding to a client and a customer service;

reference may be made specifically to the method steps of other embodiments of the present disclosure, and details are not repeated here.

S204: acquiring user behavior information aiming at the user terminal, and determining user dialogue intention through the user dialogue sentences and the user behavior information;

according to some embodiments, the user behavior information may be behavior data generated by a user at the user end in the process of browsing or using the target transaction service (provided by the service platform), and the user behavior information may be a fit of one or more of browsing behavior data, operation behavior data, data transfer amount data, user account data amount and other data types of corresponding transaction objects in the target transaction service.

Illustratively, taking the target transaction service (provided by the service platform) as a consumer financial service (i.e., a vanishing transaction service), one or more of the following forms may be included:

The user behavior information can be the browsing data of the money-eliminating object in a certain time, such as the browsing frequency characteristics of the money-managing object of the time index of the user in the time of about 1 day, about 7 days, about 14 days and about 30 days;

the user behavior information may be object option click frequency data of the financial object in a certain time, such as the click frequency characteristics of the financial object of the time index of the user such as the near 1 day, the near 7 days, the near 14 days, the near 30 days, etc.;

the user behavior information may be data of data transfer amounts related to acquiring or transferring a financial object in a certain time, for example, data transfer amounts related to a user going to a certain financial object later (i.e. data transfer amounts related to acquiring a certain financial object transferred from a user associated account to other associated accounts);

the user behavior information may be a user account data amount of a financial related data account;

the user behavior information may be a data rate characteristic (data rate of increase or data rate of decrease) of the user's vanishing data account, and so on.

The user behavior information can be the browsing times, the clicking times and the data transfer quantity characteristics of target gold-eliminating object plates (such as white wine, new energy, medical, military and other plate objects) in a certain time (such as within 30 days);

In one or more embodiments of the present disclosure, the electronic device may perform intent semantic recognition based on a user dialogue sentence (e.g., a user dialogue query) and user behavior information, so as to obtain a user dialogue intent corresponding to the user dialogue sentence.

Schematically, the electronic device may perform intent semantic recognition based on a pre-trained intent recognition model, so as to obtain a user dialogue intent corresponding to a user dialogue sentence, as follows:

a2, the electronic equipment can input user behavior information and user dialogue sentences into an intention recognition model, and the intention recognition model is used for extracting user behavior characteristics corresponding to the user behavior information and user dialogue characteristics corresponding to the user dialogue sentences;

illustratively, when the interactive dialog scene is a vanishing dialog scene (e.g., an interactive dialog scene under a vanishing transaction), the user behavior information may be referred to as user vanishing behavior information,

further, the step of extracting, by the electronic device, the user behavior feature corresponding to the user behavior information through the intent recognition model may specifically be:

and the electronic equipment extracts user degaussing behavior characteristics from the user degaussing behavior information through the intention recognition model, wherein the user degaussing behavior characteristics comprise at least one of transaction object browsing characteristics, transaction object clicking characteristics, object acquired data transfer quantity characteristics, object total account data characteristics and user data profit characteristics.

Illustratively, the user behavior characteristics corresponding to the user behavior information are extracted through the intention recognition model, and the user behavior characteristics can be directly obtained after the relevant characteristic data in the user behavior information are subjected to characteristic coding.

For example, the transaction object browsing feature may be a financial object browsing number feature of the time index of the user of near 1 day, near 7 days, near 14 days, near 30 days, etc.;

for example, the business object click feature may be a number of times the user clicks on the financial object for the time index of approximately 1 day, approximately 7 days, approximately 14 days, approximately 30 days, etc.;

for example, the object acquisition data transfer amount feature may be a feature extracted from data transfer amount data related to acquisition or transfer of a financial object in a certain time;

for example, the object general account data feature may be a user account data amount of a financial-related data account;

for example, the user data benefit feature may be a data rate of change feature (data rate of increase or data rate of decrease) of the user's vanishing data account;

schematically, extracting user behavior characteristics corresponding to the user behavior information and user dialogue characteristics corresponding to the user dialogue sentences through an intention recognition model;

a4, performing feature stitching on the user behavior features and the user dialogue features through the intention recognition model to obtain user high-order features;

And A6, carrying out potential intention recognition on the user high-order features through the intention recognition model so as to output user dialogue intention.

Illustratively, taking the intention recognition model as an example based on BERT model training in the machine learning model, as shown in fig. 4, fig. 4 is a schematic diagram of model processing, where the intention recognition model includes at least an input layer, an encoding layer, a vector stitching layer and an intention recognition score layer, and only a partial feature extraction flow related to the intention recognition model is shown in fig. 4.

The input layer in the intention recognition model comprises a plurality of Embedding layers, the Embedding layers can be token Embedding, segment Embedding and position Embedding, the intention recognition model can carry out Embedding processing on the user dialogue statement through the input layer, and the Embedding results such as token Embedding, segment Embedding and position Embedding are added to obtain an input representation of each word in the user dialogue statement, namely an Embedding vector Embedding;

the coding layer in the intention recognition model is usually based on a transducer encoder to code the representation of the input dialogue sequence, and the embedded vector Embedding is coded by the coding layer to obtain the user dialogue characteristics corresponding to the user dialogue sentences.

And the vector splicing layer in the intention recognition model performs characteristic splicing on the user behavior characteristics and the user dialogue characteristics to obtain the user high-order characteristics.

The intention recognition scoring layer in the intention recognition model is used for carrying out potential intention recognition based on the high-order characteristics of the user so as to score each intention, and feeding back the probability of a certain intention based on the intention scoring, so that the model output user dialogue intention is obtained.

S206: performing information recommendation recall processing based on the user dialogue intention to obtain at least one type of reference transaction content information aiming at the user dialogue intention;

S208: acquiring service characteristic information aiming at the customer service end, and determining customer service behavior information and data transfer conversion information aiming at the customer service end from the service characteristic information;

in one or more embodiments of the present disclosure, a target transaction service (such as a consumer financial service, a shopping service, or an internet of things service) corresponding to an electronic device is generally configured with a plurality of reference customer service ends, customer service personnel of different reference customer service ends can correspond to different recommendation favorites, different recommendation styles, and different customer service levels, when the customer service ends provide services for users, a configured recommendation information matching policy (such as a recommendation information matching engine) is adopted based on user dialogue intention, so that universal multi-class reference transaction content information can be obtained, at this time, after the electronic device performs system recall to obtain multi-class reference transaction content information, then, transaction content screening can be performed on at least one class of reference transaction content information corresponding to the user dialogue intention in combination with service feature information of the customer service ends, so as to screen out recommendation transaction content information matching with personalized recommendation characteristics of the current customer service ends.

According to some embodiments, the service feature information may be understood as a customer service feature characterizing when the customer service side performs information recommendation service on the customer side under corresponding transactions (such as a cash elimination transaction, a shopping transaction, and an internet of things transaction). The personalized recommendation characteristics of the customer service end can be represented through the service characteristic information, and the secondary personalized screening recommendation based on the customer service end for the general information recommendation content or the unified information recommendation content can be assisted.

It can be understood that the customer service behavior information is behavior data generated by customer service of the customer service end in the process of maintaining the target transaction service (provided by the service platform), and the customer service behavior information can be service recommendation result index attribute (such as exposure, adoption and adoption rate in service recommendation) in the process of performing service information recommendation on the customer service end by the maintenance target transaction service, customer service work index attribute (such as instant messaging volume, number of outbound calls, outbound call duration and the like of the current customer service), content recommendation preference when the customer service end performs information recommendation, commonly used customer service communication frame type, service style and service class.

Illustratively, taking the target transaction service as the cash elimination transaction service as an example, the customer service behavior information can be exposure, adoption and adoption rate information in service recommendations of a financing engineer for 7 days and nearly 14 days; exposure, adoption and adoption rate characteristics of type service content such as day 7, day 14 communication frames, product cards, market views, etc. of financing operators; such as IM message volume, number of outbound calls, duration of outbound call feature information, which may be near 7 days, near 14 days, of financial institute customer service;

It can be understood that the data transfer conversion information feeds back the data transfer conversion index of the user after the customer service side carries out information recommendation.

Illustratively, taking the target transaction service as the cash elimination transaction service as an example, the data transfer conversion information user side obtains the times of products, obtains the total data transfer amount of products and the like within 7 days and 14 days after dialogue communication with the financing operator customer service side in nearly half a year.

It can be understood that after the electronic device obtains the service feature information for the customer service side, the customer service behavior information and the data transfer conversion information for the customer service side are determined from the service feature information, and then transaction content screening can be performed on the at least one type of reference transaction content information corresponding to the user dialogue intention by combining the information so as to screen out recommended transaction content information which meets the personalized recommendation characteristic of the current customer service side.

S210: transaction content screening is carried out on at least one type of reference transaction content information corresponding to the user dialogue intention based on the customer service behavior information and the data transfer conversion information, so that at least one type of recommended transaction content information aiming at the customer service terminal is obtained;

the recommended transaction content information can be understood as transaction content information which is matched with personalized recommendation characteristics of the customer service side on the basis of general reference transaction content information determined by a recommendation information matching engine, and compared with the general reference transaction content information, the recommended transaction content information is more easily adopted by the customer service side and is further subjected to dialogue reply to the customer side.

In one or more embodiments of the present disclosure, information content is secondarily screened from a customer service behavior dimension and a data transfer transformation dimension, so as to better fit a customer service end, and assist the customer service end in efficiently performing personalized service recommendation.

In one possible implementation, transaction content screening can be performed on a plurality of reference transaction content information by combining customer service behavior information and data transfer conversion information through a pre-trained information recommendation model.

Illustratively, at least one type of reference transaction content information, customer service behavior information and data transfer conversion information are taken as the model input of an information recommendation model; the information recommendation model combines customer service behavior information and data transfer conversion information of a customer service side to realize transaction content screening so as to output at least one type of recommended transaction content information after content screening.

Further, the transaction content filtering may be to score the determined universal reference transaction content information, and implement content filtering based on the score.

The electronic equipment can input at least one type of reference transaction content information, customer service behavior information and data transfer conversion information corresponding to the dialogue intention of the user into an information recommendation model, and performs information recommendation scoring on the various types of reference transaction content information based on the customer service behavior information and the data transfer conversion information through the information recommendation model to obtain content information item scoring of the various types of reference transaction content information;

In one or more embodiments herein, the information recommendation model may be trained based on a machine learning model, including, but not limited to, fitting of one or more of a convolutional neural network (Convolutional Neural Network, CNN) model, a deep neural network (Deep Neural Network, DNN) model, a recurrent neural network (Recurrent Neural Networks, RNN) model, a pre-trained language model (Bidirectional Encoder Representation from Transformers, BERT), an embedding (empdding) model, a gradient-lifting decision tree (Gradient Boosting Decision Tree, GBDT) model, a logistic regression (Logistic Regression, LR) model, a BERT model, a Roberta model, and the like. And constructing an initial information recommendation model based on the machine learning model, carrying out model training on the initial information recommendation model by adopting a large number of information recommendation samples, and adjusting model parameters by adopting a back propagation learning algorithm in the model training process until the initial information recommendation model meets the model finishing training condition, so as to obtain a trained information recommendation model.

Under the actual interactive dialogue scene, the electronic equipment can input at least one type of reference transaction content information, customer service behavior information and data transfer conversion information corresponding to the dialogue intention of the user as model input into an information recommendation model, and performs information recommendation scoring on various types of reference transaction content information based on the customer service behavior information and the data transfer conversion information through the information recommendation model to obtain content information item scoring of various types of reference transaction content information

The information recommendation score is used for feeding back the fitting degree of the transaction content information and the information recommendation of the current customer service end, and generally, the larger the value of the information recommendation score is, the more the transaction content information fits the customer service end.

And B4, sorting and screening the various types of reference transaction content information based on the content information item scores through the information recommendation model so as to output at least one type of recommended transaction content information aiming at the customer service side.

Alternatively, the ranking filter may be a way of filtering and/or scoring the ranking;

in a possible implementation manner, the information recommendation model may filter the reference transaction content information based on the content information item score, specifically, may filter the reference transaction content information with the information item score smaller than the score threshold, and the model may obtain at least one type of recommended transaction content information after filtering and output the at least one type of recommended transaction content information.

In a possible implementation manner, the information recommendation model may score and sort the various types of the reference transaction content information based on the content information item scores, and adjust each type of the reference transaction content information according to the high-low order of the scoring scores, for example, adjust the display order of each type of the reference transaction content information, so as to obtain at least one type of the recommended transaction content information for the customer service side after being sequenced, and output the at least one type of the recommended transaction content information.

In a possible implementation manner, an initial information recommendation model may be created in advance, the initial information recommendation model is trained, and the model training process may be in the following form:

the method comprises the steps that C2, an electronic device creates an initial information recommendation model, information recommendation data of a plurality of reference customer service terminals aiming at least one reference user intention is obtained, and an information recommendation data sample aiming at the initial recommendation model is built based on the information recommendation data of the reference customer service terminals;

the reference customer service end is a customer service end associated with the target business service, and an information recommendation data sample is constructed by acquiring information recommendation data of different reference customer service ends aiming at different reference user intentions in an actual service scene.

Schematically, the information recommendation data is data of an information recommendation link for recommending information to a user terminal by collecting reference customer service terminals on the basis of combining the determined universal information recommendation contents when facing different reference user intentions, the information recommendation data includes, but is not limited to, data of dimensions such as reference user intention, reference customer service end identification, reference user intention corresponding general information recommendation content (typically, multiple types of communication information recommendation content generated based on reference user intention by adopting a recommendation information matching strategy), adopted information recommendation content, unapproved information recommendation content, reference customer service content recommendation preference, reference customer service communication frame type commonly used by the reference customer service, service style of the reference customer service, service class of the reference customer service, service recommendation result index attribute of the reference customer service (such as exposure, adoption amount and adoption rate in service recommendation), reference customer service work index attribute (such as instant messaging amount, number of outbound calls and outbound call duration of the current customer service), reference customer service data transfer conversion characteristics (such as data transfer conversion index of the user after the reference customer service makes the customer service information recommendation), and the like. The information recommendation data samples can be marked with label recommendation scores aiming at corresponding types of information recommendation contents, and the label recommendation scores are used for a back propagation learning adjustment model parameter process of an initial information recommendation model.

Optionally, in one or more embodiments of the present disclosure, the information recommendation data sample may at least be composed of reference customer service behavior information corresponding to the corresponding reference customer service end (one or more of the above-mentioned multi-dimensional information recommendation data), reference customer service reference data transfer conversion information, and reference user intention corresponding general information recommendation content (may be one or more types of general information recommendation content), and tag recommendation scores are labeled for the information recommendation data sample in combination with recommendation conditions of the reference customer service end in an actual application stage.

In a possible implementation manner, the electronic device executes the information recommendation data sample for constructing the initial recommendation model based on the information recommendation data of the reference customer service side, and specifically may be:

schematically, the electronic device determines first information recommendation data corresponding to a customer service adoption type and second information recommendation data corresponding to a customer service neglect type from the information recommendation data corresponding to the reference user intention by a reference customer service terminal;

the first information recommendation data corresponding to the customer service adoption type can be understood as information recommendation contents adopted by the customer service by referring to the customer service on the basis of multi-class universal information recommendation contents corresponding to the intention of the reference customer service, and the first information recommendation data at least comprises C-class universal information recommendation contents provided that the multi-class universal information recommendation contents corresponding to the intention of the reference customer service are respectively A-class universal information recommendation contents, B-class universal information recommendation contents and C-class universal information recommendation contents.

Optionally, the first information recommendation data may include data of other information recommendation dimensions in addition to the general information recommendation content adopted by the reference customer service end to assist the model in performing information recommendation processing, for example, the data of the other information recommendation dimensions may be information such as reference customer service content recommendation preference, reference customer service work index attribute (such as instant messaging message volume, outbound time length and the like of the current customer service), and reference customer service data transfer and conversion feature.

The second information recommendation data corresponding to the customer service neglect type corresponds to the first information recommendation data, the first information recommendation data corresponding to the customer service neglect type can be understood as information recommendation contents which are not adopted or ignored by the reference customer service on the basis of multi-class universal information recommendation contents corresponding to the reference customer intention, and the second information recommendation data at least comprises A-class universal information recommendation contents and B-class universal information recommendation contents.

Optionally, the second information recommendation data may include data of other information recommendation dimensions besides the general information recommendation content ignored by the reference customer service end to assist the model in performing information recommendation processing, for example, the data of the other information recommendation dimensions may be information such as reference customer service content recommendation preference, reference customer service work index attribute (such as instant messaging information amount, outbound time length, etc. of the current customer service), and reference customer service data transfer and conversion feature.

Further, after determining the first information recommendation data and the second information recommendation data, the electronic device may generate an information recommendation data sample according to a pairing sample format based on the first information recommendation data and the second information recommendation data, where the information recommendation data sample includes positive sample data corresponding to the first information recommendation data, negative sample data corresponding to the second information recommendation data, and a label recommendation score corresponding to the positive sample data/the negative sample data.

It will be appreciated that the first information recommendation data and the second information recommendation data may be used as independent information recommendation data samples, and in one or more embodiments of the present disclosure, the information recommendation data samples are sample-encoded according to a paired sample format, that is, a first proportion of the first information recommendation data is used as positive sample data and a second proportion of the second information recommendation data is used as negative sample data.

The paired sample format may be referred to as a pariwise sample format, and in one or more embodiments of the present disclosure, the paired sample format includes at least a positive sample and a negative sample, and in some embodiments, the first ratio and the second ratio of the positive sample and the negative sample indicated by the paired sample format may be the same or different. And labeling the label recommendation scores for the positive sample and the negative sample respectively according to the recommendation condition of the reference customer service end in the actual application stage, wherein the label recommendation score can be 1 for the positive sample, and 0 for the negative sample.

And C4, training the initial information recommendation model by adopting each information recommendation data sample of the plurality of reference customer service ends to obtain a trained information recommendation model.

Further, after generating an information recommendation data sample based on the first information recommendation data and the second information recommendation data according to a pairing sample format, all or part of customer service ends associated with the target transaction service correspond to a plurality of information recommendation data samples, and the electronic device trains the initial information recommendation model by adopting each information recommendation data sample of a plurality of reference customer service ends to obtain a trained information recommendation model, which specifically may be:

Optionally, the electronic device may input each of the information recommendation data samples of the plurality of reference customer service ends into the initial information recommendation model to perform model training, obtain an information recommendation score for the information recommendation data sample in each round of model training, calculate a model recommendation loss based on the information recommendation score and the label recommendation score by using a hinge loss function, and perform model adjustment on the initial information recommendation model based on the model recommendation loss until the initial information recommendation model meets the model training end condition, thereby obtaining the trained information recommendation model.

Illustratively, multiple rounds of model training can be performed on the initial information recommendation model based on each information recommendation data sample, in each round of model training, the initial information recommendation model performs information screening scoring processing on the current information recommendation data sample to obtain information recommendation scores for the current information recommendation data sample, and when multiple types of recommended information content items exist in the sample, content screening, such as content filtering, content item sorting and the like, can be performed based on the information recommendation scores;

Schematically, in the model training process of the initial information recommendation model, information recommendation scores of information recommendation data samples in each round of model training process are obtained, the information recommendation data samples are marked with mark recommendation scores in advance, model loss is calculated by combining the information recommendation scores and the mark recommendation scores obtained by actual model processing, and then model parameters are adjusted by adopting a back propagation learning algorithm according to the model loss in each round of model training process.

Illustratively, the Loss function of the initial information recommendation model may employ a Hinge Loss function, which may also be referred to as a Hinge Loss function. And inputting information recommendation scores and the mark recommendation scores as functions of a change Loss function, outputting the change Loss as model recommendation Loss, and carrying out model adjustment on the initial information recommendation model based on the model recommendation Loss until the initial information recommendation model meets the model training ending condition to obtain a trained information recommendation model.

Illustratively, the hinge loss function may be of the form:

Loss＝max(0，1-y*Y)

the Loss represents model Loss, Y represents information recommendation scores output by model prediction, and Y represents mark recommendation scores.

It can be understood that the distinction degree of the positive sample and the negative sample in the semantic space is increased through the change Loss, meanwhile, customer service behavior characteristic data and data transfer conversion characteristic data are integrated in the model processing process, customer service adoption conditions and user data transfer conversion conditions can be fully considered, scoring conditions of recommended contents of each message are processed such as sorting and filtering according to the message recommendation model, and a financing engineer realizes personalized service recommendation of customer service and reduces unnecessary exposure for filtering contents with scores lower than a certain threshold.

In one or more embodiments of the present disclosure, the model training end condition may be set based on an actual application situation, for example, the model training end condition may be that the number of training rounds of the model reaches a certain round number threshold, for example, the model training end condition may be that the model recommended loss meets a certain loss threshold, and so on.

S212: and under the interactive dialogue scene, indicating the customer service side to perform dialogue reply processing on the user side based on at least one type of recommended transaction content information.

Specific reference is made to the method steps of other embodiments of the present disclosure, and details are not repeated herein.

The information recommendation apparatus provided in the present specification will be described in detail with reference to fig. 5. Note that, the information recommendation device shown in fig. 5 is used to execute the method of the embodiment shown in fig. 1 to 4 of the present specification, and for convenience of explanation, only the portion relevant to the present specification is shown, and specific technical details are not disclosed, please refer to the embodiment shown in fig. 1 to 4 of the present specification.

Referring to fig. 5, a schematic structural diagram of the information recommendation device of the present specification is shown. The information recommendation device 1 may be implemented as all or part of the user terminal by software, hardware or a combination of both. According to some embodiments, the information recommending apparatus 1 includes a sentence acquisition module 11, a recommendation recall module 12, a content screening module 13, and an information recommending module 14, specifically configured to:

the sentence acquisition module 11 is configured to acquire a user dialogue sentence in an interactive dialogue scene, where the interactive dialogue scene is a dialogue scene corresponding to a customer service end and a user end;

a recommendation recall module 12, configured to determine a user dialogue intent based on the user dialogue sentence, and perform information recommendation recall processing based on the user dialogue intent, so as to obtain at least one type of reference transaction content information for the user dialogue intent;

The content screening module 13 is configured to obtain service feature information for the customer service end, and perform transaction content screening based on the at least one type of reference transaction content information and the service feature information corresponding to the user session intention, so as to obtain at least one type of recommended transaction content information for the customer service end;

and the information recommending module 14 is configured to instruct the customer service side to perform a session reply process on the user side based on at least one type of recommended transaction content information in the interactive session scene.

Optionally, the recommendation recall module 12 includes an intention determining unit 121, where the intention determining unit 121 is configured to:

determining a user dialogue intention through the user dialogue statement; or alternatively, the first and second heat exchangers may be,

and acquiring user behavior information aiming at the user terminal, and determining user dialogue intention through the user dialogue statement and the user behavior information.

Alternatively, as shown in fig. 6, the intention determining unit 121 includes:

a feature extraction unit 1211, configured to input the user behavior information and the user dialogue sentence into an intent recognition model, and extract, through the intent recognition model, a user behavior feature corresponding to the user behavior information and a user dialogue feature corresponding to the user dialogue sentence;

A feature stitching unit 1212, configured to perform feature stitching on the user behavior feature and the user session feature through the intent recognition model to obtain a user high-order feature;

an intention recognition unit 1213, configured to perform potential intention recognition on the user high-level feature through the intention recognition model to output a user dialogue intention.

Optionally, when the interactive session scene is a vanishing session scene, the user behavior information is user vanishing behavior information, and the feature extracting unit 1211 is configured to:

and extracting user degaussing behavior characteristics from the user degaussing behavior information through the intention recognition model, wherein the user degaussing behavior characteristics comprise at least one of transaction object browsing characteristics, transaction object clicking characteristics, object acquired data transfer quantity characteristics, object total account data characteristics and user data profit characteristics.

Optionally, the recommendation recall module 13 includes: a recommendation recall unit 122, as shown in fig. 7, the recommendation recall unit 122 comprises

A type determining subunit 1221, configured to determine a target intent type corresponding to the user dialog intention, and generate at least one type of reference transaction content information based on a recommendation information matching policy corresponding to the target intent type;

The system recall subunit 1222 is configured to cancel pushing the reference transaction content information to the customer service end and perform system recall processing on the reference transaction content information, so as to obtain the at least one type of reference transaction content information after the system recall processing.

Optionally, as shown in fig. 8, the content filtering module 13 includes:

an information determining unit 131, configured to determine customer service behavior information and data transfer conversion information for the customer service end from the service feature information;

the content filtering unit 132 is configured to perform transaction content filtering on at least one type of reference transaction content information corresponding to the user dialogue intent based on the customer service behavior information and the data transfer conversion information, so as to obtain at least one type of recommended transaction content information for the customer service side.

Alternatively, as shown in fig. 9, the content filtering unit 132 is configured to:

the scoring subunit 1321 is configured to input at least one type of reference transaction content information, the customer service behavior information and the data transfer conversion information corresponding to the user dialogue intent to an information recommendation model, and score information recommendation of each type of reference transaction content information based on the customer service behavior information and the data transfer conversion information through the information recommendation model, so as to obtain content information item scores of each type of reference transaction content information;

The sorting and screening subunit 1322 is configured to sort and screen the reference transaction content information of each category based on the content information item score through the information recommendation model to output at least one category of recommended transaction content information for the customer service.

Optionally, the device 1 is further configured to:

creating an initial information recommendation model, acquiring information recommendation data of a plurality of reference customer service terminals aiming at least one reference user intention, and constructing an information recommendation data sample aiming at the initial recommendation model based on the information recommendation data of the reference customer service terminals;

and training the initial information recommendation model by adopting each information recommendation data sample of the plurality of reference customer service ends to obtain a trained information recommendation model.

Optionally, the device 1 is further configured to:

determining first information recommendation data corresponding to a customer service adoption type and second information recommendation data corresponding to a customer service neglect type from the information recommendation data corresponding to the reference user intention by the reference customer service terminal;

and generating an information recommendation data sample based on the first information recommendation data and the second information recommendation data according to a pairing sample format, wherein the information recommendation data sample comprises positive sample data corresponding to the first information recommendation data, negative sample data corresponding to the second information recommendation data and label recommendation scores corresponding to the positive sample data/the negative sample data.

Optionally, the device 1 is further configured to:

inputting each information recommendation data sample of a plurality of reference customer service ends into the initial information recommendation model to perform model training, and acquiring an information recommendation score aiming at the information recommendation data sample in each round of model training;

and calculating model recommendation loss based on the information recommendation score and the marking recommendation score by adopting a hinge loss function, and performing model adjustment on the initial information recommendation model based on the model recommendation loss until the initial information recommendation model meets the model training ending condition to obtain a trained information recommendation model.

Optionally, the information recommendation module 14 is configured to

And sending the at least one type of recommended transaction content information to the customer service side so as to instruct the customer service side to select target transaction content information from the at least one type of recommended transaction content information and perform dialogue reply on the user side based on the target transaction content information.

It should be noted that, when the information recommending apparatus provided in the foregoing embodiment performs the information recommending method, only the division of the foregoing functional modules is used as an example, and in practical application, the foregoing functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the information recommending apparatus provided in the above embodiment and the information recommending method embodiment belong to the same concept, which embody the detailed implementation process in the method embodiment, and are not described herein again.

The foregoing description is provided for the purpose of illustration only and does not represent the advantages or disadvantages of the embodiments.

The present disclosure further provides a computer storage medium, where the computer storage medium may store a plurality of instructions, where the instructions are adapted to be loaded by a processor and execute the information recommendation method according to the embodiment shown in fig. 1 to fig. 4, and the specific execution process may refer to the specific description of the embodiment shown in fig. 1 to fig. 4, which is not repeated herein.

The present disclosure further provides a computer program product, where at least one instruction is stored, where the at least one instruction is loaded by the processor and executed by the processor to perform the information recommendation method according to the embodiment shown in fig. 1 to fig. 4, and the specific execution process may refer to the specific description of the embodiment shown in fig. 1 to fig. 4, which is not repeated herein.

Referring to fig. 10, a block diagram of an electronic device according to an exemplary embodiment of the present disclosure is shown. The electronic device in this specification may include one or more of the following: processor 110, memory 120, input device 130, output device 140, and bus 150. The processor 110, the memory 120, the input device 130, and the output device 140 may be connected by a bus 150.

Processor 110 may include one or more processing cores. The processor 110 utilizes various interfaces and lines to connect various portions of the overall electronic device, perform various functions of the electronic device 100, and process data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 120, and invoking data stored in the memory 120. Alternatively, the processor 110 may be implemented in at least one hardware form of digital signal processing (digital signal processing, DSP), field-programmable gate array (field-programmable gate array, FPGA), programmable logic array (programmable logic Array, PLA). The processor 110 may integrate one or a combination of several of a central processor (central processing unit, CPU), an image processor (graphics processing unit, GPU), and a modem, etc. The CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for being responsible for rendering and drawing of display content; the modem is used to handle wireless communications. It will be appreciated that the modem may not be integrated into the processor 110 and may be implemented solely by a single communication chip.

The memory 120 may include a random access memory (random Access Memory, RAM) or a read-only memory (ROM). Optionally, the memory 120 includes a non-transitory computer readable medium (non-transitory computer-readable storage medium). Memory 120 may be used to store instructions, programs, code, sets of codes, or sets of instructions. The memory 120 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, which may be an Android (Android) system, including an Android system-based deep development system, an IOS system developed by apple corporation, including an IOS system-based deep development system, or other systems, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing various method embodiments described below, and the like. The storage data area may also store data created by the electronic device in use, such as phonebooks, audiovisual data, chat log data, and the like.

Referring to FIG. 11, the memory 120 may be divided into an operating system space in which the operating system runs and a user space in which native and third party applications run. In order to ensure that different third party application programs can achieve better operation effects, the operating system allocates corresponding system resources for the different third party application programs. However, the requirements of different application scenarios in the same third party application program on system resources are different, for example, under the local resource loading scenario, the third party application program has higher requirement on the disk reading speed; in the animation rendering scene, the third party application program has higher requirements on the GPU performance. The operating system and the third party application program are mutually independent, and the operating system often cannot timely sense the current application scene of the third party application program, so that the operating system cannot perform targeted system resource adaptation according to the specific application scene of the third party application program.

In order to enable the operating system to distinguish specific application scenes of the third-party application program, data communication between the third-party application program and the operating system needs to be communicated, so that the operating system can acquire current scene information of the third-party application program at any time, and targeted system resource adaptation is performed based on the current scene.

Taking an operating system as an Android system as an example, as shown in fig. 12, a program and data stored in the memory 120 may be stored in the memory 120 with a Linux kernel layer 320, a system runtime library layer 340, an application framework layer 360 and an application layer 380, where the Linux kernel layer 320, the system runtime library layer 340 and the application framework layer 360 belong to an operating system space, and the application layer 380 belongs to a user space. The Linux kernel layer 320 provides the underlying drivers for various hardware of the electronic device, such as display drivers, audio drivers, camera drivers, bluetooth drivers, wi-Fi drivers, power management, and the like. The system runtime layer 340 provides the main feature support for the Android system through some C/c++ libraries. For example, the SQLite library provides support for databases, the OpenGL/ES library provides support for 3D graphics, the Webkit library provides support for browser kernels, and the like. Also provided in the system runtime library layer 340 is a An Zhuoyun runtime library (Android run) which provides mainly some core libraries that can allow developers to write Android applications using the Java language. The application framework layer 360 provides various APIs that may be used in building applications, which developers can also build their own applications by using, for example, campaign management, window management, view management, notification management, content provider, package management, call management, resource management, location management. At least one application program is running in the application layer 380, and these application programs may be native application programs of the operating system, such as a contact program, a short message program, a clock program, a camera application, etc.; and may also be a third party application developed by a third party developer, such as a game-like application, instant messaging program, photo beautification program, etc.

Taking an operating system as an IOS system as an example, the program and data stored in the memory 120 are shown in fig. 13, the IOS system includes: core operating system layer 420 (Core OS layer), core service layer 440 (Core Services layer), media layer 460 (Media layer), and touchable layer 480 (Cocoa Touch Layer). The core operating system layer 420 includes an operating system kernel, drivers, and underlying program frameworks that provide more hardware-like functionality for use by the program frameworks at the core services layer 440. The core services layer 440 provides system services and/or program frameworks required by the application, such as a Foundation (Foundation) framework, an account framework, an advertisement framework, a data storage framework, a network connection framework, a geographic location framework, a sports framework, and the like. The media layer 460 provides an interface for applications related to audiovisual aspects, such as a graphics-image related interface, an audio technology related interface, a video technology related interface, an audio video transmission technology wireless play (AirPlay) interface, and so forth. The touchable layer 480 provides various commonly used interface-related frameworks for application development, with the touchable layer 480 being responsible for user touch interactions on the electronic device. Such as a local notification service, a remote push service, an advertisement framework, a game tool framework, a message User Interface (UI) framework, a User Interface UIKit framework, a map framework, and so forth.

Among the frameworks illustrated in fig. 13, frameworks related to most applications include, but are not limited to: the infrastructure in core services layer 440 and the UIKit framework in touchable layer 480. The infrastructure provides many basic object classes and data types, providing the most basic system services for all applications, independent of the UI. While the class provided by the UIKit framework is a basic UI class library for creating touch-based user interfaces, iOS applications can provide UIs based on the UIKit framework, so it provides the infrastructure for applications to build user interfaces, draw, process and user interaction events, respond to gestures, and so on.

The manner and principle of implementing data communication between the third party application program and the operating system in the IOS system may refer to the Android system, and this description is not repeated here.

The input device 130 is configured to receive input instructions or data, and the input device 130 includes, but is not limited to, a keyboard, a mouse, a camera, a microphone, or a touch device. The output device 140 is used to output instructions or data, and the output device 140 includes, but is not limited to, a display device, a speaker, and the like. In one example, the input device 130 and the output device 140 may be combined, and the input device 130 and the output device 140 are a touch display screen for receiving a touch operation thereon or thereabout by a user using a finger, a touch pen, or any other suitable object, and displaying a user interface of each application program. Touch display screens are typically provided on the front panel of an electronic device. The touch display screen may be designed as a full screen, a curved screen, or a contoured screen. The touch display screen can also be designed to be a combination of a full screen and a curved screen, and a combination of a special-shaped screen and a curved screen is not limited in this specification.

In addition, those skilled in the art will appreciate that the configuration of the electronic device shown in the above-described figures does not constitute a limitation of the electronic device, and the electronic device may include more or less components than illustrated, or may combine certain components, or may have a different arrangement of components. For example, the electronic device further includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, a wireless fidelity (wireless fidelity, wiFi) module, a power supply, and a bluetooth module, which are not described herein.

In this specification, the execution subject of each step may be the electronic device described above. Optionally, the execution subject of each step is an operating system of the electronic device. The operating system may be an android system, an IOS system, or other operating systems, which is not limited in this specification.

The electronic device of the present specification may further have a display device mounted thereon, and the display device may be various devices capable of realizing a display function, for example: cathode ray tube displays (cathode ray tubedisplay, CR), light-emitting diode displays (light-emitting diode display, LED), electronic ink screens, liquid crystal displays (liquid crystal display, LCD), plasma display panels (plasma display panel, PDP), and the like. A user may utilize a display device on electronic device 101 to view displayed text, images, video, etc. The electronic device may be a smart phone, a tablet computer, a gaming device, an AR (Augmented Reality ) device, an automobile, a data storage device, an audio playing device, a video playing device, a notebook, a desktop computing device, a wearable device such as an electronic watch, electronic glasses, an electronic helmet, an electronic bracelet, an electronic necklace, an electronic article of clothing, etc.

In the electronic device shown in fig. 10, where the electronic device may be a service platform, the processor 110 may be configured to invoke an application program stored in the memory 120 and specifically perform the following operations:

In one embodiment, the processor 110, when executing the determining the user dialogue intent based on the user dialogue statement, specifically executes the following steps:

In one embodiment, the processor 110, when executing the determining the user dialogue intent through the user behavior information and the user dialogue statement, specifically executes the following steps:

inputting the user behavior information and the user dialogue sentences into an intention recognition model, and extracting user behavior characteristics corresponding to the user behavior information and user dialogue characteristics corresponding to the user dialogue sentences through the intention recognition model;

performing feature stitching on the user behavior features and the user dialogue features through the intention recognition model to obtain user high-order features;

and carrying out potential intention recognition on the user high-order characteristics through the intention recognition model so as to output user dialogue intention.

In one embodiment, when the interactive session scene is a vanishing session scene, the user behavior information is user vanishing behavior information, and the processor 110 specifically performs the following steps when executing the extracting, by using the intent recognition model, the user behavior feature corresponding to the user behavior information:

In one embodiment, the processor 110 performs the information recommendation recall process based on the user session intention to obtain at least one type of reference transaction content information for the user session intention, and specifically performs the following steps:

determining a target intention type corresponding to the user dialogue intention, and generating at least one type of reference transaction content information based on a recommendation information matching strategy corresponding to the target intention type;

and canceling pushing the reference transaction content information to the customer service side and carrying out system recall processing on the reference transaction content information to obtain the at least one type of reference transaction content information after the system recall processing.

In one embodiment, the processor 110 performs transaction content filtering based on the at least one type of reference transaction content information corresponding to the user session intention and the service feature information to obtain at least one type of recommended transaction content information for the customer service, and specifically performs the following steps:

Determining customer service behavior information and data transfer conversion information aiming at the customer service end from the service characteristic information;

and carrying out transaction content screening on at least one type of reference transaction content information corresponding to the user dialogue intention based on the customer service behavior information and the data transfer conversion information to obtain at least one type of recommended transaction content information aiming at the customer service terminal.

In one embodiment, the processor 110 performs transaction content filtering on at least one type of reference transaction content information corresponding to the user session intention based on the customer service behavior information and the data transfer conversion information to obtain at least one type of recommended transaction content information for the customer service, and specifically performs the following steps:

inputting at least one type of reference transaction content information, the customer service behavior information and the data transfer conversion information corresponding to the user dialogue intention into an information recommendation model, and carrying out information recommendation scoring on each type of reference transaction content information based on the customer service behavior information and the data transfer conversion information through the information recommendation model to obtain content information item scores of each type of reference transaction content information;

And sorting and screening the reference transaction content information of each category based on the content information item scores through the information recommendation model so as to output at least one category of recommended transaction content information aiming at the customer service side.

In one embodiment, the processor 110, when executing the information recommendation method, specifically executes the steps of:

In one embodiment, the processor 110 constructs an information recommendation data sample for the initial recommendation model based on the information recommendation data of the reference customer service side by performing the following steps:

In one embodiment, the processor 110 performs the training on the initial information recommendation model using the information recommendation data samples of the plurality of reference customer service ends to obtain a trained information recommendation model, and specifically performs the following steps:

In one embodiment, the processor 110 instructs the customer service side to perform a session reply process on the user side based on the at least one type of recommended transaction content information, and specifically performs the following steps:

Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a read-only memory, a random access memory, or the like.

The foregoing disclosure is only illustrative of the preferred embodiments of the present invention and is not to be construed as limiting the scope of the claims, which follow the meaning of the claims of the present invention.

Claims

1. An information recommendation method, the method comprising:

2. The method of claim 1, the determining a user dialog intent based on the user dialog statement, comprising:

3. The method of claim 2, the determining a user dialog intention from the user behavior information and the user dialog statement comprising:

4. The method of claim 3, wherein when the interactive dialog scene is a vanishing dialog scene, the user behavior information is user vanishing behavior information,

the extracting, by the intent recognition model, the user behavior feature corresponding to the user behavior information includes:

5. The method of claim 1, the performing information recommendation recall processing based on the user dialog intention to obtain at least one type of reference transaction content information for the user dialog intention, comprising:

6. The method of claim 1, wherein the performing transaction content screening based on the at least one type of reference transaction content information corresponding to the user dialogue intention and the service feature information to obtain at least one type of recommended transaction content information for the customer service side comprises:

7. The method of claim 6, wherein the performing transaction content screening on at least one type of reference transaction content information corresponding to the user dialogue intent based on the customer service behavior information and the data transfer conversion information to obtain at least one type of recommended transaction content information for the customer service side comprises:

8. The method of claim 7, the method further comprising:

9. The method of claim 8, the constructing an information recommendation data sample for the initial recommendation model based on the information recommendation data of the reference customer service side, comprising:

10. The method of claim 9, wherein the training the initial information recommendation model using the information recommendation data samples of the plurality of reference customer service ends to obtain a trained information recommendation model comprises:

11. The method of claim 1, the instructing the customer service side to perform a dialogue reply process on the user side based on at least one type of recommended transaction content information, including:

12. An information recommendation apparatus, the apparatus comprising:

13. A computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the method steps of any one of claims 1 to 11.

14. A computer program product storing at least one instruction for loading by a processor and performing the method steps of any one of claims 1 to 11.

15. An electronic device, comprising: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the method steps of any of claims 1-11.