Artificial intelligence-based sentence analysis method and device, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to an artificial intelligence-based sentence analysis method, which comprises the following steps of: performing conversion processing on acquired composite sentence data to obtain a word sequence; labeling the character sequence to obtain a target character sequence; training the initial sentence analysis model based on the target word sequence to obtain a corresponding training output result; determining the training loss of the initial sentence analysis model based on the training output result and the label data; optimizing the sentence analysis model based on the training loss to obtain a target sentence analysis model; and analyzing the target sentence data based on the target analysis model to generate a corresponding simple sentence result. The invention further provides a sentence analysis device based on artificial intelligence, computer equipment and a storage medium. In addition, the simple sentence result can be stored in the block chain. According to the method, analysis and simplification processing is performed on the composite sentence data based on the target sentence analysis model, so that the processing efficiency and accuracy of analysis and simplification processing are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to a sentence analysis method, apparatus, computer equipment and storage medium based on artificial intelligence. Background Art

[0002] Syntactic analysis is one of the key technologies in natural language processing. Its basic task is to determine the syntactic structure of a sentence or the dependency relationship of words in a sentence. At present, for the financial insurance industry, there are roughly three types of business needs in insurance business scenarios: 1. Change a business to a business; 2. Cancel a business; 3. Handle a business. For users served by insurance companies in insurance business scenarios, the business request sentence data contained in the service requests raised by these users are usually permutations and combinations of the above three types. Each business request sentence data may contain 1 to 3 types, each type can be freely combined, and each type may contain multiple businesses, so that the business request sentence data usually has a more complex content structure. If it is necessary to extract the user's intention from the business request sentence data raised by the user, it is usually necessary to first perform sentence analysis and simplification on the business request sentence data, and then extract the user's intention from the simplified sentence.

[0003] The current processing method used by insurance companies to analyze and simplify the business request sentence data is usually based on rule matching. This method requires manual writing of rules to analyze all possible grammatical structures of the sentence, and the processing efficiency is not high. In addition, there may be many complex structural ambiguities in the business request sentence data that do not conform to common grammatical rules. The processing method based on rule matching often cannot accurately analyze and simplify the business request sentence data, resulting in a low processing accuracy rate. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a sentence analysis method, device, computer equipment and storage medium based on artificial intelligence to solve the problem that the processing method used by existing enterprises is usually a rule-matching-based processing method that requires manual writing of rules to analyze all possible grammatical structures of sentences, and the processing efficiency is low. In addition, there may be many complex structural ambiguities in the business request sentence data that do not conform to common grammatical rules. The rule-matching-based processing method often cannot accurately analyze and simplify the business request sentence data, resulting in a technical problem of low processing accuracy.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a sentence analysis method based on artificial intelligence, which adopts the following technical solution:

[0006] Acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions;

[0007] Converting the compound sentence data to obtain a corresponding word sequence;

[0008] The word sequence is annotated based on a preset annotation method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence;

[0009] Training a pre-constructed initial sentence analysis model based on the target word sequence to obtain a training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field, and a prediction layer;

[0010] Determining a training loss of the initial sentence analysis model based on the training output result and the label data;

[0011] Performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model;

[0012] The target sentence data input by the user is analyzed and processed based on the target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0013] Furthermore, the step of obtaining pre-constructed complex sentence data specifically includes:

[0014] Obtain input initial compound sentence data;

[0015] Performing a replacement process on the initial compound sentence data based on a preset word replacement strategy to obtain corresponding first sentence data;

[0016] Combining the initial compound sentence data based on a preset sentence combination strategy to obtain corresponding second sentence data;

[0017] The compound sentence data is constructed based on the initial compound sentence data, the first sentence data and the second sentence data.

[0018] Furthermore, the step of training the pre-built initial sentence analysis model based on the target word sequence to obtain the training output result of the initial sentence analysis model specifically includes:

[0019] Inputting the target word sequence into the initial sentence analysis model;

[0020] Performing data conversion processing on the target word sequence through the pre-trained language model to obtain a corresponding hidden state vector;

[0021] Performing data mapping processing on the hidden state vector through the linear layer to obtain a corresponding sentence feature matrix;

[0022] The sentence matrix features are processed by the conditional random field to obtain corresponding scoring data;

[0023] Based on the prediction layer, the scoring data is normalized to obtain corresponding prediction probability distribution data;

[0024] The predicted probability distribution data is used as the training output result of the initial sentence analysis model.

[0025] Furthermore, the step of determining the training loss of the initial sentence analysis model based on the training output result and the label data specifically includes:

[0026] Determine the target loss function;

[0027] Based on the target loss function, the training output and the label data are calculated and processed to obtain corresponding difference data;

[0028] The difference data is used as the training loss of the initial sentence parsing model.

[0029] Further, after the step of performing model optimization processing on the sentence analysis model based on the training loss to obtain the corresponding target sentence analysis model, the method further includes:

[0030] Constructing a verification data set based on the compound sentence data;

[0031] Performing performance verification on the target sentence analysis model based on the verification data set;

[0032] If the target sentence analysis model fails the performance verification, obtaining a preset model optimization strategy;

[0033] Optimizing the target sentence analysis model based on the model optimization strategy to obtain an optimized target sentence analysis model;

[0034] The step of analyzing and processing the target sentence data input by the user based on the target analysis model to generate a simple sentence result corresponding to the target sentence data includes:

[0035] The target sentence data input by the user is analyzed and processed based on the optimized target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0036] Further, after the step of performing performance verification on the target sentence analysis model based on the verification data set, the method further includes:

[0037] If the target sentence analysis model passes the performance verification, retaining the model parameters and configuration data of the target sentence analysis model;

[0038] Determining a target storage mode corresponding to the target sentence analysis model;

[0039] The target sentence analysis model is stored based on the target storage method.

[0040] Furthermore, after the step of analyzing and processing the target sentence data input by the user based on the target analysis model to generate a simple sentence result corresponding to the target sentence data, the method further includes:

[0041] Calling pre-built intent recognition models;

[0042] Performing intent recognition on the simple sentence result based on the intent recognition model to obtain a user intent corresponding to the simple sentence result;

[0043] A corresponding response process is executed based on the user intention.

[0044] In order to solve the above technical problems, the embodiment of the present application further provides a sentence analysis device based on artificial intelligence, which adopts the following technical solution:

[0045] A first acquisition module is used to acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions;

[0046] A conversion module, used for converting the compound sentence data to obtain a corresponding word sequence;

[0047] A labeling module, used to label the word sequence based on a preset labeling method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence;

[0048] A training module, used to train a pre-built initial sentence analysis model based on the target word sequence, and obtain the training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field and a prediction layer;

[0049] A first determination module, configured to determine a training loss of the initial sentence analysis model based on the training output result and the label data;

[0050] A first optimization module, configured to perform model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model;

[0051] The analysis module is used to analyze and process the target sentence data input by the user based on the target analysis model, and generate a simple sentence result corresponding to the target sentence data.

[0052] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0053] Acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions;

[0054] Converting the compound sentence data to obtain a corresponding word sequence;

[0055] The word sequence is annotated based on a preset annotation method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence;

[0056] Training a pre-constructed initial sentence analysis model based on the target word sequence to obtain a training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field, and a prediction layer;

[0057] Determining a training loss of the initial sentence analysis model based on the training output result and the label data;

[0058] Performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model;

[0059] The target sentence data input by the user is analyzed and processed based on the target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0060] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0061] Acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions;

[0062] Converting the compound sentence data to obtain a corresponding word sequence;

[0063] The word sequence is annotated based on a preset annotation method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence;

[0064] Training a pre-constructed initial sentence analysis model based on the target word sequence to obtain a training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field, and a prediction layer;

[0065] Determining a training loss of the initial sentence analysis model based on the training output result and the label data;

[0066] Performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model;

[0067] The target sentence data input by the user is analyzed and processed based on the target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0068] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0069] The present application first obtains pre-constructed compound sentence data, and converts the compound sentence data to obtain a corresponding word sequence; then the word sequence is annotated based on a preset annotation method to obtain a corresponding target word sequence; then a pre-constructed initial sentence analysis model is trained based on the target word sequence to obtain a training output result of the initial sentence analysis model; subsequently, the training loss of the initial sentence analysis model is determined based on the training output result and the label data; further, the sentence analysis model is optimized based on the training loss to obtain a corresponding target sentence analysis model; finally, the target sentence data input by the user is analyzed based on the target analysis model to generate a simple sentence result corresponding to the target sentence data. The present application obtains a target word sequence by converting and annotating pre-constructed compound sentence data, and then uses the target word sequence to train and optimize a pre-constructed initial sentence analysis model to obtain a target sentence analysis model. Subsequently, the target analysis model is used to analyze and process the target sentence data input by the user to achieve rapid and accurate generation of simple sentence results corresponding to the target sentence data. Compared with the processing method of analyzing and simplifying sentences based on rule matching only available in the prior art, it does not require manual writing of rules to analyze all possible grammatical structures of the sentence, and can accurately complete the analysis and simplification of compound sentence data with many complex and ambiguous structures that do not conform to common grammatical rules, effectively improving the processing efficiency of analyzing and processing the sentence data and ensuring the accuracy of the obtained simple sentence results. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0071] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0072] Figure 2 A flowchart of an embodiment of a sentence analysis method based on artificial intelligence according to the present application;

[0073] Figure 3 is a structural schematic diagram of an embodiment of an artificial intelligence-based sentence analysis device according to the present application;

[0074] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0075] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0076] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0077] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0078] like Figure 1As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0079] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0080] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0081] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0082] It should be noted that the artificial intelligence-based sentence analysis method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based sentence analysis device is generally set in the server / terminal device.

[0083] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0084] Continue to refer Figure 2, showing a flow chart of an embodiment of an artificial intelligence-based sentence analysis method according to the present application. According to different needs, the order of the steps in the flow chart can be changed, and some steps can be omitted. The artificial intelligence-based sentence analysis method provided in the embodiment of the present application can be applied to any scenario that requires complex sentence analysis, and the artificial intelligence-based sentence analysis method can be applied to products in these scenarios, for example, in the policy revision scenario in the insurance and finance field, the analysis and processing of sentence data that is modified for the policy amount, the insured, the policy time, etc. The artificial intelligence-based sentence analysis method comprises the following steps:

[0085] Step S201, obtaining pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions.

[0086] In this embodiment, the electronic device (e.g. Figure 1 The server / terminal device shown in the figure) can obtain the compound sentence data through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The above-mentioned compound sentence data is sentence data containing multiple intentions. The compound sentence data can be complex sentence data with multiple intentions that modify the policy amount, policyholder, policy time and other contents in the policy revision scenario in the insurance and finance field. Among them, the specific implementation process of obtaining the pre-constructed compound sentence data above will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0087] Step S202: converting the compound sentence data to obtain a corresponding word sequence.

[0088] In this embodiment, the input compound sentence is converted into a word sequence and recorded as: x = (x1, x2, ..., x n ), where x i Represents the id of the i-th word in the sentence in the dictionary.

[0089] Step S203 , annotating the word sequence based on a preset annotation method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence.

[0090] In the present embodiment, the above-mentioned annotation method can specifically adopt a target annotation method that improves the BIO annotation method. The principle of BIO annotation is to mark each character of a word (including the starting character, middle character and end character of the word) as one of the three types of B, I and O. Among them, B means that the character is the starting character of the word, I means that the character is the middle character of the word, and O means that the character is not the starting or middle character of any word. The target annotation method adopted by this application is: based on the BIO annotation method, the label of the first word of the simple sentence in each data is "B-SEN", and the following words are all "I-SEN", until the next simple sentence is encountered, and it is repeated in sequence. Among them, a compound sentence refers to a sentence data with multiple intentions in a data, and a simple sentence refers to a sentence data with only one intention. In addition, in order to avoid the loss of a part of the data, this application will not have a label of "O" in the sentence data. For example, suppose there is a piece of business data "Cancel business, process business, change business A to business B". This sentence contains three intentions, namely 1. Cancel business, 2. Process business, 3. Change business A to business B. The word sequence is annotated according to the above annotation method. The corresponding target word sequence obtained by annotating each simple sentence contained in the business data is as follows:

[0091] Cancel business Handle business Business A Change business B

[0092] B-SEN I-SEN I-SEN I-SEN B-SEN I-SEN I-SENI-SEN B-SEN I-SEN I-SEN I-SEN I-SEN I-SEN I-SEN I-SEN

[0093] Step S204, training a pre-constructed initial sentence analysis model based on the target word sequence to obtain the training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field and a prediction layer.

[0094] In this embodiment, the specific implementation process of training the pre-constructed initial sentence analysis model based on the target word sequence to obtain the training output result of the initial sentence analysis model will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated in detail here.

[0095] Step S205: determining the training loss of the initial sentence analysis model based on the training output result and the label data.

[0096] In this embodiment, the specific implementation process of determining the training loss of the initial sentence analysis model based on the training output results and the label data will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0097] Step S206, performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model.

[0098] In this embodiment, after obtaining the training loss, the model parameters of the sentence analysis model can be reversed according to the training loss. Specifically, the sentence analysis model can be optimized for multiple rounds, and the model training is stopped when the optimization cutoff condition is reached, thereby obtaining the target sentence analysis model. Among them, the setting of the optimization cutoff condition is not specifically limited. For example, the training cutoff condition can be that the number of training iterations reaches a preset number, or that the training loss is lower than a preset loss threshold. The value of the loss threshold can be set according to actual usage requirements. The above-mentioned target sentence analysis model is the sentence analysis model obtained by the final training after the optimization process is completed. The trained target sentence analysis model can autonomously learn the correct grammatical analysis method, and can realize the function of simplifying a complex sentence containing complex intentions into multiple simple sentences.

[0099] Step S207: Analyze and process the target sentence data input by the user based on the target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0100] In this embodiment, the target sentence data input by the user is input into the target analysis model, the target sentence data is analyzed and processed by the target analysis model, and a simple sentence result corresponding to the target sentence data is output. Exemplarily, if the target sentence data input by the user is: "Cancel business, handle business, change business A to business B", then the corresponding simple sentence result corresponding to the target sentence data output by the target analysis model is: the result of 3 simple sentences corresponding to "cancel business", "handle business" and "change business A to business B".

[0101] The present application obtains a target word sequence by converting and annotating pre-constructed compound sentence data, and then uses the target word sequence to train and optimize a pre-constructed initial sentence analysis model to obtain a target sentence analysis model. Subsequently, the target analysis model is used to analyze and process the target sentence data input by the user to achieve rapid and accurate generation of simple sentence results corresponding to the target sentence data. Compared with the processing method of analyzing and simplifying sentences based on rule matching only available in the prior art, it does not require manual writing of rules to analyze all possible grammatical structures of the sentence, and can accurately complete the analysis and simplification of compound sentence data with many complex and ambiguous structures that do not conform to common grammatical rules, effectively improving the processing efficiency of analyzing and processing the sentence data and ensuring the accuracy of the obtained simple sentence results.

[0102] In some optional implementations, step S201 includes the following steps:

[0103] Get the input initial compound sentence data.

[0104] In this embodiment, the above-mentioned initial compound sentence data may be a plurality of compound sentences pre-collected and written by humans according to actual business needs. A compound sentence refers to a sentence data in which there are multiple intentions. A compound sentence is a compound sentence composed of multiple intentions, and one simple sentence corresponds to one intention. Among them, the compound sentence may be sentence data for modifying the policy amount, the insured, the policy time and other contents in the policy modification scenario in the insurance and finance field. Exemplarily, the compound sentence may include: I hope to remove the third-party insurance, add scratch insurance, increase 300,000 for wading, and pedestrian insurance on the car. This sentence data contains 4 simple sentences, corresponding to 4 different types of modification information.

[0105] The initial compound sentence data is replaced based on a preset word replacement strategy to obtain corresponding first sentence data.

[0106] In this embodiment, the above-mentioned word replacement strategy may specifically include a synonym replacement method. For any initial compound sentence data, a keyword can be found from the initial compound sentence data, and then the keyword in the initial compound sentence data is replaced with a synonym corresponding to the keyword, thereby obtaining the first sentence data corresponding to the initial compound sentence data.

[0107] The initial compound sentence data is combined based on a preset sentence combination strategy to obtain corresponding second sentence data.

[0108] In this embodiment, the number of the initial compound sentence data includes multiple ones, and the sentence combination strategy may specifically include combining any two initial compound sentence data included in all the initial compound sentence data to obtain corresponding second sentence data.

[0109] The compound sentence data is constructed based on the initial compound sentence data, the first sentence data and the second sentence data.

[0110] In this embodiment, the initial compound sentence data, the first sentence data, and the second sentence data are integrated to obtain integrated data, and the integrated data is used as the compound sentence data.

[0111] The present application performs data enhancement processing on the input initial compound sentence data by using a word replacement strategy and a sentence combination strategy, thereby constructing the compound sentence data based on the initial compound sentence data, the first sentence data, and the second sentence data, effectively increasing the quantity and diversity of the compound sentence data, which is beneficial for the subsequent use of the constructed compound sentence data to train the initial sentence analysis model, and helps to improve the robustness and generalization ability of the sentence analysis model. This enables the accuracy of the subsequent use of the target analysis model to analyze and process the target sentence data input by the user, and effectively improves the data accuracy of the generated simple sentence results corresponding to the target sentence data.

[0112] In some optional implementations of this embodiment, step S204 includes the following steps:

[0113] The target word sequence is input into the initial sentence analysis model.

[0114] In this embodiment, when the target word sequence is used to train the initial sentence analysis model, the target word sequence is first used as a model input and input into the initial sentence analysis model.

[0115] The target word sequence is subjected to data conversion processing by the pre-trained language model to obtain a corresponding hidden state vector.

[0116] In this embodiment, the above-mentioned pre-trained language model body can adopt the BERT (Bidirectional Encoder Representation from Transformers) model, which is a pre-trained model for language representation. It uses the bidirectional encoder representation of Transformer and emphasizes the generation of a deep bidirectional language representation model. After the target word sequence is input into the BERT model, the BERT model outputs a hidden state vector for each word position in the target word sequence, and these vectors contain context information.

[0117] The hidden state vector is subjected to data mapping processing through the linear layer to obtain a corresponding sentence feature matrix.

[0118] In this embodiment, the linear layer may specifically be a fully connected layer. After the hidden state vector is input into the linear layer, the linear layer maps the hidden state vector from m dimensions to k dimensions, where k is the number of labels in the annotation set (i.e., label data), thereby obtaining automatically extracted sentence features, which are recorded as a matrix P = (p1, p2, p3, ..., p n )∈R n*m , p n It refers to the label probability distribution of the nth position, that is, the probability distribution of the nth word belonging to each label. Through the linear layer, a k-dimensional vector can be output for each position of the target word sequence to represent the probability distribution of the position belonging to each label.

[0119] The sentence matrix features are processed through the conditional random field to obtain corresponding scoring data.

[0120] In this embodiment, the conditional random field is used for sentence-level sequence tagging. The conditional random field ensures that the output label sequence is globally optimal by modeling the dependencies between labels. Specifically, the parameters in the conditional random field are a (k+2)×(k+2) transfer matrix A, A ij It represents the transfer score from the i-th label to the j-th label, so that when labeling a position, the previously labeled label can be used. The conditional random field can calculate the score of all possible label sequences of the entire sentence based on the label probability distribution P and the transfer matrix A output by the linear layer. If a label sequence with a length equal to the length of the sentence is recorded as y = (y1, y2, y3, ..., y n ), then the initial sentence analysis model scores the sentence x’s label equal to y as:

[0121]

[0122] in, refers to the i-th row, y-th row in the matrix P i The elements of the column indicate that the i-th element belongs to label y i probability. is the transition matrix A from label y i-1 Move to label y i score.

[0123] The scoring data is normalized based on the prediction layer to obtain corresponding prediction probability distribution data.

[0124] In this embodiment, the above prediction layer can specifically adopt a Softmax layer. By inputting the scoring data into the Softmax layer, the scoring data is normalized by the Softmax layer, so that the prediction probability distribution data corresponding to the data is obtained, and the specific formula is:

[0125]

[0126] Among them, y ′ Represents all possible label sequences, i.e., training outputs.

[0127] The predicted probability distribution data is used as the training output result of the initial sentence analysis model.

[0128] After inputting the target word sequence into the initial sentence analysis model, the present application processes the target word sequence by using the pre-trained language model, linear layer, conditional random field and prediction layer included in the initial sentence analysis model, thereby quickly and accurately generating the training output results of the initial sentence analysis model, thereby ensuring the accuracy of the obtained training output results.

[0129] In some optional implementations, step S205 includes the following steps:

[0130] Determine the target loss function.

[0131] In this embodiment, there is no specific limitation on the requirements of the above-mentioned target loss function, and specifically, any one of the negative log-likelihood loss function and the cross entropy loss function can be used.

[0132] The training output and the label data are calculated and processed based on the target loss function to obtain corresponding difference data.

[0133] In this embodiment, the training loss is an indicator used in machine learning to measure the difference between the model prediction result and the true label. The training loss can be calculated by a loss function. Specifically, the corresponding loss value, i.e., the difference data, can be obtained by calculating and processing the training output and the label data using the target loss function.

[0134] The difference data is used as the training loss of the initial sentence parsing model.

[0135] In this embodiment, during the training process of the initial sentence analysis model, the initial sentence analysis model minimizes the training loss obtained by continuously adjusting parameters, thereby effectively improving the model performance.

[0136] The present application determines a target loss function; then calculates and processes the training output and the label data based on the target loss function to obtain corresponding difference data; and subsequently uses the difference data as the training loss of the initial sentence analysis model. The present application calculates and processes the training output and the label data based on the target loss function, so that the training loss of the initial sentence analysis model can be calculated quickly and accurately, ensuring the data accuracy of the obtained training loss.

[0137] In some optional implementations, after step S206, the electronic device may further perform the following steps:

[0138] A verification data set is constructed based on the compound sentence data.

[0139] In this embodiment, a specific proportion of data can be randomly selected from the compound sentence data and used as the verification data set. The value of the specific proportion is not specifically limited and can be determined according to actual usage requirements, for example, it can be set to 0.2.

[0140] The performance of the target sentence analysis model is verified based on the verification data set.

[0141] In this embodiment, the process of performing performance verification on the target sentence analysis model based on the verification data set may include: inputting the verification data set into the target sentence analysis model, analyzing and processing the verification data set through the target sentence analysis model, and then performing performance verification processing on the target sentence analysis model according to the target simple sentence result output by the target sentence analysis model, that is, performing performance evaluation on the target sentence analysis model to obtain the performance index value of the target sentence analysis model. Among them, the above-mentioned performance index value may include the specific index values ​​corresponding to the calculated accuracy, recall rate and other indicators used to evaluate the model performance of the target sentence analysis model.

[0142] If the target sentence analysis model fails the performance verification, a preset model optimization strategy is obtained.

[0143] In this embodiment, the performance indicator value obtained by performance verification of the target sentence analysis model can be compared with the preset indicator threshold. If all performance indicator values ​​are greater than the corresponding indicator threshold, it indicates that the target sentence analysis model is a qualified model that meets the performance requirements of the constructed model, and the target sentence analysis model will be determined to have passed the performance verification. If at least one performance indicator value among all performance indicator values ​​is less than the corresponding indicator threshold, it indicates that the target sentence analysis model is an unqualified model that does not meet the performance requirements of the constructed model, and the target sentence analysis model will be determined to have failed the performance verification. Among them, the above-mentioned model optimization strategy may include at least one or more of the strategies such as hyperparameter adjustment, replacement of model layer structure, modification of model architecture or data enhancement strategy.

[0144] The target sentence analysis model is optimized based on the model optimization strategy to obtain an optimized target sentence analysis model.

[0145] In this embodiment, the target sentence analysis model can be optimized according to the policy content in the model optimization policy until a model that meets the model performance requirements is obtained and used as the above-mentioned target sentence analysis model.

[0146] Step S207 includes:

[0147] The target sentence data input by the user is analyzed and processed based on the optimized target analysis model to generate a simple sentence result corresponding to the target sentence data.

[0148] In this embodiment, the target sentence data input by the user is input into the optimized target analysis model, so that the target sentence data is analyzed and processed by the optimized target analysis model and a corresponding simple sentence result is output.

[0149] After completing the construction of the target sentence analysis model, the present application will also intelligently use the verification data set constructed based on the compound sentence data to perform performance verification on the target sentence analysis model, and when it is detected that the target sentence analysis model has not passed the performance verification, the model optimization strategy will be automatically used to perform corresponding optimization processing on the target sentence analysis model to ensure that the final optimized target sentence analysis model can meet the corresponding model performance requirements, which is conducive to further improving the subsequent use of the optimized target analysis model to analyze and process the target sentence data input by the user, thereby effectively ensuring the data accuracy of the generated simple sentence results corresponding to the target sentence data.

[0150] In some optional implementations of this embodiment, after the step of performing performance verification on the target sentence analysis model based on the verification data set, the electronic device may further perform the following steps:

[0151] If the target sentence analysis model passes the performance verification, the model parameters and configuration data of the target sentence analysis model are retained.

[0152] In this embodiment, if the target sentence analysis model passes the performance verification, it indicates that the target sentence analysis model is a qualified model that meets the performance requirements of the constructed model, and thus there is no need to perform additional adjustment processing on the target sentence analysis model. It is only necessary to retain the model parameters and configuration data of the target sentence analysis model.

[0153] Determine a target storage mode corresponding to the target sentence analysis model.

[0154] In this embodiment, there is no specific limitation on the selection of the above-mentioned target storage method, which can be determined according to the actual storage business needs. Exemplarily, the target storage method can be any one of blockchain storage, network disk storage, online server storage, local database storage, etc. Among them, blockchain storage is preferably used as the above-mentioned target storage method. By using the storage method of blockchain storage to store and manage the target sentence analysis model, the security and non-tamperability of the target sentence analysis model can be effectively guaranteed.

[0155] The target sentence analysis model is stored based on the target storage method.

[0156] In this embodiment, the target storage medium corresponding to the target storage method can be determined, and then the target sentence analysis model can be stored in the target storage medium to complete the storage processing of the target sentence analysis model. Exemplarily, if the target storage method is blockchain storage, the corresponding target storage medium is blockchain, and then the storage processing of the target sentence analysis model can be performed by using blockchain.

[0157] After detecting that the target sentence analysis model has passed the performance verification, the present application will intelligently retain the model parameters and configuration data of the target sentence analysis model to complete the final construction of the target sentence analysis model, thereby ensuring the accuracy of the obtained target sentence analysis model. In addition, the target sentence analysis model will be stored and processed using a target storage method that matches the target sentence analysis model, which can effectively ensure the data security and non-tamperability of the target sentence analysis model, thereby improving the storage intelligence of the target sentence analysis model.

[0158] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps:

[0159] Call a pre-built intent recognition model.

[0160] In this embodiment, the above-mentioned intent recognition model can adopt an intent recognition model based on a deep learning algorithm or an intent recognition model based on a joint model. Specifically, a convolutional neural network can be used as an intent recognition model, and local features of the input sentence can be extracted through operations such as convolution and pooling, thereby realizing intent recognition. Alternatively, LSTM (Long Short-Term Memory Network) and GRU (Gated Recurrent Unit) can be used as intent recognition models to process variable-length sequence data, and long-term dependencies in the input sentence are captured through memory units to realize intent recognition. Alternatively, a bidirectional GRU+CRF model can be used as an intent recognition model. Since the bidirectional GRU+CRF model has a high accuracy in intent recognition, it is preferred to use a bidirectional GRU+CRF model as the above-mentioned intent recognition model.

[0161] The intention recognition is performed on the simple sentence result based on the intention recognition model to obtain the user intention corresponding to the simple sentence result.

[0162] In this embodiment, the simple sentence results specifically include a plurality of simple sentences outputted after the target analysis model analyzes and processes the target sentence data inputted by the user. And a simple sentence has an intention. The obtained simple sentences can be respectively inputted into the intention recognition model for intention recognition, thereby obtaining the intentions corresponding to each simple sentence. The above user intention is composed of a plurality of intentions corresponding to each simple sentence. Compared with the method of using the intention recognition model to perform intent recognition on compound sentence data, the method of using the intention recognition model to perform intent recognition on each simple sentence obtained by splitting the compound sentence data can accurately extract the user's user intention and ensure the data accuracy of the obtained user intention.

[0163] A corresponding response process is executed based on the user intention.

[0164] In this embodiment, according to the intentions corresponding to the simple sentences contained in the above user intentions, response processing corresponding to each intention can be performed respectively to accurately complete the service processing that meets the user intentions.

[0165] This application calls a pre-built intent recognition model; then performs intent recognition on the simple sentence result based on the intent recognition model to obtain the user intent corresponding to the simple sentence result; and subsequently performs corresponding response processing based on the user intent. This application uses an intent recognition model to perform intent recognition on each simple sentence obtained by splitting the compound sentence data. Compared with the method of using an intent recognition model to perform intent recognition on compound sentence data, this can effectively improve the accuracy of identifying the user's intent, ensure the accuracy of the user intent corresponding to the simple sentence result, and perform corresponding response processing based on the obtained user intent, so that the service processing that meets the user's intent can be accurately completed, which is conducive to improving the user's experience.

[0166] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0167] It should be emphasized that in order to further ensure the privacy and security of the above simple sentence results, the above simple sentence results can also be stored in a node of a blockchain.

[0168] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0169] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0170] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0171] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0172] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0173] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a sentence analysis device based on artificial intelligence, and the embodiment of the device is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0174] like Figure 3 As shown, the artificial intelligence-based sentence analysis device 300 described in this embodiment includes: a first acquisition module 301, a conversion module 302, a labeling module 303, a training module 304, a first determination module 305, a first optimization module 306 and an analysis module 307. Among them:

[0175] The first acquisition module 301 is used to acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions;

[0176] A conversion module 302 is used to convert the compound sentence data to obtain a corresponding word sequence;

[0177] A labeling module 303 is used to label the word sequence based on a preset labeling method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence;

[0178] A training module 304 is used to train a pre-built initial sentence analysis model based on the target word sequence to obtain a training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field, and a prediction layer;

[0179] A first determination module 305, configured to determine the training loss of the initial sentence analysis model based on the training output result and the label data;

[0180] A first optimization module 306, configured to perform model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model;

[0181] The analysis module 307 is used to analyze and process the target sentence data input by the user based on the target analysis model, and generate a simple sentence result corresponding to the target sentence data.

[0182] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based sentence analysis method in the aforementioned implementation mode, and will not be repeated here.

[0183] In some optional implementations of this embodiment, the first acquisition module 301 includes:

[0184] The acquisition submodule is used to obtain the input initial compound sentence data;

[0185] A replacement submodule, configured to perform replacement processing on the initial compound sentence data based on a preset word replacement strategy to obtain corresponding first sentence data;

[0186] A combining submodule, configured to combine the initial compound sentence data based on a preset sentence combination strategy to obtain corresponding second sentence data;

[0187] A construction submodule is used to construct the compound sentence data based on the initial compound sentence data, the first sentence data and the second sentence data.

[0188] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based sentence analysis method in the aforementioned implementation mode, and will not be repeated here.

[0189] In some optional implementations of this embodiment, the training module 304 includes:

[0190] An input submodule, used for inputting the target word sequence into the initial sentence analysis model;

[0191] A replacement submodule, used for performing data conversion processing on the target word sequence through the pre-trained language model to obtain a corresponding hidden state vector;

[0192] A mapping submodule, used for performing data mapping processing on the hidden state vector through the linear layer to obtain a corresponding sentence feature matrix;

[0193] A first processing submodule is used to process the sentence matrix features through the conditional random field to obtain corresponding scoring data;

[0194] A second processing submodule is used to perform normalization processing on the scoring data based on the prediction layer to obtain corresponding prediction probability distribution data;

[0195] The first determination submodule is used to use the predicted probability distribution data as the training output result of the initial sentence analysis model.

[0196] In some optional implementations of this embodiment, the first determining module 305 includes:

[0197] The second determination submodule is used to determine the target loss function;

[0198] A calculation submodule, used for calculating and processing the training output and the label data based on the target loss function to obtain corresponding difference data;

[0199] The third determination submodule is used to use the difference data as the training loss of the initial sentence analysis model.

[0200] In some optional implementations of this embodiment, the sentence analysis device based on artificial intelligence further includes:

[0201] A construction module, used to construct a verification data set based on the compound sentence data;

[0202] A verification module, used to perform performance verification on the target sentence analysis model based on the verification data set;

[0203] A second acquisition module is used to acquire a preset model optimization strategy if the target sentence analysis model fails the performance verification;

[0204] A second optimization module, configured to optimize the target sentence analysis model based on the model optimization strategy to obtain an optimized target sentence analysis model;

[0205] The analysis module 307 includes:

[0206] The analysis submodule is used to analyze and process the target sentence data input by the user based on the optimized target analysis model, and generate a simple sentence result corresponding to the target sentence data.

[0207] In some optional implementations of this embodiment, the sentence analysis device based on artificial intelligence further includes:

[0208] A retaining module, used for retaining the model parameters and configuration data of the target sentence analysis model if the target sentence analysis model passes the performance verification;

[0209] A second determination module, used to determine a target storage mode corresponding to the target sentence analysis model;

[0210] A storage module is used to store the target sentence analysis model based on the target storage method.

[0211] In some optional implementations of this embodiment, the sentence analysis device based on artificial intelligence further includes:

[0212] A calling module is used to call a pre-built intent recognition model;

[0213] An identification module, used to perform intent identification on the simple sentence result based on the intent identification model to obtain a user intent corresponding to the simple sentence result;

[0214] A response module is used to perform corresponding response processing based on the user intention.

[0215] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0216] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0217] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0218] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a sentence analysis method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0219] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based sentence analysis method.

[0220] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0221] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the artificial intelligence-based sentence analysis method as described above.

[0222] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0223] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A sentence analysis method based on artificial intelligence, characterized in that: The steps include: Acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions; Converting the compound sentence data to obtain a corresponding word sequence; The word sequence is annotated based on a preset annotation method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence; Training a pre-constructed initial sentence analysis model based on the target word sequence to obtain a training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field, and a prediction layer; Determining a training loss of the initial sentence analysis model based on the training output result and the label data; Performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model; The target sentence data input by the user is analyzed and processed based on the target analysis model to generate a simple sentence result corresponding to the target sentence data.

2. The sentence analysis method based on artificial intelligence according to claim 1, characterized in that: The step of obtaining pre-constructed complex sentence data specifically includes: Obtain input initial compound sentence data; Performing a replacement process on the initial compound sentence data based on a preset word replacement strategy to obtain corresponding first sentence data; Combining the initial compound sentence data based on a preset sentence combination strategy to obtain corresponding second sentence data; The compound sentence data is constructed based on the initial compound sentence data, the first sentence data and the second sentence data.

3. The sentence analysis method based on artificial intelligence according to claim 1, characterized in that: The step of training the pre-built initial sentence analysis model based on the target word sequence to obtain the training output result of the initial sentence analysis model specifically includes: Inputting the target word sequence into the initial sentence analysis model; Performing data conversion processing on the target word sequence through the pre-trained language model to obtain a corresponding hidden state vector; Performing data mapping processing on the hidden state vector through the linear layer to obtain a corresponding sentence feature matrix; The sentence matrix features are processed by the conditional random field to obtain corresponding scoring data; Based on the prediction layer, the scoring data is normalized to obtain corresponding prediction probability distribution data; The predicted probability distribution data is used as the training output result of the initial sentence analysis model.

4. The sentence analysis method based on artificial intelligence according to claim 1, characterized in that: The step of determining the training loss of the initial sentence analysis model based on the training output result and the label data specifically includes: Determine the target loss function; Based on the target loss function, the training output and the label data are calculated and processed to obtain corresponding difference data; The difference data is used as the training loss of the initial sentence parsing model.

5. The sentence analysis method based on artificial intelligence according to claim 1, characterized in that: After the step of performing model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model, the method further includes: Constructing a verification data set based on the compound sentence data; Performing performance verification on the target sentence analysis model based on the verification data set; If the target sentence analysis model fails the performance verification, obtaining a preset model optimization strategy; Optimizing the target sentence analysis model based on the model optimization strategy to obtain an optimized target sentence analysis model; The step of analyzing and processing the target sentence data input by the user based on the target analysis model to generate a simple sentence result corresponding to the target sentence data includes: The target sentence data input by the user is analyzed and processed based on the optimized target analysis model to generate a simple sentence result corresponding to the target sentence data.

6. The sentence analysis method based on artificial intelligence according to claim 5 is characterized in that: After the step of performing performance verification on the target sentence analysis model based on the verification data set, the method further includes: If the target sentence analysis model passes the performance verification, retaining the model parameters and configuration data of the target sentence analysis model; Determining a target storage mode corresponding to the target sentence analysis model; The target sentence analysis model is stored based on the target storage method.

7. The sentence analysis method based on artificial intelligence according to claim 1 is characterized in that: After the step of analyzing and processing the target sentence data input by the user based on the target analysis model to generate a simple sentence result corresponding to the target sentence data, the method further includes: Calling pre-built intent recognition models; Performing intent recognition on the simple sentence result based on the intent recognition model to obtain a user intent corresponding to the simple sentence result; A corresponding response process is executed based on the user intention.

8. A sentence analysis device based on artificial intelligence, characterized in that: include: A first acquisition module is used to acquire pre-constructed complex sentence data; wherein the complex sentence data is sentence data containing multiple intentions; A conversion module, used for converting the compound sentence data to obtain a corresponding word sequence; A labeling module, used to label the word sequence based on a preset labeling method to obtain a corresponding target word sequence; wherein the target word sequence includes label data corresponding to the word sequence; A training module, used to train a pre-built initial sentence analysis model based on the target word sequence, and obtain the training output result of the initial sentence analysis model; wherein the sentence analysis model includes a pre-trained language model, a linear layer, a conditional random field and a prediction layer; A first determination module, configured to determine a training loss of the initial sentence analysis model based on the training output result and the label data; A first optimization module, configured to perform model optimization processing on the sentence analysis model based on the training loss to obtain a corresponding target sentence analysis model; The analysis module is used to analyze and process the target sentence data input by the user based on the target analysis model, and generate a simple sentence result corresponding to the target sentence data.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the sentence analysis method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the sentence analysis method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images