WO2021047189A1 - Method and apparatus for interactive session information processing, computer device and storage medium - Google Patents

Abstract

A method and an apparatus for interactive session information processing based on relationship network analysis, a computer device, and a storage medium. The method comprises: obtaining session information sent by a user terminal; performing slot recognition on the session information to recognize slots and slot values in the session information; inputting the slot information in the session information into a trained relationship analysis model to analyze associated speculative slots according to the plurality of slots, and calculating the correlation between each slot and the plurality of speculative slots; extracting the speculative slots having a correlation reaching a threshold, and calculating corresponding speculative slot values of the speculative slots according to the plurality of slot information; determining target slot information of a next interactive node according to the slot information and the speculative slot information; and generating interactive information of the next interactive node according to the target slot information, and pushing the interactive information to the corresponding user terminal. By means of the method, the slot information in the session information can be accurately and effectively analyzed and predicted, thereby effectively improving the interactive session information processing efficiency.

Description

Session information interactive processing method, device, computer equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 9, 2019, the application number is 201910848656.1, and the invention title is "Conversational Information Interaction Processing Method, Device, Computer Equipment, and Storage Medium". The entire content is approved The reference is incorporated in this application.

Technical field

This application relates to the field of computer technology, and in particular to a method, device, computer equipment and storage medium for interactive processing of session information.

Background technique

With the rapid development of computer technology, there have been many human-computer interaction systems based on task-based dialogues, which can effectively assist users in some business. Previously, the rule-based slot filling method was only widely used. By observing the text data in the training expectation, based on linguistic knowledge, manually constructing recognition templates for each slot to form a template set, and then setting the order in which the templates are used after the template set is obtained. Use the template to extract the slot information for the oranges input by the user one by one in order.

The inventor realizes that this traditional method requires a lot of labor costs, and the scope of use is relatively narrow, which makes it difficult to cover multiple situations. There is also a slot filling task method based on a classification model. This method requires slot-based classification for each word, which cannot effectively establish the association between the slot values, and cannot accurately and effectively analyze and predict the slots in the conversation information. Bit information, resulting in low interaction efficiency of the session.

Summary of the invention

Based on this, it is necessary to address the above technical problems and provide a session information interactive processing method, device, and device that can accurately and effectively identify and further analyze and predict the slot information in the session information, thereby effectively improving the efficiency of interactive processing of the session information. Computer equipment and storage media.

A method for processing session information interaction, the method comprising:

Acquiring session information sent by the user terminal, where the session information includes a user ID and a service type;

Perform slot identification on the session information, and identify the slot and the slot value in the session information;

Obtain the trained relationship analysis model according to the business type, input the identified slot and slot value into the relationship analysis model, and calculate the slot and the associated speculative slot based on multiple slot identifications The correlation between multiple speculative slots;

Extracting the speculative slot whose relevance reaches the threshold, and calculating the speculative slot value corresponding to the speculative slot according to the multiple slot values in the session information;

Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

The interaction information of the next interaction node is generated according to the target slot information, and the interaction information is pushed to the user terminal corresponding to the user identifier.

In one of the embodiments, important factors of multiple slots are deployed in the relationship analysis model, and the step of identifying the associated speculative slots according to the multiple slots includes: comparing multiple slots with the relationship analysis model Perform feature extraction of slots and slot values to obtain corresponding slot vectors; calculate the relevance between multiple slot vectors based on the important factors of multiple slots; calculate multiple based on the relevance between multiple slot vectors The correlation between a slot vector and a candidate slot; extract the candidate slot whose correlation reaches a preset threshold, and use the candidate slot as a speculative slot.

In one of the embodiments, the step of calculating the speculative slot value corresponding to the speculative slot according to the plurality of slot values in the session information includes: calculating the speculative slot corresponding to the speculative slot according to the plurality of slots and the slot value Probability distribution values of multiple elements; calculate the confidence levels of multiple elements according to the probability distribution values; if there is no element whose confidence level meets the threshold, use the speculative slot as the target slot of the next node session.

In one of the embodiments, the method further includes: if there is an element whose confidence level meets the threshold, determining the element as the speculative slot value corresponding to the speculative slot; and comparing the speculative slot with the speculative slot The slot value is added to the slot information set of the user identification; the slot information set is matched with the slot definition table of the service type, and the candidate slot is determined according to the matching result; the known slot information is calculated The correlation with the candidate slot is extracted, and the candidate slot whose correlation reaches a preset threshold is extracted, and the candidate slot is used as the target slot of the next interactive node.

In one of the embodiments, before obtaining the trained relationship analysis model, the method further includes: obtaining a plurality of sample data, and dividing the sample data into a training set and a verification set, and the sample data includes a plurality of slot information; Input the training data into a preset network model, train the dependency relationships between multiple slots and the corresponding probability distributions according to the preset network model, and generate an initial relationship analysis model; The initial relationship analysis model is further trained and verified to obtain category probabilities corresponding to multiple verification data; until the number of category probabilities corresponding to the verification data within the preset range reaches a preset threshold, the training is stopped to obtain the required Relationship analysis model.

In one of the embodiments, the method further includes: when the slot information in the slot information set identified by the user meets a preset threshold, obtaining product data corresponding to the service type, the product data including attributes Information; calculate the matching degree between the slot information of the user ID and the attribute information of multiple product data; obtain the product data whose matching degree reaches the matching degree threshold, and push the product data to the user corresponding to the user ID terminal.

A device for processing session information interaction, the device comprising:

A data acquisition module, configured to acquire session information sent by a user terminal, where the session information includes user identification and service type;

A slot identification module, configured to perform slot identification on the session information, and identify the slot and the slot value in the session information;

The slot analysis module obtains the trained relationship analysis model according to the service type, inputs the identified slot and slot value into the relationship analysis model, identifies the associated speculative slot according to multiple slots, and calculates The correlation between the slot and a plurality of speculative slots; extract the speculative slot whose correlation reaches the threshold, and calculate the speculative slot corresponding to the speculative slot according to the multiple slot values in the session information Value; Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

The interactive information sending module is configured to generate the interactive information of the next interactive node according to the target slot information, and push the interactive information to the user terminal corresponding to the user identifier.

In one of the embodiments, the device further includes a product data push module, configured to obtain product data corresponding to the service type when the slot information in the slot information set identified by the user meets a preset threshold, The product data includes attribute information; the matching degree between the slot information of the user identification and the attribute information of a plurality of product data is calculated; the product data whose matching degree reaches the matching degree threshold is obtained, and the product data is pushed to all The user terminal corresponding to the user ID.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the session information interaction processing method provided in any embodiment of the present application when the processor executes the computer program.

A computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, it implements the steps of the session information interactive processing method provided in any one of the embodiments of the present application.

In the foregoing session information interactive processing method, device, computer equipment and storage medium, after the server obtains the session information sent by the user terminal, it performs slot identification on the session information, and identifies the slot and slot value in the session information. The server then obtains the trained relationship analysis model according to the business type of the current session, and analyzes related speculation slots based on the identified slots and slot values through the relationship analysis model, so as to accurately and effectively identify and analyze the associated speculations Slot and speculative slot value. The server can then effectively determine the target slot information of the next interactive node based on the identified slot information and the analyzed speculative slot information, and generate the query information of the next interactive node based on the target slot information, and then query The information is sent to the corresponding user terminal, so that the user terminal can further input corresponding session information according to the query information, so as to effectively conduct interactive query for the user slot information, thereby accurately and effectively pushing accurate push data to the user terminal. The relationship analysis model can accurately and effectively infer and analyze the associated slot information based on the known information, which can effectively save unnecessary dialogue branches, thereby effectively improving the efficiency of conversational information interaction processing.

Description of the drawings

Fig. 1 is an application scenario diagram of a session information interaction processing method in an embodiment;

FIG. 2 is a schematic flowchart of a method for processing session information interaction in an embodiment;

FIG. 3 is a schematic flow chart of the step of calculating the inferred slot value in an embodiment;

FIG. 4 is a schematic flowchart of the step of calculating the inferred slot value in another embodiment;

Figure 5 is a structural block diagram of a session information interaction processing device in an embodiment;

Fig. 6 is an internal structure diagram of a computer device in an embodiment.

detailed description

The session information interaction processing method provided in this application can be applied to the application environment as shown in FIG. 1. Among them, the user terminal 102 communicates with the server 104 via the network. The user terminal 102 may send session information to the server 104. After the server 104 obtains the session information sent by the user terminal, the server 104 performs slot identification on the session information, and identifies the slot and the slot value in the session information. Furthermore, the trained relationship analysis model is obtained according to the business type of the current session, and the related speculative slot is analyzed according to the identified slot and the slot value through the relationship analysis model. Then, the target slot information of the next interactive node is determined according to the identified slot information and the analyzed speculative slot information, and the query information of the next interactive node is generated according to the target slot information, and the query information is sent to the corresponding The user terminal 102 of the user terminal 102 enables the user terminal 102 to further input corresponding session information for interaction according to the query information. The user terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server 104 may be implemented by an independent server or a server cluster composed of multiple servers.

In one embodiment, as shown in FIG. 2, a method for processing session information interaction is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:

Step 202: Obtain the session information sent by the user terminal. The session information includes the user ID and the service type.

When the user performs human-computer interaction through the interactive application of the corresponding user terminal, the session information can be input and sent to the corresponding server, and the server can recognize the session information sent by the user and return corresponding reply information for human-computer interaction. For example, the common intelligent consulting interactive system.

The server may obtain the session information sent by the user terminal. The session information includes the service type and may also include the user identification. The session information may include historical session information and current session information sent by the user terminal, and reply information returned by the server to the user terminal.

Step 204: Perform slot identification on the session information, and identify the slot and the slot value in the session information.

Wherein, the slot may refer to the key information attribute in the session information, for example, it may be the key information that the server needs to obtain; the slot value may refer to the specific content corresponding to the key information attribute in the session information. For example, it can be specific key information expressed by the user. For example, the information attribute corresponding to "sex" can be a slot, and "male" can be the slot value corresponding to the "sex" slot. The server may pre-configure a slot definition table corresponding to the service type, and the slot definition table stores various slot information required by the service type.

In one of the embodiments, the value of the slot may include an enumerable string type, an enumerable numeric type, a continuous numeric type, an enumerable numeric range, and so on. Among them, the dependence of the slot can include dependence on other variables and independent variables. For example, in the financial field, you can simplify the dependency and independent slots. For example, the gender ratio is basically a stable ratio between 0-70 years old. If the ratio is more than 70 years old, the proportion of women will increase significantly. Therefore, the gender ratio can be reduced. Ratio is an independent variable and does not have a specific distribution with age.

After obtaining the session information sent by the user terminal, the server identifies the slot information in the session information. Specifically, the server may collect slot information. A plurality of slot keywords are stored in the slot definition table. The slot keywords correspond to corresponding slot information, and the slot information includes the slot and the corresponding slot value. The server can perform keyword identification on the session information according to the slot definition table, and identify the slot information in the query information. Among them, the server recognizes the current session information and the session information adjacent to the historical session information according to multiple slot keywords, extracts the text information corresponding to the slot that matches the slot keyword in the slot definition table, and extracts The text information is used as the recognized slot and the corresponding slot value.

Further, the server can also obtain corresponding user information according to the user identification, identify the user information and the obtained historical session information and current session information, and identify multiple slots and corresponding slot values. The slot and the corresponding slot value are complete slot information.

Step 206: Obtain the trained relationship analysis model according to the service type, input the identified slot and the value of the slot into the relationship analysis model, and identify the associated speculative slot according to the multiple slots, and calculate the slot and multiple speculatives The correlation between the slots.

Step 208: Extract the speculative slot whose relevance reaches the threshold, and calculate the slot value corresponding to the speculative slot according to the multiple slot values in the session information.

Among them, the server can construct a relationship analysis model in advance, and the relationship analysis model can be an intelligent decision model based on a Bayesian network.

After identifying the slot information in the session information, the server obtains the trained relationship analysis model according to the service type, and inputs the identified slots and slot values into the relationship analysis model. The server then identifies the associated speculative slots according to the multiple slot information through the relationship analysis model, and calculates the relevance between the slots and the multiple speculative slots.

The server further extracts the speculative slot whose relevance reaches the threshold, and calculates the slot value corresponding to the speculative slot according to the multiple slot values in the session information. Specifically, the server analyzes and infers the element probability distribution value of the slot according to the slot value in the session information, and when there is a slot value that satisfies the threshold, it determines the slot value as the slot value of the inferred slot.

Step 210: Determine the target slot information of the next interactive node according to the slot information and the inferred slot information.

Step 212: Generate interactive information of the next interactive node according to the target slot, and push the interactive information to the user terminal corresponding to the user identifier.

After the server analyzes the slot information and the inferred slot information in the session information through the relationship analysis model, it determines the target slot information of the next interactive node according to the slot information and the inferred slot information. Specifically, the to-be-identified slot information corresponding to the service type can be calculated according to the existing slot information and the inferred slot information, and the to-be-identified slot information corresponding to the service type can be determined as the target slot information of the next interactive node.

The server then generates the interaction information of the next interaction node according to the target slot, and pushes the interaction information to the user terminal corresponding to the user identification. In this way, the user terminal can input corresponding session information according to the interactive information, and the server can obtain complete slot information from the session information, so that it can accurately and effectively identify the key slot information in the session information, so as to effectively improve the efficiency of human-computer interaction. .

For example, according to the current session information input by the user, the slots of "gender" and "age" can be extracted, as well as the corresponding slot values of "gender: male" and "age: 0-4 years old". Then through the relationship analysis model, the associated slot "education level" can be analyzed, and the slot value of the "education level" slot can be further analyzed according to the slot value "not attended elementary school", then it can be analyzed Obtain the slot value with higher confidence, and then skip the inquiry information about the "education level" slot. Further analyze the target slot of the next conversation node. The relationship analysis model can accurately and effectively infer and analyze the associated slot information based on the known information, which can effectively save unnecessary dialogue branches, thereby effectively improving the efficiency of conversational information interaction processing.

In the foregoing session information interaction processing method, after the server obtains the session information sent by the user terminal, it performs slot identification on the session information, and identifies the slot and the slot value in the session information. The server then obtains the trained relationship analysis model according to the business type of the current session, and analyzes related speculation slots based on the identified slots and slot values through the relationship analysis model, so as to accurately and effectively identify and analyze the associated speculations Slot and speculative slot value. The server can then effectively determine the target slot information of the next interactive node based on the identified slot information and the analyzed speculative slot information, and generate the query information of the next interactive node based on the target slot information, and then query The information is sent to the corresponding user terminal, so that the user terminal can further input corresponding session information according to the query information, so as to effectively perform interactive query for the user slot information, so as to accurately and effectively push accurate push data to the user terminal. The relationship analysis model can accurately and effectively infer and analyze the associated slot information based on the known information, which can effectively save unnecessary dialogue branches, thereby effectively improving the efficiency of conversational information interaction processing.

In one embodiment, the step of identifying associated speculative slots based on multiple slots includes: performing feature extraction on multiple slots and slot values through a relationship analysis model to obtain corresponding slot vectors; Calculate the relevance between multiple slot vectors; calculate the relevance between multiple slot vectors and candidate slots according to the relevance between multiple slot vectors; extract the relevance to a preset threshold Candidate slots are considered as speculative slots.

Among them, the important factors corresponding to each slot are pre-deployed in the relationship analysis model.

After the server receives the session information sent by the user terminal, it performs slot identification on the session information, and identifies the slot and slot value in the session information. The server then obtains the trained relationship analysis model according to the service type of the current session, and analyzes the relevant speculative slot according to the identified slot and the slot value through the relationship analysis model. Specifically, the server performs feature extraction on the identified multiple slots and corresponding slot values through the relationship analysis model, and extracts the corresponding slot vector.

The server then calculates the correlation between the multiple slot vectors based on the important factors of the multiple slots deployed in the relationship analysis model, and then calculates the multiple slot vectors and candidate slots based on the correlation between the multiple slot vectors The correlation between. The server obtains a candidate slot whose relevance reaches a preset threshold, and uses the candidate slot as a speculative slot. In this way, the server can effectively base on the identified slot information and the analyzed inferred slot information.

For example, based on user information and current session information, the slot types of "sex" and "age" can be extracted, as well as the corresponding slot values of "sex: male" and "age: 15 years old". Among them, age and gender can be two important factors that determine the educational procedure. The relationship analysis model can be used to analyze the associated "education level" slot, and the two candidate slots of "marital status" and "reproductive status" can be directly excluded. The server then obtains the associated slots, and then It can effectively analyze the slot value of the associated slot.

In one embodiment, as shown in FIG. 3, the step of calculating the corresponding speculative slot value of the speculative slot according to multiple slot values in the session information specifically includes the following content:

Step 302: Calculate the probability distribution values of the multiple elements corresponding to the inferred slots according to the multiple slots and the slot values.

Step 304: Calculate the confidence of multiple elements according to the probability distribution value.

In step 306, if there is no element whose confidence level meets the threshold, the inferred slot is used as the target slot of the next node session.

After obtaining the session information sent by the user terminal, the server performs slot identification on the session information, and identifies the slot and the slot value in the session information. The server then obtains the trained relationship analysis model according to the service type of the current session, and analyzes the relevant speculative slot according to the identified slot and the slot value through the relationship analysis model. Specifically, the server performs feature extraction on multiple slots and slot values through the relationship analysis model to obtain corresponding slot vectors, and then calculates the correlation between the multiple slot vectors according to the important factors of the multiple slots. The server calculates the relevance between the multiple slot vectors and the candidate slots according to the relevance between the multiple slot vectors, extracts the preset slots whose relevance reaches the preset threshold, and then extracts the candidate slots As a speculative slot.

After the server extracts the inferred slot based on the existing slot information, it further calculates the inferred slot value corresponding to the inferred slot. Specifically, the server calculates the probability distribution values of multiple elements corresponding to the inferred slot according to the slot value corresponding to the slot. Among them, the slot value can be a numerical interval. The server further calculates the probability confidence of multiple slot values based on the distribution probability. When there is no slot value with the probability confidence meeting the threshold, it means that the slot value corresponding to the speculative slot is unknown, and the user needs to be further prompted to enter the corresponding Answer information. The server can directly use the speculative slot as the target speculative slot of the next interactive node, generate corresponding interactive information according to the target speculative slot, and send the interactive information to the user terminal so that the user can input the interactive information through the user terminal The corresponding answer information is interacted so that the server can push the corresponding push data to the user after obtaining the required slot information.

For example, the relationship analysis model can be a Bayesian network-based model. Let G = (I, E) represent a directed acyclic graph (DAG), where I represents the set of all nodes in the graph, and E represents The set of connected line segments, and let X=(Xi)i∈I be the random variable represented by a certain node i in the directed acyclic graph. If the formula of the joint probability distribution of node X can be:

P(x)=∏ _i∈I p(x _i |x _pa(i) )

Among them, X is called a Bayesian network relative to a directed acyclic graph G, pa(i) represents the "cause" of node i. For any random variable, its joint distribution can be compared with the respective local conditional probability distributions. Multiply and get:

According to the above formula, the joint probability distribution of a Bayesian network can be:

Among them, Xi corresponds to each corresponding "dependent" variable Xj. The difference between the above two expressions lies in the part of conditional probability. In a Bayesian network, if its "dependent" variable is known, some nodes will be conditionally independent from its "dependent" variable, and only those related to the "dependent" variable The node will have conditional probability.

If the number of dependencies of the joint distribution is very rare, the Bayesian function method can save considerable memory capacity. For example, if you want to store 10 variables whose values are all 0 or 1 as a conditional probability table type, an intuitive idea can be seen that a total of 2 ¹⁰ =1024 values must be calculated; but if there are none of these 10 variables If there are more than three related "dependent" variables, the conditional probability table of the Bayesian network only needs to calculate 10*2 ³ =80 values at most. Another advantage of the Bayesian network is that it is easier to know whether each variable is conditionally independent or dependent on the type of local distribution to obtain the joint distribution of all random variables.

For example, if the slot information of "age" and "sex" is known, the correlation between "age" and "sex" and "education level" should be analyzed. The unknown speculative slot and its probability distribution value can be inferred based on the known slot information. For example, E is the education level, A is the age, and G is the gender. Among them, age and gender are two important factors that determine the educational procedure, that is, E=(A, G). E, A, G form a directed acyclic graph.

The formula for calculating the probability distribution value of multiple elements can be:

P(E)=∏ _i∈I p(E _i |A,G)

Among them, E can be education level, A can be age, and G can be gender. A and G can be two independent variables. G can have two values, namely G0 and G1.

As shown in Table 1 below, the probability distribution value corresponding to the "sex" slot can be as follows:

Gender value	Probability	Description
G0	0.5085	male
G1	0.4915	Female

Table 1

As shown in Table 2 below, the "age" slot can have 20 values, namely A0-A19, and the probability distribution value can be as follows:

Age value	Probability	Description
A0	0.0585	0-4 years old
A1	0.0548	5-9 years old
…	…	…
A20	0.000001	>100 years old

Table 2

As shown in Table 3 below, the probability distribution value of the "Education Level" slot can be as follows:

education level	Never went to school	primary school	junior high school	High school	Technical secondary school	Junior college	the University	Postgraduate
To	E0	E1	E2	E3	E4	E5	E6	E7

table 3

The CPD (conditional probability distribution) relationship between the "sex" slot, the "age" slot, and the "education level" slot can be shown in Table 4 below:

To	E0	E1	E2	E3	E4	E5	E6	E7
G0, A0	0.99	0.01	0	0	0	0	0	0
G0, A1	0.04	0.95	0.01	0	0	0	0	0
…	To	To	To	To	To	To	To	To
G1, A19	0.66	0.3	0.03	0.01	0	0	0	0

Table 4

According to the above formula and table data, it can be calculated that the value of E0 is 0.99 under the conditions of known slot information: gender: male, age: 0-4 years old. Among them, the set threshold can be 0.95. Therefore, if the confidence of E0 satisfies the preset threshold, it can be concluded that the value of the slot that satisfies the confidence of the "education level" slot is "has not attended elementary school". The server can skip the question corresponding to this slot and does not need to further send the inquiry information corresponding to the "education level" slot to the user terminal. The server can accurately and effectively infer and analyze the inferred slot and the corresponding inferred slot value through the relationship analysis model based on the known slot information, so as to effectively generate the corresponding interactive information according to the target inferred slot to improve the interaction efficiency.

In one embodiment, as shown in FIG. 4, the step of calculating the corresponding speculative slot value of the speculative slot according to multiple slot values in the session information specifically includes the following content:

Step 402: If there is an element whose confidence level meets the threshold, the element is determined as the speculative slot value corresponding to the speculative slot.

Step 404: Add the speculative slot and the speculative slot value to the slot information set identified by the user.

Step 406: Perform matching according to the slot information set and the slot definition table of the service type, and determine the candidate slot according to the matching result.

Step 408: Calculate the correlation between the known slot information and the candidate slot, extract the candidate slot whose correlation reaches the preset threshold, and use the candidate slot as the target slot of the next interactive node.

After obtaining the session information sent by the user terminal, the server performs slot identification on the session information, and identifies the slot and the slot value in the session information. The server then obtains the trained relationship analysis model according to the service type of the current session, and analyzes the relevant speculative slot according to the identified slot and the slot value through the relationship analysis model.

Specifically, the server performs feature extraction on multiple slots and slot values through the relationship analysis model to obtain corresponding slot vectors, and then calculates the correlation between the multiple slot vectors according to the important factors of the multiple slots. The server calculates the relevance between the multiple slot vectors and the candidate slots according to the relevance between the multiple slot vectors, extracts the preset slots whose relevance reaches the preset threshold, and then extracts the candidate slots As a speculative slot.

After the server extracts the inferred slot based on the existing slot information, it further calculates the inferred slot value corresponding to the inferred slot. Specifically, the server calculates the probability distribution values of multiple elements corresponding to the inferred slot according to the slot value corresponding to the slot. The server further calculates the probability confidence of multiple slot values based on the distribution probability. When there is no slot value with the probability confidence meeting the threshold, it means that the slot value corresponding to the speculative slot is unknown, and the user needs to be further prompted to enter the corresponding Answer information. The server can directly use the speculative slot as the target speculative slot of the next interactive node.

If there is an element whose confidence level meets the threshold, the server determines the element as the speculative slot value corresponding to the speculative slot. Among them, the server may configure a slot definition table corresponding to the service type in advance, and the slot definition table stores various slot information required by the service type. The server can pre-establish a set of slot information corresponding to the user ID. After the server analyzes and obtains the known slot information corresponding to the user ID, it adds the identified slot information and the analyzed speculative slot information to the corresponding user ID. Slot information collection.

The server further matches the slot definition table of the service type according to the slot information set, and determines the candidate slot according to the matching result. For example, the server can match the slot information in the slot definition table of the service type according to the known slot information in the slot information set, identify the remaining unknown slot information in the slot definition table, and replace the unknown slot with the slot information in the slot definition table of the service type. The information is determined as a candidate slot.

The server calculates the correlation between the multiple known slot information and the candidate slot, extracts the candidate slot whose correlation reaches the preset threshold, and uses the candidate slot as the target slot of the next interactive node.

The server can then generate query information for the next interactive node based on the target slot information, and send the query information to the corresponding user terminal, so that the user terminal can further input corresponding session information according to the query information, thereby effectively targeting the user slot information By conducting interactive inquiry, it is possible to accurately and effectively push accurate push data to the user terminal. The relationship analysis model can accurately and effectively infer and analyze the associated slot information based on the known information, which can effectively save unnecessary dialogue branches, thereby effectively improving the efficiency of conversational information interaction processing.

In one embodiment, before acquiring the trained relationship analysis model, the server further includes the step of constructing the relationship analysis model, which specifically includes: acquiring multiple sample data, dividing the sample data into a training set and a validation set, and the sample data includes multiple Slot information; input the training data into the preset network model, train the dependency relationship between multiple slots and the corresponding probability distribution according to the preset network model, and generate the initial relationship analysis model; use the verification set to compare the initial relationship The analysis model is further trained and verified to obtain the category probabilities corresponding to multiple verification data; until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, the training is stopped, and the required relationship analysis model is obtained.

Before the server obtains the preset relationship analysis model, it can also construct and train the relationship analysis model in advance. Specifically, the server may obtain a large amount of sample data from a local database or a third-party database in advance, and generate a training set and a validation set from the large amount of sample data. Among them, the sample data in the training set may be multiple slot information manually labeled, and the verification set contains multiple unlabeled slot information.

The server first performs data cleaning and data preprocessing on the training sample data in the training set to obtain multiple preprocessed slot information. The server inputs multiple slot information into a preset network model, where the preset network model may be a model based on a Bayesian network. The server trains and learns the dependency relationship between the multiple slots according to the initial network model, and the probability distribution interval corresponding to the multiple slot information, and trains to obtain the initial relationship analysis model.

The server further uses multiple slot information in the verification set to further train and verify the generated initial relationship analysis model to obtain category probabilities corresponding to multiple verification data. Until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, the training is stopped, and the required relationship analysis model is obtained. By analyzing and training a large amount of training data, it is possible to effectively analyze the dependencies between multiple slots, and effectively construct a relationship analysis model, which can then accurately and effectively infer and analyze the associated slots based on the known information. information.

In one embodiment, the method further includes: when the slot information in the slot information set identified by the user meets a preset threshold, obtaining product data corresponding to the service type, and the product data includes attribute information; and calculating the slot identified by the user The matching degree between the information and the attribute information of multiple product data; obtaining product data whose matching degree reaches the matching degree threshold, and pushing the product data to the user terminal corresponding to the user identification.

After obtaining the session information sent by the user terminal, the server performs slot identification on the session information, and identifies the slot and the slot value in the session information. The server then obtains the trained relationship analysis model according to the business type of the current session, and analyzes related speculation slots based on the identified slots and slot values through the relationship analysis model, so as to accurately and effectively identify and analyze the associated speculations Slot and speculative slot value. The server can then effectively determine the target slot information of the next interactive node based on the identified slot information and the analyzed speculative slot information, and generate the query information of the next interactive node based on the target slot information, and then query The information is sent to the corresponding user terminal, so that the user terminal further inputs the corresponding session information according to the query information, so that the interactive query for the user slot information can be effectively performed.

During the interaction between the user terminal and the server, the server continuously obtains the slot information corresponding to the user ID and adds it to the slot information collection of the user ID.

When the server detects that the slot information in the slot information set of the user ID meets the preset threshold, it means that the known slot information corresponding to the user ID has met the amount of slot information required by the service type. The server obtains the product data corresponding to the business type, and the product data includes corresponding attribute information. For example, product data may include financial product data, insurance product data, and so on. Then, the corresponding product data is matched according to the slot information of the user identification. Specifically, the server may calculate the matching degree between the slot information of the user ID and the attribute information of multiple product data, obtain product data whose matching degree reaches the matching degree threshold, and push the product data to the user terminal corresponding to the user ID. The relationship analysis model can accurately and effectively infer and analyze the associated slot information based on the known information, which can effectively save unnecessary dialogue branches, thereby effectively improving the efficiency of interactive processing of session information, and accurately pushing matches to users High-degree product data.

In one embodiment, as shown in FIG. 5, a session information interactive processing device is provided, including: a data acquisition module 502, a slot identification module 504, a slot analysis module 506, and an interactive information sending module 508, wherein:

The data acquisition module 502 is configured to acquire session information sent by the user terminal, and the session information includes a user ID and service type;

The slot identification module 504 is used to identify the slot in the session information, and identify the slot and the slot value in the session information;

The slot analysis module 506 obtains the trained relationship analysis model according to the service type, inputs the identified slot and the value of the slot into the relationship analysis model, identifies the associated speculative slot according to multiple slots, and calculates the slot and Relevance between multiple speculative slots; extract speculative slots whose relevance reaches the threshold, and calculate the corresponding speculative slot value of the speculative slot according to the multiple slot values in the session information; according to the slot information and speculative slot The position information determines the target slot information of the next interactive node;

The interactive information sending module 508 is configured to generate the interactive information of the next interactive node according to the target slot information, and push the interactive information to the user terminal corresponding to the user identifier.

In one embodiment, important factors of multiple slots are deployed in the relationship analysis model, and the slot analysis module 506 is also used to extract features of multiple slots and slot values through the relationship analysis model to obtain corresponding slots. Vector; calculate the relevance between multiple slot vectors based on the important factors of multiple slots; calculate the relevance between multiple slot vectors and candidate slots based on the relevance between multiple slot vectors; extract Candidate slots whose relevance reaches the preset threshold are regarded as speculative slots.

In one embodiment, the slot analysis module 506 is further configured to calculate the probability distribution values of multiple elements corresponding to the inferred slots according to the multiple slots and the slot values; calculate the confidence levels of the multiple elements according to the probability distribution values; if There is no element whose confidence level meets the threshold, and the speculative slot is used as the target slot for the next node session.

In one embodiment, the slot analysis module 506 is further configured to determine the element as the speculative slot value corresponding to the speculative slot if there is an element whose confidence level meets the threshold; add the speculative slot and the speculative slot value to the user identification According to the slot information set of the slot information set and the slot definition table of the service type, the candidate slot is determined according to the matching result; the correlation between the known slot information and the candidate slot is calculated, and the correlation is extracted For candidate slots that reach the preset threshold, the candidate slot is used as the target slot of the next interactive node.

In one embodiment, the device further includes a model building module for obtaining a plurality of sample data, dividing the sample data into a training set and a verification set, the sample data includes a plurality of slot information; the training data is input to the preset network In the model, train the dependencies between multiple slots and the corresponding probability distribution according to the preset network model, and generate the initial relationship analysis model; use the verification set to further train and verify the initial relationship analysis model, and obtain multiple verification data Corresponding category probability; until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, stop training to obtain the required relationship analysis model.

In one embodiment, the device further includes a product data push module, configured to obtain product data corresponding to the service type when the slot information in the slot information set identified by the user meets a preset threshold, and the product data includes attribute information; Calculate the matching degree between the slot information of the user ID and the attribute information of multiple product data; obtain the product data whose matching degree reaches the matching degree threshold, and push the product data to the user terminal corresponding to the user ID.

For the specific limitation of the session information interaction processing device, please refer to the above limitation on the session information interaction processing method, which will not be repeated here. Each module in the apparatus for interactive processing of session information may be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. 6. The computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The computer equipment database is used to store data such as session information, slot information, product data, and slot definition tables. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the steps of the session information interaction processing method provided in any embodiment of the present application.

In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile with a computer program stored thereon. When the computer program is executed by a processor, the present application is implemented. Steps of the session information interactive processing method provided in any one of the embodiments.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage. In the medium, when the computer program is executed, it may include the procedures of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Claims (20)

1. A method for processing session information interaction, the method comprising:

   Acquiring session information sent by the user terminal, where the session information includes a user ID and a service type;

   Perform slot identification on the session information, and identify the slot and the slot value in the session information;

   Obtain the trained relationship analysis model according to the business type, input the identified slot and slot value into the relationship analysis model, and calculate the slot and the associated speculative slot based on multiple slot identifications The correlation between multiple speculative slots;

   Extracting the speculative slot whose relevance reaches the threshold, and calculating the speculative slot value corresponding to the speculative slot according to the multiple slot values in the session information;

   Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

   The interaction information of the next interaction node is generated according to the target slot information, and the interaction information is pushed to the user terminal corresponding to the user identifier.
2. The method according to claim 1, wherein important factors of multiple slots are deployed in the relationship analysis model, and the step of identifying the associated speculative slots according to the multiple slots comprises:

   Perform feature extraction on multiple slots and slot values through the relationship analysis model to obtain corresponding slot vectors;

   Calculate the correlation between multiple slot vectors based on the important factors of multiple slots;

   Calculate the correlation between multiple slot vectors and candidate slots according to the correlation between multiple slot vectors;

   The candidate slots whose relevance reaches a preset threshold are extracted, and the candidate slots are used as speculative slots.
3. The method according to claim 1, wherein the step of calculating a speculative slot value corresponding to a speculative slot according to a plurality of slot values in the session information comprises:

   Calculate the probability distribution value of multiple elements corresponding to the inferred slot based on multiple slots and slot values;

   Calculating the confidence levels of multiple elements according to the probability distribution value;

   If there is no element whose confidence level meets the threshold, the speculative slot is used as the target slot of the next node session.
4. The method according to claim 3, further comprising:

   If there is an element whose confidence level meets the threshold, determine the element as the speculative slot value corresponding to the speculative slot;

   Adding the speculative slot and the speculative slot value to the slot information set identified by the user;

   Matching the slot information set with the slot definition table of the service type, and determining candidate slots according to the matching result;

   Calculate the correlation between the known slot information and the candidate slot, extract the candidate slot whose correlation reaches a preset threshold, and use the candidate slot as the target slot of the next interactive node.
5. The method according to claim 1, before obtaining the trained relationship analysis model, further comprising:

   Acquiring a plurality of sample data, dividing the sample data into a training set and a verification set, the sample data including a plurality of slot information;

   Input the training data into a preset network model, train the dependency relationships and corresponding probability distributions among a plurality of slots according to the preset network model, and generate an initial relationship analysis model;

   Further training and verification of the initial relationship analysis model by using the verification set to obtain category probabilities corresponding to multiple verification data;

   Until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, the training is stopped, and the required relationship analysis model is obtained.
6. The method according to any one of claims 1 to 5, further comprising:

   When the slot information in the slot information set identified by the user meets a preset threshold, acquiring product data corresponding to the service type, where the product data includes attribute information;

   Calculating the degree of matching between the slot information of the user identification and the attribute information of the multiple product data;

   Acquire product data whose matching degree reaches a matching degree threshold, and push the product data to the user terminal corresponding to the user identifier.
7. A device for processing session information interaction, the device comprising:

   A data acquisition module, configured to acquire session information sent by a user terminal, where the session information includes user identification and service type;

   A slot identification module, configured to perform slot identification on the session information, and identify the slot and the slot value in the session information;

   The slot analysis module obtains the trained relationship analysis model according to the service type, inputs the identified slot and slot value into the relationship analysis model, identifies the associated speculative slot according to multiple slots, and calculates The correlation between the slot and a plurality of speculative slots; extract the speculative slot whose correlation reaches the threshold, and calculate the speculative slot corresponding to the speculative slot according to the multiple slot values in the session information Value; Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

   The interactive information sending module is configured to generate the interactive information of the next interactive node according to the target slot information, and push the interactive information to the user terminal corresponding to the user identifier.
8. The device according to claim 7, the device further comprising a product data push module, configured to obtain the product corresponding to the service type when the slot information in the slot information set of the user identification meets a preset threshold Data, the product data includes attribute information; the matching degree between the slot information of the user identification and the attribute information of multiple product data is calculated; the product data whose matching degree reaches the matching degree threshold is obtained, and the product data is pushed To the user terminal corresponding to the user identifier.
9. A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements a session information interactive processing method when the computer program is executed:

   Wherein, the session information interaction processing method includes:

   Acquiring session information sent by the user terminal, where the session information includes a user ID and a service type;

   Perform slot identification on the session information, and identify the slot and the slot value in the session information;

   Obtain the trained relationship analysis model according to the business type, input the identified slot and slot value into the relationship analysis model, and calculate the slot and the associated speculative slot based on multiple slot identifications The correlation between multiple speculative slots;

   Extracting the speculative slot whose relevance reaches the threshold, and calculating the speculative slot value corresponding to the speculative slot according to the multiple slot values in the session information;

   Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

   The interaction information of the next interaction node is generated according to the target slot information, and the interaction information is pushed to the user terminal corresponding to the user identifier.
10. The computer device according to claim 9, wherein the important factors of multiple slots are deployed in the relationship analysis model, and the step of identifying the associated speculative slots according to the multiple slots comprises:

    Perform feature extraction on multiple slots and slot values through the relationship analysis model to obtain corresponding slot vectors;

    Calculate the correlation between multiple slot vectors based on the important factors of multiple slots;

    Calculate the correlation between multiple slot vectors and candidate slots according to the correlation between multiple slot vectors;

    The candidate slots whose relevance reaches a preset threshold are extracted, and the candidate slots are used as speculative slots.
11. 8. The computer device according to claim 9, wherein the step of calculating a speculative slot value corresponding to a speculative slot according to a plurality of slot values in the session information comprises:

    Calculate the probability distribution value of multiple elements corresponding to the inferred slot based on multiple slots and slot values;

    Calculating the confidence levels of multiple elements according to the probability distribution value;

    If there is no element whose confidence level meets the threshold, the speculative slot is used as the target slot of the next node session.
12. A computer device according to claim 11, comprising:

    If there is an element whose confidence level meets the threshold, determine the element as the speculative slot value corresponding to the speculative slot;

    Adding the speculative slot and the speculative slot value to the slot information set identified by the user;

    Matching the slot information set with the slot definition table of the service type, and determining candidate slots according to the matching result;

    Calculate the correlation between the known slot information and the candidate slot, extract the candidate slot whose correlation reaches a preset threshold, and use the candidate slot as the target slot of the next interactive node.
13. The computer device according to claim 9, before acquiring the trained relationship analysis model, further comprising:

    Acquiring a plurality of sample data, dividing the sample data into a training set and a verification set, the sample data including a plurality of slot information;

    Input the training data into a preset network model, train the dependency relationships and corresponding probability distributions among a plurality of slots according to the preset network model, and generate an initial relationship analysis model;

    Further training and verification of the initial relationship analysis model by using the verification set to obtain category probabilities corresponding to multiple verification data;

    Until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, the training is stopped, and the required relationship analysis model is obtained.
14. The computer device according to any one of claims 9 to 13, when the slot information in the slot information set of the user identification meets a preset threshold, the product data corresponding to the service type is acquired, and the Product data includes attribute information;

    Calculating the degree of matching between the slot information of the user identification and the attribute information of the multiple product data;

    Acquire product data whose matching degree reaches a matching degree threshold, and push the product data to the user terminal corresponding to the user identifier.
15. A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, a method for interactive processing of session information is realized,

    Wherein, the session information interaction processing method includes the following steps:

    Acquiring session information sent by the user terminal, where the session information includes a user ID and a service type;

    Perform slot identification on the session information, and identify the slot and the slot value in the session information;

    Obtain the trained relationship analysis model according to the business type, input the identified slot and slot value into the relationship analysis model, and calculate the slot and the associated speculative slot based on multiple slot identifications The correlation between multiple speculative slots;

    Extracting the speculative slot whose relevance reaches the threshold, and calculating the speculative slot value corresponding to the speculative slot according to the multiple slot values in the session information;

    Determine the target slot information of the next interactive node according to the slot information and the inferred slot information;

    The interaction information of the next interaction node is generated according to the target slot information, and the interaction information is pushed to the user terminal corresponding to the user identifier.
16. The computer-readable storage medium according to claim 15, wherein important factors of multiple slots are deployed in the relationship analysis model, and the step of identifying the associated speculative slots according to the multiple slots comprises:

    Perform feature extraction on multiple slots and slot values through the relationship analysis model to obtain corresponding slot vectors;

    Calculate the correlation between multiple slot vectors based on the important factors of multiple slots;

    Calculate the correlation between multiple slot vectors and candidate slots according to the correlation between multiple slot vectors;

    The candidate slots whose relevance reaches a preset threshold are extracted, and the candidate slots are used as speculative slots.
17. 15. The computer-readable storage medium according to claim 15, wherein the step of calculating a speculative slot value corresponding to a speculative slot according to a plurality of slot values in the session information comprises:

    Calculate the probability distribution value of multiple elements corresponding to the inferred slot based on multiple slots and slot values;

    Calculating the confidence levels of multiple elements according to the probability distribution value;

    If there is no element whose confidence level meets the threshold, the speculative slot is used as the target slot of the next node session.
18. The computer-readable storage medium according to claim 17, if there is an element whose confidence level meets a threshold, determining the element as the speculative slot value corresponding to the speculative slot;

    Adding the speculative slot and the speculative slot value to the slot information set identified by the user;

    Matching the slot information set with the slot definition table of the service type, and determining candidate slots according to the matching result;

    Calculate the correlation between the known slot information and the candidate slot, extract the candidate slot whose correlation reaches a preset threshold, and use the candidate slot as the target slot of the next interactive node.
19. 15. The computer-readable storage medium according to claim 15, before acquiring the trained relationship analysis model, further comprising:

    Acquiring a plurality of sample data, dividing the sample data into a training set and a verification set, the sample data including a plurality of slot information;

    Input the training data into a preset network model, train the dependency relationships and corresponding probability distributions among a plurality of slots according to the preset network model, and generate an initial relationship analysis model;

    Further training and verification of the initial relationship analysis model by using the verification set to obtain category probabilities corresponding to multiple verification data;

    Until the number of category probabilities corresponding to the verification data within the preset range reaches the preset threshold, the training is stopped, and the required relationship analysis model is obtained.
20. The computer-readable storage medium according to claims 15-19, when the slot information in the slot information set identified by the user meets a preset threshold, the product data corresponding to the service type is acquired, and the product data Including attribute information;

    Calculating the degree of matching between the slot information of the user identification and the attribute information of the multiple product data;

    Acquire product data whose matching degree reaches a matching degree threshold, and push the product data to the user terminal corresponding to the user identifier.

Images