CN116737894A - Intelligent robot service system based on model training - Google Patents

Abstract

The invention relates to the field of artificial intelligence, in particular to an intelligent robot service system based on model training. The system comprises a selection module, a feedback module and a feedback module, wherein the selection module is used for selecting any problem to be fed back from a first database; the input module is used for inputting the problem to be fed back into a language model of a transducer architecture to form a first feedback list; the receiving module is used for receiving the feedback information of the user when the feedback information presents different information characterization sequences, so as to form a second feedback list; the extraction module is used for extracting first feedback information positioned at the first position in the first feedback list, comparing the first feedback information with the service satisfaction degree of the first feedback information in the second feedback list and obtaining a comparison result; and the adjustment module adjusts the adjustment strategy of the language model according to the comparison result so as to enable the first feedback information in the first feedback list to be the optimal feedback information. According to the invention, the language model is adjusted, so that the satisfaction degree of a person can be maximized in the language of the information fed back by the robot system for the problem raised by the user.

Description

Intelligent robot service system based on model training

Technical field

The invention relates to the field of artificial intelligence, and in particular to an intelligent robot service system based on model training.

Background technique

With the rapid development of science and technology, artificial intelligence has penetrated into all walks of life. In some industries, intelligent machines can retrieve and answer simple questions asked by users. Although these machines can answer questions, they still have limitations, especially In some cases, users want to get more warm answers. In order to optimize the answers to artificial intelligence questions and improve the interactivity with people, it is necessary to design an artificial intelligence optimization system.

The patent document with the publication number CN115759123A discloses an intelligent question and answer robot system, which includes an acquisition model, a question and answer model and an output model. The acquisition model is used to obtain questions raised by users; the question and answer model is used to find the answer that best matches the user's query. Matching questions and then giving corresponding answers, the question and answer model adopts two forms: converting words into vectors and word movement distance; the output model is used to receive the corresponding answers given by the question and answer model and then output.

Intelligent question-and-answer robots in the prior art answer questions raised by users by retrieving databases. The feedback information given by the intelligent question-and-answer robots is in a single form, has limitations, and results in poor interactive effects.

Contents of the invention

To this end, the present invention provides an intelligent robot service system based on model training, which can solve the problem of a single form of robot feedback information, limitations and poor interaction effects.

In order to achieve the above object, the present invention provides an intelligent robot service system based on model training, including: a selection module for selecting any question to be fed back from a first database, where a number of questions to be fed back are pre-stored in the first database;

An input module, connected to the selection module, is used to input the question to be fed back into the language model based on the Transformer architecture, and output at least one feedback information based on the question to be fed back, and a plurality of the feedback information forms the first feedback A list, the feedback information includes at least two information representations;

A receiving module, configured to receive user feedback information when the feedback information presents different information representation orders. The user feedback information is used to represent the user's service satisfaction with the feedback information, and is arranged from high to low based on the service satisfaction. Sort the feedback information to form a second feedback list;

An extraction module, connected to the input module, used to extract the first feedback information located first in the first feedback list, and combine the first feedback information with the service of the first feedback information in the second feedback list Compare satisfaction levels and obtain comparison results;

An adjustment module, connected to the extraction module, is used to adjust the adjustment strategy of the language model of the Transformer architecture according to the comparison result, so that the first feedback information in the first feedback list is the optimal feedback information.

Further, the adjustment module includes a joining unit, a calculation unit and an iteration unit;

The adding unit is used to add KL divergence as an item of the loss function in the loss function of the language model of the Transformer architecture;

The calculation unit is used to calculate the KL divergence of the loss function according to the comparison result, where the expression of the KL divergence r _KL is The π ^RL represents the output distribution probability of the first feedback information of the language model of the Transformer architecture, and the π ^SFT represents the output distribution probability of the feedback information after the adjustment module adjusts the language model of the Transformer architecture. ;

The iteration unit is connected to the calculation unit and is used to iterate the first feedback information in the first feedback list into optimal feedback information by minimizing the KL divergence.

Further, the iteration unit is preset with a standard threshold during the iteration process. When obtaining the comparison result, if the service satisfaction between the first feedback information and the first feedback information in the second feedback list is greater than or equal to The standard threshold value is reduced by reducing the KL divergence value multiple times to update the language model of the Transformer architecture multiple times until the KL divergence value is minimum, and the first feedback list is updated The first feedback information outputs the optimal feedback information.

Further, if the service satisfaction of the first feedback information and the first feedback information in the second feedback list is less than the standard threshold, the language model of the Transformer architecture is updated once so that the KL divergence value becomes minimum, the first feedback information in the first feedback list outputs optimal feedback information.

Further, the language model of the Transformer architecture includes: input embedding layer, encoder, decoder and output layer;

The input embedding layer is used to convert the word sequence in the input question to be fed back into a vector representation;

The encoder is used to convert the vector representation into an output vector of the encoder;

The decoder is used to convert the output vector of the encoder into a query vector, and calculate the query vector and the output vector representation of the encoder to obtain the context representation of the problem to be fed back, and obtain the decoder's output vector;

The output layer is used to map the output vector of the decoder into the answer to the feedback question.

Further, the encoder includes an encoder position encoder, an encoder multi-head self-attention mechanism and an encoder forward neural network layer. The working process of the encoder is:

Use the encoder position encoder to encode the position information of the word in each of the questions to be fed back into a vector;

Input the vector into the encoder multi-head self-attention mechanism to establish the relationship between words in the question to be fed back in the input sequence;

The encoder forward neural network is used to process the representation of the words in each of the questions to be fed back in the encoder multi-head self-attention mechanism to obtain the output vector of the encoder.

Further, the decoder includes a decoder position encoder, a decoder multi-head self-attention mechanism, a multi-head attention mechanism and a decoder forward neural network layer. The working process of the decoder is:

Use the decoder position encoder to encode the position information of the word in each of the questions to be fed back into a vector;

Input the vector into the multi-head self-attention mechanism of the decoder, establish the relationship between words in the decoder, and obtain the intermediate layer representation of the decoder;

Input the intermediate layer representation of the decoder into the multi-head attention mechanism;

The decoder forward neural network is used to process the representation of each word in the multi-head attention mechanism to obtain the output vector of the decoder.

The working process of the language model of the Transformer architecture is:

Input the word sequence in the question to be fed back into the input embedding layer and convert it into a vector representation;

The encoder multi-head self-attention mechanism processes the words of the question to be fed back, and calculates the contribution of each question word to be fed back to the question to be fed back based on the current position and semantic relationship of the question word to be fed back. The importance of the sentence is used to obtain the representation of the word to be fed back in the input;

The representation of the question word to be fed back in the input obtained by the multi-head self-attention mechanism of the encoder is encoded through the feed-forward fully connected layer to enhance the semantic expression ability of the representation;

The output vector passing through the encoder multi-head self-attention mechanism and the feedforward fully connected layer is used as input, and the answer to the feedback question is output through the encoder self-attention mechanism and the multi-layer decoder;

A probability distribution of answers to the output feedback question is obtained using a normalized exponential function.

Further, the receiving module collects statistics on user feedback information through big data, gives the user's service satisfaction with the feedback information based on the statistical results, and ranks the feedback information from high to low based on the service satisfaction. Sort.

Further, the minimization of KL divergence is achieved through a gradient optimization algorithm. The gradient optimization algorithm is a method of finding the minimum value of a function by calculating the gradient of the objective function. The method is to continuously adjust the function parameters to continuously reduce the value of the objective function. until a local minimum or global minimum is reached.

Compared with the existing technology, the beneficial effect of the present invention is that the intelligent robot service system based on model training includes a selection module, an input module, a receiving module, an extraction module and an adjustment module. The selection module obtains data from the first database Select any question to be fed back in the first database, which reduces duplication and improves the efficiency of the subsequent work process; the input module will output at least one question from any question to be fed back selected in the first database through a language model based on the Transformer architecture. For feedback information, the language model of the Transformer architecture has parallel computing capabilities, which reduces the consumption of computing resources; the receiving module sorts the feedback information from high to low based on the service satisfaction, improving the scoring consistency of the annotators. property; the extraction module extracts the first feedback information in the first feedback list and compares the service satisfaction of the first feedback information in the second feedback list to obtain the comparison result; through the adjustment module The language model of the Transformer architecture is adjusted so that the language of the information fed back by the robot system to the questions raised by the user can maximize human satisfaction.

Further, the adjustment module includes a joining unit, a calculation unit and an iteration unit. Through the adjustment module, the language model of the Transformer architecture can be continuously updated, so that the answer output by the language model of the Transformer architecture is close to the highest satisfaction level. s answer.

Further, the iteration unit sets a standard threshold, through which the update process of the language model of the Transformer architecture can be judged, the update efficiency of the language model of the Transformer architecture can be improved, and the language model of the Transformer architecture can be implemented. Optimal update results.

Furthermore, the language model of the Transformer architecture has the advantages of unlimited position-related operations, strong modeling capabilities, strong versatility, and strong scalability.

Furthermore, through the multi-head self-attention mechanism, the language model of the Transformer architecture can focus on different parts of the problem to be fed back from different positions and angles, improving the representation ability of the model.

Furthermore, the gradient optimization algorithm can select a reasonable parameter update direction and improve the efficiency of KL divergence minimization.

Description of drawings

Figure 1 is a schematic structural diagram of an intelligent robot service system based on model training provided by an embodiment of the present invention;

Figure 2 is another structural schematic diagram of an intelligent robot service system based on model training provided by an embodiment of the present invention.

Detailed ways

In order to make the purpose and advantages of the present invention more clear, the present invention will be further described below in conjunction with the examples; it should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of the present invention.

It should be noted that in the description of the present invention, the terms "upper", "lower", "left", "right", "inner", "outer" and other terms indicating the direction or positional relationship are based on the figures. The directions or positional relationships shown are only for convenience of description and do not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation of the present invention.

In addition, it should be noted that in the description of the present invention, unless otherwise clearly stated and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense. For example, it can be a fixed connection or a fixed connection. It is a detachable connection or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be an internal connection between two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

Please refer to Figure 1. The present invention provides an intelligent robot service system based on model training, which includes:

The selection module 10 is used to select any question to be fed back from the first database, where a number of questions to be fed back are pre-stored in the first database;

The input module 20 is connected to the selection module 10, and is used to input the question to be fed back into the language model based on the Transformer architecture, and output at least one feedback information based on the question to be fed back. A plurality of the feedback information forms a first A feedback list, the feedback information includes at least two information representations;

The receiving module 30 is used to receive user feedback information when the feedback information presents different information representation orders. The user feedback information is used to represent the user's service satisfaction with the feedback information, and based on the service satisfaction, from high to Sort the feedback information to form a second feedback list;

The extraction module 40 is connected to the input module 20 to extract the first feedback information in the first feedback list, and combine the first feedback information with the first feedback information in the second feedback list. Compare service satisfaction and obtain comparison results;

The adjustment module 50 is connected to the extraction module 40 and is used to adjust the adjustment strategy of the language model of the Transformer architecture according to the comparison result, so that the first feedback information in the first feedback list is the optimal feedback information.

Specifically, the selection module 10 selects any question to be fed back from the first database, and the selected questions reduce duplication and improve the efficiency of the subsequent work process; the input module 20 uses a language model based on the Transformer architecture At least one feedback information will be output from any question to be fed back selected from the first database. The language model of the Transformer architecture has the capability of parallel computing, which reduces the consumption of computing resources; the receiving module 30 is based on the service satisfaction from high to The feedback information is sorted to improve the consistency of the satisfaction level; the extraction module 40 extracts the first feedback information located first in the first feedback list and the first feedback information in the second feedback list Compare the service satisfaction of the feedback information to obtain the comparison results to help subsequent calculations; adjust the language model of the Transformer architecture through the adjustment module 50 so that the robot system can respond to the questions raised by the user in the language of the information. This maximizes people's satisfaction, thereby enabling better optimization of the feedback information output during the human-computer interaction process, and improving the practicality of the feedback information.

Specifically, please refer to Figure 2, the adjustment module 50 includes a joining unit 51, a calculation unit 52 and an iteration unit 53;

The adding unit 51 is used to add KL divergence as an item of the loss function in the loss function of the language model of the Transformer architecture;

The calculation unit 52 is used to calculate the KL divergence of the loss function according to the comparison result, where the expression of the KL divergence r _KL is The π ^RL represents the output distribution probability of the first feedback information of the language model of the Transformer architecture, and the π ^SFT represents the output distribution of the feedback information after the adjustment module 50 adjusts the language model of the Transformer architecture. probability;

The iteration unit 53 is connected to the calculation unit 52 and is used to iterate the first feedback information in the first feedback list into optimal feedback information by minimizing the KL divergence.

Specifically, the adding unit 51 adds KL divergence to the loss function in the language model of the Transformer architecture to help update the language model of the Transformer architecture, so that the language model of the Transformer architecture can be updated with minimal changes. In the calculation of KL divergence, the output calculation results of the first feedback information and the adjusted feedback information of the language model of the Transformer architecture are close to each other.

Specifically, the iteration unit 53 is preset with a standard threshold during the iteration process. When obtaining the comparison result, if the service satisfaction of the first feedback information and the first feedback information in the second feedback list is is greater than or equal to the standard threshold, then the KL divergence value is reduced multiple times to update the language model of the Transformer architecture multiple times, until the KL divergence value is minimum, the first feedback The first feedback information in the list outputs the optimal feedback information.

Specifically, the standard threshold can be used to judge the update process of the language model of the Transformer architecture, improve the update efficiency of the language model of the Transformer architecture, achieve the optimal update result of the language model of the Transformer architecture, and then achieve Effectively optimize feedback information and improve optimization efficiency.

Specifically, if the service satisfaction of the first feedback information and the first feedback information in the second feedback list is less than the standard threshold, the language model of the Transformer architecture is updated once so that the KL divergence value becomes is minimum, the first feedback information in the first feedback list outputs optimal feedback information.

Specifically, when the service satisfaction between the first feedback information and the first feedback information in the second feedback list is less than the standard threshold, the language model update process of the Transformer architecture is reduced, improving the language model update efficiency of the Transformer architecture. .

Specifically, the language model of the Transformer architecture includes: input embedding layer, encoder, decoder and output layer;

The input embedding layer is used to convert the word sequence in the input question to be fed back into a vector representation;

The encoder is used to convert the vector representation into an output vector of the encoder;

The decoder is used to convert the output vector of the encoder into a query vector, and calculate the query vector and the output vector representation of the encoder to obtain the context representation of the problem to be fed back, and obtain the decoder's output vector;

The output layer is used to map the output vector of the decoder into the answer to the feedback question.

Specifically, the structure of the language model of the Transformer architecture has the advantages of fully parallel computing, flexible and scalable structure, good pre-training effect, and processing of multi-modal data.

Specifically, the encoder includes an encoder position encoder, an encoder multi-head self-attention mechanism and an encoder forward neural network layer. The working process of the encoder is:

Use the encoder position encoder to encode the position information of the word in each of the questions to be fed back into a vector;

Input the vector into the encoder multi-head self-attention mechanism to establish the relationship between words in the question to be fed back in the input sequence;

The encoder forward neural network is used to process the representation of the words in each of the questions to be fed back in the encoder multi-head self-attention mechanism to obtain the output vector of the encoder.

Specifically, the encoder adopts the encoder self-attention mechanism, so that the model can simultaneously process the information of all positions in the input sequence in one calculation, improving the parallelism and efficiency of the model. The encoder simultaneously The encoder multi-head self-attention mechanism is used to improve the model's ability to learn the relationship between different positions in the input sequence.

Specifically, the decoder includes a decoder position encoder, a decoder multi-head self-attention mechanism, a multi-head attention mechanism and a decoder forward neural network layer. The working process of the decoder is:

Use the decoder position encoder to encode the position information of the word in each of the questions to be fed back into a vector;

Input the vector into the multi-head self-attention mechanism of the decoder, establish the relationship between words in the decoder, and obtain the intermediate layer representation of the decoder;

Input the intermediate layer representation of the decoder into the multi-head attention mechanism;

The decoder forward neural network is used to process the representation of each word in the multi-head attention mechanism to obtain the output vector of the decoder.

Specifically, the decoder uses the decoder multi-head self-attention mechanism to improve the modeling ability of the model and make it more accurate when processing the input sequence.

Specifically, the working process of the language model of the Transformer architecture is:

Input the word sequence in the question to be fed back into the input embedding layer and convert it into a vector representation;

The encoder multi-head self-attention mechanism processes the words of the question to be fed back, and calculates the contribution of each question word to be fed back to the question to be fed back based on the current position and semantic relationship of the question word to be fed back. The importance of the sentence is used to obtain the representation of the word to be fed back in the input;

The representation of the question word to be fed back in the input obtained by the multi-head self-attention mechanism of the encoder is encoded through the feed-forward fully connected layer to enhance the semantic expression ability of the representation;

The output vector passing through the encoder multi-head self-attention mechanism and the feedforward fully connected layer is used as input, and the answer to the feedback question is output through the encoder self-attention mechanism and the multi-layer decoder;

A probability distribution of answers to the output feedback question is obtained using a normalized exponential function.

Specifically, the language model of the Transformer architecture has strong context awareness capabilities when analyzing text and sequence data, making the model have long-term dependencies in learning text and sequences. The structure of the language model of the Transformer architecture is highly efficient. performance, parallelism and scalability, the model’s inference speed is fast for large-scale data sets.

Specifically, the receiving module 30 counts user feedback information through big data, provides the user's service satisfaction with the feedback information based on the statistical results, and ranks the feedback from high to low based on the service satisfaction. Information is sorted.

Specifically, the user feedback information data requires a large amount, so that the ranking of user satisfaction based on the statistical results of the user feedback information is more accurate.

Specifically, the KL divergence is minimized through a gradient optimization algorithm. The gradient optimization algorithm calculates the gradient of the objective function to find the minimum value of the function. It continuously adjusts the function parameters to continuously reduce the value of the objective function. , until reaching a local minimum or global minimum.

Specifically, the gradient optimization algorithm can select a reasonable parameter update direction and improve the efficiency of KL divergence minimization.

Specifically, the intelligent robot service system based on model training provided by the embodiment of the present invention can be used for natural language understanding and generation of intelligent question and answer robots for credit business, to achieve accurate understanding and rapid response to various questions of users in the field of credit business. Reduce the workload of manual customer service, improve the response speed and accuracy of customer inquiries, and provide users with more professional and convenient financial services.

So far, the technical solution of the present invention has been described with reference to the preferred embodiments shown in the drawings. However, those skilled in the art can easily understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or replacements to relevant technical features, and the technical solutions after these changes or replacements will fall within the protection scope of the present invention.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention; for those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (10)

1. An intelligent robot service system based on model training, comprising:

the selection module is used for selecting any problem to be fed back from a first database, and a plurality of problems to be fed back are prestored in the first database;

the input module is connected with the selection module and used for inputting the to-be-fed-back problem into a language model based on a transducer architecture, outputting at least one piece of feedback information based on the to-be-fed-back problem, wherein a plurality of pieces of feedback information form a first feedback list, and the feedback information comprises at least two information characterizations;

the receiving module is used for receiving user feedback information when the feedback information presents different information characterization sequences, wherein the user feedback information is used for representing service satisfaction degree of a user on the feedback information, and ordering the feedback information from high to low based on the service satisfaction degree to form a second feedback list;

the extraction module is connected with the input module and used for extracting first feedback information positioned at the first position in the first feedback list, comparing the first feedback information with the service satisfaction degree of the first feedback information in the second feedback list and obtaining a comparison result;

the adjusting module is connected with the extracting module and used for adjusting the adjusting strategy of the language model of the transducer architecture according to the comparison result so as to enable the first feedback information in the first feedback list to be the optimal feedback information.

2. The model training-based intelligent robot service system of claim 1, wherein the adjustment module comprises a joining unit, a calculation unit, and an iteration unit;

the adding unit is used for adding KL divergence into a loss function of a language model of the transducer architecture as one item of the loss function;

the calculation unit is used for calculating the KL divergence of the loss function according to the comparison result, wherein the KL divergence r _KL The expression of (2) isSaid pi ^RL Output distribution probability representing the first feedback information of the language model outputting the transducer architecture, the pi ^SFT The output distribution probability of the feedback information after the adjustment module adjusts the language model of the transducer architecture is represented;

the iteration unit is connected with the calculation unit and used for enabling the first feedback information in the first feedback list to be iterated into optimal feedback information by minimizing the KL divergence.

3. The intelligent robot service system based on model training according to claim 2, wherein the iteration unit presets a standard threshold in the process of iteration, and when a comparison result is obtained, if the service satisfaction degree of the first feedback information in the first feedback list and the first feedback information in the second feedback list is greater than or equal to the standard threshold, the KL divergence value is reduced for multiple times, so as to update the language model of the transducer architecture for multiple times, and the first feedback information in the first feedback list is output to the optimal feedback information until the KL divergence value is minimum.

4. The intelligent robot service system based on model training according to claim 3, wherein if the service satisfaction degree of the first feedback information in the first feedback list and the first feedback information in the second feedback list is smaller than the standard threshold value, the language model of the transducer architecture is updated once so that the KL divergence value becomes minimum, and the first feedback information in the first feedback list outputs the optimal feedback information.

5. The model training-based intelligent robot service system of claim 4, wherein the language model of the transducer architecture comprises: an input embedded layer, an encoder, a decoder, and an output layer;

the input embedding layer is used for converting word sequences input into questions to be fed back into vector representations;

the encoder is configured to convert the vector representation into an output vector of the encoder;

the decoder is used for converting the output vector of the encoder into a query vector, and calculating the query vector and the output vector representation of the encoder to obtain the context representation of the problem to be fed back to obtain the output vector of the decoder;

the output layer is used for mapping the output vector of the decoder into the answer of the feedback question.

6. The model training-based intelligent robot service system of claim 5, wherein the encoder comprises an encoder position encoder, an encoder multi-headed self-attention mechanism and an encoder forward neural network layer, the encoder working process is as follows:

encoding position information of words in each of the questions to be fed back into a vector by using the encoder position encoder;

inputting the vector into the multi-head self-attention mechanism of the encoder, and establishing the relation between words in the to-be-fed-back problem in an input sequence;

and using the encoder forward neural network to process the representation of the word in each to-be-fed-back problem in the multi-head self-attention mechanism of the encoder to obtain an output vector of the encoder.

7. The model training based intelligent robotic service system of claim 6, wherein the decoder comprises a decoder position encoder, a decoder multi-headed self-attention mechanism, a multi-headed attention mechanism, and a decoder forward neural network layer, the decoder operating as:

encoding the position information of the words in each of the questions to be fed back into a vector by using the decoder position encoder;

inputting the vector into a multi-head self-attention mechanism of a decoder, and establishing a relation between words in the decoder to obtain an intermediate layer representation of the decoder;

inputting an intermediate layer representation of the decoder into a multi-headed attention mechanism;

the decoder forward neural network is used to process the representation of each word in the multi-headed attention mechanism to obtain the output vector of the decoder.

8. The intelligent robot service system based on model training of claim 7, wherein the language model of the transducer architecture works as follows:

inputting word sequences in a problem to be fed back into the input embedding layer to be converted into vector representations;

the multi-head self-attention mechanism of the encoder processes the words of the problem to be fed back, and calculates the importance degree of each word of the problem to be fed back to the sentence of the problem to be fed back according to the current position and semantic relation of the word of the problem to be fed back, so as to obtain the representation of the word of the problem to be fed back in the input process;

encoding the representation of the to-be-fed-back problem word obtained by the multi-head self-attention mechanism of the encoder in input through the feedforward full-connection layer, and enhancing the semantic expression capability of the representation;

taking output vectors passing through the encoder multi-head self-attention mechanism and the feedforward full-connection layer as input, and outputting answers to feedback questions through the encoder self-attention mechanism and the multi-layer decoder;

and obtaining the probability distribution of the answer of the output feedback problem by using a normalized exponential function.

9. The intelligent robot service system based on model training according to claim 8, wherein the receiving module counts the feedback information of the user through big data, gives the service satisfaction of the user for the feedback information based on the statistics result, and orders the feedback information from high to low based on the service satisfaction.

10. The intelligent robot service system based on model training according to claim 9, wherein the minimized KL divergence is obtained by a gradient optimization algorithm that finds the minimum of the function by calculating the gradient of the objective function by continuously adjusting the function parameters such that the value of the objective function is continuously reduced until a local minimum or global minimum is reached.