阅读说明：本技术 基于神经网络的信息推送方法、系统、设备及介质 (Information pushing method, system, equipment and medium based on neural network ) 是由 舒畅 陈又新 于 2021-09-30 设计创作，主要内容包括：本发明涉及人工智能领域,提供了一种基于神经网络的信息推送方法,所述方法包括：获取目标用户的第一对话信息和目标对象根据所述第一对话信息回复的第二对话信息；根据所述第一对话信息获取第一产品信息,并根据所述第一产品信息判断所述第二对话信息中是否存在所述第一产品信息对应的第二产品信息；如果所述第二对话信息中存在所述第二产品信息,则根据所述第二产品信息确定目标产品信息,并返回与所述目标产品信息对应的目标推送信息；如果所述第二对话信息中不存在所述第二产品信息,则判断所述第二对话信息是否为异议信息；及如果第二对话信息为异议信息,则返回所述第二对话信息对应的异议推送信息。本发明提高了信息推送的效率和准确性。(The invention relates to the field of artificial intelligence, and provides an information pushing method based on a neural network, which comprises the following steps: acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information; acquiring first product information according to the first dialogue information, and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information; if the second product information exists in the second dialogue information, determining target product information according to the second product information, and returning target push information corresponding to the target product information; if the second product information does not exist in the second dialogue information, judging whether the second dialogue information is objection information; and if the second dialogue information is the objection information, returning objection push information corresponding to the second dialogue information. The invention improves the efficiency and accuracy of information push.)

1. An information pushing method based on a neural network is characterized by comprising the following steps:

acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information;

acquiring first product information according to the first dialogue information, and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information;

if the second product information exists in the second dialogue information, determining target product information according to the second product information, and sending target push information corresponding to the target product information to a user terminal associated with the target user;

if the second product information does not exist in the second dialogue information, judging whether the second dialogue information is objection information; and

and if the second session information is objection information, sending objection push information corresponding to the second session information to the user terminal.

2. The information pushing method based on the neural network as claimed in claim 1, wherein the step of obtaining first product information according to the first dialogue information and determining whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information includes:

judging whether the first dialogue information and the second dialogue information are text information or not;

if the first dialogue information and the second dialogue information are not text information, respectively carrying out transcription operation on the first dialogue information and the second dialogue information to obtain first text information and second text information; and

and acquiring first product information according to the first text information, and judging whether second product information corresponding to the first product information exists in the second text information according to the first product information.

3. The information pushing method based on the neural network as claimed in claim 2, wherein the step of obtaining first product information according to the first text information and determining whether second product information corresponding to the first product information exists in the second text information according to the first product information includes:

extracting a plurality of first keywords from the first text information;

determining the first product information from a preset mapping table according to the first keywords;

acquiring a plurality of associated keywords associated with the first product information, wherein each associated keyword corresponds to one piece of second product information;

extracting a plurality of second keywords from the second text information;

respectively calculating the keyword similarity of each second keyword and each associated keyword to obtain a plurality of keyword similarities corresponding to each log keyword; and

and judging whether second product information corresponding to the first product information exists in the second text information according to the similarity of a plurality of keywords corresponding to each second keyword and the similarity of a preset keyword, wherein when at least one keyword in the similarity of the plurality of keywords is greater than the similarity of the preset keyword, the second product information corresponding to the first product information exists in the second text information.

4. The information push method based on the neural network as claimed in claim 3, wherein the step of determining target product information according to the second product information and transmitting target push information corresponding to the target product information to the user terminal associated with the target user if the second product information exists in the second session information comprises:

acquiring a plurality of attributes corresponding to the second product information and a predicted attribute value corresponding to each attribute according to a plurality of attribute classification models configured in advance;

extracting an attribute prediction vector of the second product information according to the plurality of attributes and the plurality of predicted attribute values; and

and according to the attribute prediction vector, obtaining a target vector with the highest similarity with the attribute prediction vector from a database, and determining the target product information according to the target vector.

5. The information pushing method based on the neural network as claimed in claim 1, wherein the step of sending the objection pushing information corresponding to the second session information to the user terminal if the second session information is objection information includes:

extracting an objection vector corresponding to the objection information, and acquiring a target objection vector with the highest acquaintance degree with the objection vector from the database according to the objection vector; and

and determining a target objection according to the target objection vector, and acquiring objection push information from a database according to the target objection.

6. The information pushing method based on the neural network as claimed in claim 1, further comprising:

if the second dialogue information does not contain objection information, inputting the second dialogue information into a link classification model trained in advance so as to output the current dialogue link type of the second dialogue information through the link classification model; and

and acquiring conversation push information corresponding to the second conversation information from a database according to the current conversation link type and the second conversation information, and sending the conversation push information to the user terminal.

7. The information pushing method based on the neural network as claimed in any one of claims 1 to 6, further comprising: and uploading the target push information or the objection push information to a block chain.

8. An information push system based on a neural network is characterized by comprising:

the information acquisition module is used for acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information;

the first judgment module is used for acquiring first product information according to the first dialogue information and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information;

a first sending module, configured to determine, if the second session information includes the second product information, target product information according to the second product information, and send target push information corresponding to the target product information to a user terminal associated with the target user;

a second judging module, configured to judge whether the second session information is objection information if the second product information does not exist in the second session information; and

and a second sending module, configured to send, if the second session information is objection information, objection push information corresponding to the second session information to the user terminal.

9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the neural network-based information pushing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, in which a computer program is stored, and the computer program is executable by at least one processor to cause the at least one processor to execute the steps of the neural network-based information pushing method according to any one of claims 1 to 7.

Technical Field

The embodiment of the invention relates to the field of artificial intelligence, in particular to a method, a system, equipment and a medium for pushing information based on a neural network.

Background

In the online or offline sales process, sales personnel are often required to communicate and introduce commodities to customers, but some sales personnel often have problems that the sales personnel cannot quickly and accurately deal with the customer proposal because of insufficient experience or insufficient comprehensive knowledge of relevant information of products and relevant competitive products of the products, and the like, so that the customer satisfaction is reduced and the sales opportunity is lost. At present, the related art can only provide pre-collected dialogs for a salesperson to solve the problem of insufficient experience of the salesperson, but cannot solve the problems that the salesperson does not know the relevant information of the product comprehensively enough and cannot push the relevant information of competitive products related to the product to the salesperson. Therefore, how to solve the problems that the prior art cannot realize the pushing of the information related to the competitive products to the salespersons, and the information pushing accuracy is low and the effect is poor becomes a technical problem which needs to be solved urgently at present.

Disclosure of Invention

In view of the above, it is necessary to provide an information pushing method, system, device and readable storage medium based on a neural network, so as to solve the problems that the related information of a competitive product cannot be pushed to a salesperson in the prior art, and the information pushing accuracy is low and the effect is poor.

In order to achieve the above object, an embodiment of the present invention provides an information pushing method based on a neural network, where the method includes:

acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information;

acquiring first product information according to the first dialogue information, and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information;

if the second product information exists in the second dialogue information, determining target product information according to the second product information, and sending target push information corresponding to the target product information to a user terminal associated with the target user;

if the second product information does not exist in the second dialogue information, judging whether the second dialogue information is objection information; and

and if the second session information is objection information, sending objection push information corresponding to the second session information to the user terminal.

Optionally, the step of obtaining first product information according to the first session information and determining whether second product information corresponding to the first product information exists in the second session information according to the first product information includes:

judging whether the first dialogue information and the second dialogue information are text information or not;

if the first dialogue information and the second dialogue information are not text information, respectively carrying out transcription operation on the first dialogue information and the second dialogue information to obtain first text information and second text information; and

and acquiring first product information according to the first text information, and judging whether second product information corresponding to the first product information exists in the second text information according to the first product information.

Optionally, the step of obtaining first product information according to the first text information and determining whether second product information corresponding to the first product information exists in the second text information according to the first product information includes:

extracting a plurality of first keywords from the first text information;

determining the first product information from a preset mapping table according to the first keywords;

acquiring a plurality of associated keywords associated with the first product information, wherein each associated keyword corresponds to one piece of second product information;

extracting a plurality of second keywords from the second text information; and

respectively calculating the keyword similarity of each second keyword and each associated keyword to obtain a plurality of keyword similarities corresponding to each log keyword;

and judging whether second product information corresponding to the first product information exists in the second text information according to the similarity of a plurality of keywords corresponding to each second keyword and the similarity of a preset keyword, wherein when at least one keyword in the similarity of the plurality of keywords is greater than the similarity of the preset keyword, the second product information corresponding to the first product information exists in the second text information.

Optionally, if the second session information includes the second product information, determining target product information according to the second product information, and sending target push information corresponding to the target product information to a user terminal associated with the target user, where the step includes:

acquiring a plurality of attributes corresponding to the second product information and a predicted attribute value corresponding to each attribute according to a plurality of attribute classification models configured in advance;

extracting an attribute prediction vector of the second product information according to the plurality of attributes and the plurality of predicted attribute values; and

and according to the attribute prediction vector, obtaining a target vector with the highest similarity with the attribute prediction vector from a database, and determining the target product information according to the target vector.

Optionally, if the second session information is objection information, the step of sending objection push information corresponding to the second session information to the user terminal includes:

extracting an objection vector corresponding to the objection information, and acquiring a target objection vector with the highest acquaintance degree with the objection vector from the database according to the objection vector; and

and determining a target objection according to the target objection vector, and acquiring objection push information from a database according to the target objection.

Optionally, the method further includes:

if the second dialogue information does not contain objection information, inputting the second dialogue information into a link classification model trained in advance so as to output the current dialogue link type of the second dialogue information through the link classification model; and

and acquiring conversation push information corresponding to the second conversation information from a database according to the current conversation link type and the second conversation information, and sending the conversation push information to the user terminal.

Optionally, the method further includes: and testing the target method according to the target calling chain, and uploading the target calling chain to a block chain.

In order to achieve the above object, an embodiment of the present invention further provides an information pushing system based on a neural network, including:

the information acquisition module is used for acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information;

the first judgment module is used for acquiring first product information according to the first dialogue information and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information;

a first sending module, configured to determine, if the second session information includes the second product information, target product information according to the second product information, and send target push information corresponding to the target product information to a user terminal associated with the target user;

a second judging module, configured to judge whether the second session information is objection information if the second product information does not exist in the second session information; and

and a second sending module, configured to send, if the second session information is objection information, objection push information corresponding to the second session information to the user terminal.

In order to achieve the above object, an embodiment of the present invention further provides a computer device, where the computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the computer program implements the steps of the information push method based on the neural network as described above.

To achieve the above object, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, where the computer program is executable by at least one processor, so as to cause the at least one processor to execute the steps of the neural network-based information pushing method as described above.

According to the information pushing method, the information pushing system, the computer device and the computer readable storage medium based on the neural network, whether second product information corresponding to the first product information exists in the second dialogue information is judged, if the second product information exists, corresponding pushing information is returned, the problem that related information of competitive products cannot be pushed to sales staff in the prior art is solved, whether the second dialogue information is objection information is judged, and if the objection information exists, objection pushing information corresponding to the second dialogue information is returned, so that the efficiency and the accuracy of information pushing are improved.

Drawings

Fig. 1 is a schematic flowchart of an information push method based on a neural network according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of program modules of a second embodiment of a neural network-based information pushing system of the present invention;

fig. 3 is a schematic diagram of a hardware structure of a third embodiment of the computer device according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the description relating to "first", "second", etc. in the present invention is for descriptive purposes only and is not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In addition, technical solutions between various embodiments may be combined with each other, but must be realized by a person skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination should not be considered to exist, and is not within the protection scope of the present invention.

Example one

Referring to fig. 1, a flowchart illustrating steps of a neural network-based information pushing method according to an embodiment of the present invention is shown. It is to be understood that the flow charts in the embodiments of the present method are not intended to limit the order in which the steps are performed. The information push system based on the neural network in the present embodiment may be implemented in the computer device 2, and the computer device 2 is taken as an execution subject to be exemplarily described below. The details are as follows.

Step S100, acquiring first dialogue information of a target user and second dialogue information replied by a target object according to the first dialogue information.

The target user can be a salesperson in the product sale process, and the target object can be a customer of the salesperson; in the online or offline sales process, sales staff (target users) are often required to communicate and introduce commodities to customers (target objects), but some sales staff often have problems that the sales staff cannot quickly and accurately deal with the customer and propose due to insufficient experience or incomplete knowledge of relevant information of the products, and the like, so that the customer satisfaction is reduced and the sales opportunity is lost. In view of this, the present embodiment provides a data pushing method based on a neural network, which can be applied to a sales scenario of a target product, so as to help a salesperson to quickly and accurately answer a question posed by a client, where the target product may be various insurance products.

The dialog scene of the target object and the target user can be a face-to-face real person dialog scene, a telephone voice dialog scene, an online character dialog scene and the like. The first dialogue information may be dialogue information of the salesperson, that is, a word spoken by the salesperson, such as a word spoken by the salesperson (target user) when the salesperson (target object) communicates with the customer (target object) to introduce the commodity; the second dialogue information is the words spoken by the customer.

Step S102, acquiring first product information according to the first dialogue information, and judging whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information.

It should be noted that, a general salesperson introduces a product to a customer (target object) at the beginning of a conversation, and in this embodiment, the first product information may be obtained from a product introduced by the salesperson (target user) to the customer (target object), where the first product information is the product information introduced by the salesperson (target user) to the customer (target object).

After obtaining the first product information, the computer device 2 may determine whether second product information corresponding to the first product information exists in the second dialogue information according to the first product information. Wherein the second product information may be competitive product information of the first product; for example, if the product corresponding to the first product information is a insurance, and if the second dialogue information is "i see X insurance of B company yesterday, and feel that is better", then "X insurance" which is a competitive product of "X insurance" and "a insurance" may be the second product information corresponding to the first product information.

In an exemplary embodiment, step S102 may further include step S200 to step S202, wherein: step S200, judging whether the first dialogue information and the second dialogue information are text information; step S202, if the first dialogue information and the second dialogue information are not text information, respectively performing transcription operation on the first dialogue information and the second dialogue information to obtain first text information and second text information; and step S204, acquiring first product information according to the first text information, and judging whether second product information corresponding to the first product information exists in the second text information according to the first product information. In the embodiment, whether the first dialog information and the second dialog information are text information is judged, so that the embodiment can be applied to various dialog scenes, and the application range is improved.

It should be noted that the first dialog information and the second dialog information may be text information, or the first dialog information and the second dialog information may not be text information (for example, may be voice information).

When the first dialog information and the second dialog information are obtained, the computer device 2 may further detect whether the first dialog information and the second dialog information are text information, and if not, the first dialog information and the second dialog information need to be converted into text information.

Illustratively, when the conversation scene is a face-to-face real person conversation scene or a telephone voice conversation scene or other voice conversation scenes, the acquired first conversation information and the acquired second conversation information are voice information, and at this time, voice during conversation needs to be converted into characters. For example, the computer device 2 may convert the first dialog information and the second dialog information into the first text information and the second text information by way of transcription.

In this embodiment, the computer device 2 may perform a transcription operation on the dialog information (the first dialog information or the second dialog information) in the form of voice to obtain a plurality of voice texts; then calculating the matching degree between each voice text and the dialogue information in the voice form; and taking the voice text with the highest matching degree with the voice-form dialogue information in the plurality of voice texts as the word-form dialogue information to obtain the first text information or the second text information. Wherein the calculation between each speech text w and the dialog information in speech formThe matching degree may specifically be a voice text with the highest matching degree with the voice-form dialogue information as the character-form dialogue information

The calculation can be carried out by converting p (w | x) into p (x | w) p (w) through a Bayesian formula. Wherein p (x | w) represents an acoustic model; p (w) represents a language model; p (x) represents the probability of an acoustic feature, the quantity being constant for different w, being calculatedAnd can be ignored. According to the embodiment, the text-form dialogue information with the highest matching degree with the speech-form dialogue information is acquired from the plurality of speech texts, so that the transcription accuracy of speech recognition is improved.

When the dialog scene is a text dialog scene such as an online text dialog scene, the computer device 2 may directly obtain the dialog information (the first text information and the second text information) in a text form.

In an exemplary embodiment, step S204 may further include step S300 to step S310, wherein: step S300, extracting a plurality of first keywords from the first text information; step S302, determining the first product information from a preset mapping table according to the plurality of first keywords; step S304, a plurality of associated keywords associated with the first product information are obtained, wherein each associated keyword corresponds to one piece of second product information; step S306, extracting a plurality of second keywords from the second text information; step S308, respectively calculating the keyword similarity of each second keyword and each associated keyword to obtain a plurality of keyword similarities corresponding to each log keyword; step S310, judging whether second product information corresponding to the first product information exists in the second text information according to the similarity of a plurality of keywords corresponding to each second keyword and the similarity of preset keywords, wherein when at least one keyword in the similarity of the plurality of keywords is larger than the similarity of the preset keywords, the second product information corresponding to the first product information exists in the second text information. In the embodiment, the identification accuracy of the second text information is improved by extracting the plurality of first keywords, the plurality of second keywords and the plurality of associated keywords, and judging whether the second text information corresponding to the first product information exists in the second text information according to the plurality of second keywords and the plurality of associated keywords. It should be noted that, in order to improve the matching range, in this embodiment, the keyword similarity between each second keyword and each associated keyword may also be calculated through a cosine similarity algorithm, and whether the second product information corresponding to the first product information exists in the second text information is determined according to the similarity between a plurality of keywords corresponding to each second keyword and a preset keyword similarity.

In this embodiment, the computer device 2 may input the first text information into a pre-trained neural network model to obtain a first keyword output by the neural network model; the computer device 2 may further input the second text information into the neural network model to obtain a second keyword output by the neural network model; the neural network model may be a bidirectional long-short term memory neural network, and the bidirectional long-short term memory neural network is configured to extract main information from the first text information and the second text information to obtain the first keywords and the second keywords.

The preset mapping table may be an information mapping table pre-constructed according to basic information of a plurality of products, for example, the computer device 2 may pre-acquire basic information of a plurality of products, for example, attribute information, each product may include a plurality of attributes, the computer device 2 may extract a plurality of tags of each product according to the plurality of attributes of the product, so as to establish the information mapping table of the product according to the plurality of tags, and when the plurality of first keywords and the plurality of tags of one product in the preset mapping table are matched, the information of the product is first product information corresponding to the first text information.

And a plurality of competitive products are configured for each product in the preset mapping table in advance, and a plurality of associated keywords are extracted for each product according to the basic information of the plurality of competitive products of each product. When the first product information is obtained, the computer device 2 may obtain a plurality of associated keywords of the first product information from the preset mapping table according to the first product information.

After obtaining the plurality of associated keywords and the plurality of second keywords, the computer device 2 may determine whether a corresponding product exists in the preset mapping table according to the plurality of associated keywords, and if a corresponding product exists, determine whether the product is the second product information according to the plurality of second keywords, that is, determine whether the product is a competitive product of the first product.

Step S104, if the second product information exists in the second dialogue information, determining target product information according to the second product information, and sending target push information corresponding to the target product information to a user terminal associated with the target user.

And if the product information corresponding to the second keywords is matched with any one of the associated keywords, the second product information corresponding to the first product information exists in the second text information.

In an exemplary embodiment, step S104 may further include step S400 to step S404, where: step S400, acquiring a plurality of attributes corresponding to the second product information and a predicted attribute value corresponding to each attribute according to a plurality of attribute classification models configured in advance; step S402, extracting an attribute prediction vector of the second product information according to the attributes and the prediction attribute values; and step S404, acquiring a target vector with the highest similarity with the attribute prediction vector from a database according to the attribute prediction vector, and determining the target product information according to the target vector. In the process of sales, when a salesperson (target user) communicates and introduces a commodity to a customer (target object), a product (first product information) mentioned by the salesperson (target user) is often a proper name of the product, and the computer device 2 may directly recognize the corresponding product, but the product (target object) mentioned by the customer (target user) may not be the proper name of the product, so that when recognizing a product (second product information) mentioned by the customer (target object), the target product information corresponding to the second product information needs to be determined through a plurality of attribute classification models trained in advance to improve the recognition accuracy of the second product information.

It should be noted that each product may include multiple attributes, such as product: attributes of X insurance may include insurance type (e.g., severe insurance), sales region (e.g., Beijing and Tianjin), age bracket (e.g., 70 years), and the like.

Illustratively, the attribute classification models include an insurance type classification model, a selling area classification model, an age group classification model, and the like, wherein a probability value (predicted attribute value) of each attribute of the second product information may be obtained through a corresponding attribute classification model. For example, "x insurance" is input into an insurance type classification model, and output values output by the softmax layer of the insurance type classification model can be used for representing probability distribution of the "x insurance" on different insurance types; the computer device 2 may select the maximum value of the output values output by the softmax layer as the insurance type judgment result of the "x insurance"; for example, if the type corresponding to the maximum value of the output values is "heavy risk", the insurance type of "x insurance" is "heavy risk", and the predicted attribute value corresponding to "heavy risk" is the probability of "heavy risk". The range of output values output by the softmax layer of each attribute classification model is 0-1, and each attribute classification model can be trained through a plurality of labeled data.

After obtaining the plurality of attributes corresponding to the second product information and the predicted attribute values corresponding to the attributes, the computer device 2 may extract an attribute prediction vector of the second product information according to the plurality of attributes and the plurality of predicted attribute values, where the attribute prediction vector may be used to match a target vector with a highest similarity to the attribute prediction vector from a database, and determine the target product information according to the target vector. The computer device 2 may calculate, by a cosine similarity algorithm, a similarity between each pre-configured vector in the database and the attribute prediction vector, and select, as the target vector, a vector having a highest similarity to the attribute prediction vector from among the similarities.

Step S106, if the second product information does not exist in the second dialogue information, whether the second dialogue information is objection information or not is judged.

If the second session information does not include the corresponding second product information, that is, the target object does not refer to the product related to the first product in the session, this embodiment may further determine whether the second session information is objection information, and the objection information may be information such as disagreement and refutation of the session information, where the identification of the objection information may be identified by an objection intention identification model to determine whether the second session information is objection information.

For example, in order to improve the efficiency of recognizing the objection information, the embodiment may further perform objection information recognition on the second session information through a pre-trained objection classification model. The objection classification model can be a two-class classifier, and the objection classification model can be trained by using labeled objection and non-objection data as training samples. The present embodiment may also recognize "i am not empty" as "objection" by using MLP as a classifier.

Step S108, if the second dialogue information is objection information, the objection push information corresponding to the second dialogue information is sent to the user terminal.

In an exemplary embodiment, step S106 may further include step S500 to step S502, where: step S500, extracting an objection vector corresponding to the objection information, and acquiring a target objection vector with the highest acquaintance degree with the objection vector from the database according to the objection vector; and step S502, determining a target objection according to the target objection vector, and acquiring the objection push information from a database according to the target objection. The similarity between each pre-configured vector in a database and the objection vector can be calculated by a cosine similarity algorithm to obtain a target objection vector with the highest degree of dissimilarity with the objection vector, and the vector with the highest degree of similarity with the objection vector is selected from the degrees of dissimilarity to serve as the target objection vector. In the embodiment, the objection pushing information is acquired from the database by extracting the objection vector corresponding to the objection information, so that the objection information matching efficiency and accuracy are improved.

In an exemplary embodiment, the information push method based on the neural network further includes steps S600 to S602, where: step S600, if no objection information exists in the second dialogue information, inputting the second dialogue information into a pre-trained link classification model so as to output the current dialogue link type of the second dialogue information through the link classification model; and step S602, obtaining the dialogue pushing information corresponding to the second dialogue information from a database according to the current dialogue link type and the second dialogue information, and sending the dialogue pushing information to the user terminal. In this embodiment, if the objection information does not exist in the second session information, that is, the target object does not propose objection to the first product, the computer device 2 may obtain, according to the current session link, session push information corresponding to the second session information from a database, where the session push information may be general talk or the like for the current link, and the current session link of the second session information is obtained in this embodiment; and obtaining the dialogue pushing information corresponding to the second dialogue information from a database according to the current dialogue link, so that the accuracy of information pushing is improved. The present embodiment may input the second session information into a link classification model trained in advance, so as to output the current session link type of the second session information through the link classification model, where the link classification model may be obtained by training a softmax multi-class classifier through a large number of training samples, and the training samples of the link classification model may be a session text labeled with link labels in advance, for example, the session text: "hello" may belong to the "open scene" link, i.e., the dialog text: the link label of "hello" may be "open field white" (i.e., the dialog text: "hello" is labeled "open field white"). Specifically, the computer device 2 may encode the second dialog information through a GRU (Gated current Unit Gated loop Unit) neural network, and input the encoding of the second dialog information to the pre-trained link classification model, so as to output the link type of the second dialog information through the pre-trained link classification model.

In some embodiments, if the conversation scene is a face-to-face real person conversation scene or a telephone voice conversation scene, the text information is converted into voice and returned to the earphone associated with the user terminal of the salesperson or the display screen associated with the user terminal. And if the conversation scene is a character conversation scene such as an online character conversation scene, returning the character information to a display screen associated with the user terminal. The embodiment can also greatly improve the success rate of insurance sales personnel, reduce the training and practice time of the sales personnel, and can also obtain real-time high-quality sales call prompts through the assistant even for primary personnel.

In an exemplary embodiment, the information push method based on the neural network further includes step S700: and uploading the target push information or the objection push information to a block chain.

For example, uploading the target push information or the objection push information to the blockchain may ensure security and fair transparency. The blockchain referred to in this example is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm, and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

Example two

Fig. 2 is a schematic diagram of program modules of a second embodiment of the information push system based on a neural network according to the present invention. The information push system 20 based on neural network may include or be divided into one or more program modules, and the one or more program modules are stored in a storage medium and executed by one or more processors to implement the present invention and implement the information push method based on neural network described above. The program module referred to in the embodiments of the present invention refers to a series of computer program instruction segments capable of performing specific functions, and is more suitable for describing the execution process of the information push system 20 based on the neural network in the storage medium than the program itself. The following description will specifically describe the functions of the program modules of the present embodiment:

the information obtaining module 200 is configured to obtain first session information of a target user and second session information replied by a target object according to the first session information.

The first determining module 202 is configured to obtain first product information according to the first dialog information, and determine whether second product information corresponding to the first product information exists in the second dialog information according to the first product information.

A first sending module 204, configured to determine, if the second product information exists in the second dialogue information, target product information according to the second product information, and send target push information corresponding to the target product information to a user terminal associated with the target user.

A second determining module 206, configured to determine whether the second session information is objection information if the second product information does not exist in the second session information.

A second sending module 208, configured to send, if the second session information is objection information, objection push information corresponding to the second session information to the user terminal.

Illustratively, the first determining module 202 is further configured to: judging whether the first dialogue information and the second dialogue information are text information or not; if the first dialogue information and the second dialogue information are not text information, respectively carrying out transcription operation on the first dialogue information and the second dialogue information to obtain first text information and second text information; and acquiring first product information according to the first text information, and judging whether second product information corresponding to the first product information exists in the second text information according to the first product information.

Illustratively, the first determining module 202 is further configured to: extracting a plurality of first keywords from the first text information; determining the first product information from a preset mapping table according to the first keywords; acquiring a plurality of associated keywords associated with the first product information, wherein each associated keyword corresponds to one piece of second product information; extracting a plurality of second keywords from the second text information; respectively calculating the keyword similarity of each second keyword and each associated keyword to obtain a plurality of keyword similarities corresponding to each log keyword; and judging whether second product information corresponding to the first product information exists in the second text information according to the similarity of a plurality of keywords corresponding to each second keyword and the similarity of a preset keyword, wherein when at least one keyword in the similarity of the plurality of keywords is greater than the similarity of the preset keyword, a second product corresponding to the first product information exists in the second text information.

Illustratively, the first sending module 204 is further configured to: acquiring a plurality of attributes corresponding to the second product information and a predicted attribute value corresponding to each attribute according to a plurality of attribute classification models configured in advance; extracting an attribute prediction vector of the second product information according to the plurality of attributes and the plurality of predicted attribute values; and acquiring a target vector with the highest similarity with the attribute prediction vector from a database according to the attribute prediction vector, and determining the target product information according to the target vector.

Illustratively, the second sending module 208 is further configured to: extracting an objection vector corresponding to the objection information, and acquiring a target objection vector with the highest acquaintance degree with the objection vector from the database according to the objection vector; and determining a target objection according to the target objection vector, and acquiring objection push information from a database according to the target objection.

Illustratively, the information push system 20 based on a neural network further includes a third sending module, where the third sending module is configured to, if there is no objection information in the second session information, input the second session information into a link classification model trained in advance, so as to output a current session link type of the second session information through the link classification model; and acquiring dialogue push information corresponding to the second dialogue information from a database according to the current dialogue link type and the second dialogue information, and sending the dialogue push information to the user terminal.

Illustratively, the information push system 20 based on a neural network further includes an uploading module, configured to upload the target push information or the objection push information to a blockchain.

EXAMPLE III

Fig. 3 is a schematic diagram of a hardware architecture of a computer device according to a third embodiment of the present invention. In the present embodiment, the computer device 2 is a device capable of automatically performing numerical calculation and/or information processing in accordance with a command set in advance or stored. The computer device 2 may be a rack server, a blade server, a tower server or a rack server (including an independent server or a server cluster composed of a plurality of servers), and the like. As shown, the computer device 2 includes, but is not limited to, at least a memory 21, a processor 22, a network interface 23, and a neural network-based information push system 20, which are communicatively connected to each other via a system bus.

In this embodiment, the memory 21 includes at least one type of computer-readable storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the storage 21 may be an internal storage unit of the computer device 2, such as a hard disk or a memory of the computer device 2. In other embodiments, the memory 21 may also be an external storage device of the computer device 2, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like provided on the computer device 2. Of course, the memory 21 may also comprise both internal and external memory units of the computer device 2. In this embodiment, the memory 21 is generally used to store an operating system installed in the computer device 2 and various types of application software, such as the program code of the neural network-based information pushing system 20 in the second embodiment. Further, the memory 21 may also be used to temporarily store various types of data that have been output or are to be output.

Processor 22 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 22 is typically used to control the overall operation of the computer device 2. In this embodiment, the processor 22 is configured to execute the program code stored in the memory 21 or process data, for example, execute the neural network based information push system 20, so as to implement the neural network based information push method according to the first embodiment.

The network interface 23 may comprise a wireless network interface or a wired network interface, and the network interface 23 is typically used for establishing a communication connection between the computer device 2 and other electronic apparatuses. For example, the network interface 23 is used to connect the computer device 2 to an external terminal through a network, establish a data transmission channel and a communication connection between the computer device 2 and the external terminal, and the like. The network may be a wireless or wired network such as an Intranet (Intranet), the Internet (Internet), a Global System of Mobile communication i/On (GSM), a Wideband Code Division Multiple Access (WCDMA), a 4G network, a 5G network, Bluetooth (Bluetooth), Wi-Fi, and the like.

It is noted that fig. 3 only shows the computer device 2 with components 20-23, but it is to be understood that not all shown components are required to be implemented, and that more or less components may be implemented instead.

In this embodiment, the information push system 20 based on neural network stored in the memory 21 can be further divided into one or more program modules, and the one or more program modules are stored in the memory 21 and executed by one or more processors (in this embodiment, the processor 22) to complete the present invention.

For example, fig. 2 is a schematic diagram illustrating program modules for implementing the neural network-based information push system 20 according to a second embodiment of the present invention, in which the neural network-based information push system 20 may be divided into an information obtaining module 200, a first determining module 202, a first sending module 204, a second determining module 206, and a second sending module 208. The program module referred to in the present invention refers to a series of computer program instruction segments capable of performing specific functions, and is more suitable than a program for describing the execution process of the neural network based information push system 20 in the computer device 2. The specific functions of the program modules 200 and 208 have been described in detail in the second embodiment, and are not described herein again.

Example four

The present embodiment also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application mall, etc., on which a computer program is stored, which when executed by a processor implements corresponding functions. The computer-readable storage medium information push system 20 based on a neural network of the embodiment is implemented by a processor to implement the information push method based on a neural network of the first embodiment.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.