Respiratory disease treatment feedback data processing method and system based on artificial intelligence

Abstract

According to the artificial intelligence-based respiratory disease treatment feedback data processing method and system provided by the invention, the efficient decision performance of the first treatment feedback matching model and the low decision error characteristic of the second treatment feedback matching model are combined; and the evaluation precision of the feedback content semantic relevance between the respiratory disease treatment feedback data is remarkably improved. By introducing a training cost mechanism and dynamically adjusting the number of iterations of each training sequence according to the training cost mechanism, the distinguishing capability of the model on complex feedback data is effectively enhanced, and the performance is particularly outstanding when counter-example data which is obviously different from target respiratory disease treatment feedback data and is high in distinguishing difficulty is processed. Therefore, the convergence speed of the model is accelerated, the advantages of the two models are fused through the iterative parameter interaction training strategy, and the finally generated target treatment feedback matching model greatly reduces the decision error while maintaining the efficient processing capability.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to an artificial intelligence-based respiratory disease treatment feedback data processing method and system. Background Art

[0002] In the treatment of respiratory diseases, timely and accurate acquisition and analysis of patient feedback data is crucial for adjusting treatment plans and evaluating treatment effectiveness. However, traditional methods for processing feedback data often rely on manual review and analysis, which is not only inefficient but also susceptible to subjective factors, making it difficult to quickly and accurately process large-scale and diverse feedback data.

[0003] The rapid development of artificial intelligence (AI), particularly breakthroughs in natural language processing and deep learning, has provided new approaches for the automated processing of respiratory disease treatment feedback data. Existing methods have emerged that use single models to analyze the semantic relevance of feedback data. While these methods have improved processing efficiency to a certain extent, they still suffer from large model decision errors and insufficient generalization capabilities. Especially when faced with complex and ever-changing respiratory disease treatment feedback data, a single model often struggles to achieve both efficiency and accuracy.

[0004] To address these issues, relevant technologies have begun exploring strategies that combine the strengths of multiple models. One effective approach is to introduce multiple treatment feedback matching models and leverage their complementarity to optimize overall performance. Specifically, the efficient decision-making performance of one model is leveraged to rapidly process large amounts of data, while the low decision-making error characteristics of another model are combined to ensure the accuracy of the evaluation results. However, how to effectively integrate these models so that they can work together and generate a more accurate target treatment feedback matching model remains an urgent problem.

[0005] Furthermore, a major challenge facing current technology is how to dynamically adjust training strategies for feedback data of varying complexity during training to fully utilize limited computing resources and accelerate model convergence. The traditional approach of evenly distributing training resources often results in insufficient model recognition of certain complex feedback data, thus impacting overall performance. Summary of the Invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides a method for processing respiratory disease treatment feedback data based on artificial intelligence, the method comprising: Obtaining a first treatment feedback matching model and a second treatment feedback matching model, wherein the first treatment feedback matching model and the second treatment feedback matching model are configured to determine feedback content semantic relevance between respiratory disease treatment feedback data; a model decision error of the first treatment feedback matching model is greater than a model decision error of the second treatment feedback matching model, and a model decision performance of the first treatment feedback matching model is greater than a model decision performance of the second treatment feedback matching model; Acquire multiple respiratory disease treatment feedback data training sequences, each of the respiratory disease treatment feedback data training sequences including target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data; the counterexample respiratory disease treatment feedback data in different respiratory disease treatment feedback data training sequences correspond to different training costs, the training costs being used to represent a complexity index for distinguishing the target respiratory disease treatment feedback data from the corresponding counterexample respiratory disease treatment feedback data; Determining the number of training iterations for each of the respiratory disease treatment feedback data training sequences, wherein the number of training iterations for any of the respiratory disease treatment feedback data training sequences is positively correlated with the training cost of counterexample respiratory disease treatment feedback data in the corresponding respiratory disease treatment feedback data training sequence; According to the ascending order of the number of training iterations, the first treatment feedback matching model is used to perform iterative parameter interactive training based on the multiple respiratory disease treatment feedback data training sequences and the second treatment feedback matching model to generate a target treatment feedback matching model.

[0007] For example, in a possible implementation of the first aspect, before performing iterative parameter interactive training, the method further includes: Obtaining a first sample sequence, the first sample sequence including at least one first combined sample set, each first combined sample set including: target respiratory disease treatment feedback data, a first positive example corresponding to the target respiratory disease treatment feedback data, and a first negative example corresponding to the target respiratory disease treatment feedback data; wherein the target respiratory disease treatment feedback data in different first combined sample sets is different; Based on the first sample sequence, feature comparison learning is performed on the first treatment feedback matching model.

[0008] For example, in a possible implementation of the first aspect, the first treatment feedback matching model extracts semantic feature data of each respiratory disease treatment feedback data from two respiratory disease treatment feedback data respectively, and uses the semantic feature matching degree between the two extracted semantic feature data as the semantic correlation of the feedback content between the two respiratory disease treatment feedback data; the second treatment feedback matching model fuses the two respiratory disease treatment feedback data to obtain fused respiratory disease treatment feedback data, estimates the semantic feature data of the fused respiratory disease treatment feedback data, and generates the semantic correlation of the feedback content between the two respiratory disease treatment feedback data based on the extracted semantic feature data.

[0009] On the other hand, an embodiment of the present application also provides an artificial intelligence-based respiratory disease treatment feedback data processing system, including a processor and a machine-readable storage medium, the machine-readable storage medium being connected to the processor, the machine-readable storage medium being used to store programs, instructions or codes, and the processor being used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above-mentioned method.

[0010] Based on the above aspects, the embodiment of the present application significantly improves the evaluation accuracy of the semantic relevance of feedback content between respiratory disease treatment feedback data by combining the efficient decision-making performance of the first treatment feedback matching model with the low decision-making error characteristics of the second treatment feedback matching model. By introducing a training cost mechanism and dynamically adjusting the number of iterations of each training sequence accordingly, the model's ability to discriminate complex feedback data is effectively enhanced, especially when dealing with counterexample data that are significantly different from the target respiratory disease treatment feedback data and difficult to distinguish. As a result, not only is the convergence speed of the model accelerated, but the advantages of the two models are also integrated through an iterative parameter interaction training strategy. The target treatment feedback matching model ultimately generated significantly reduces decision errors while maintaining efficient processing capabilities, providing more accurate and reliable feedback data support for the adjustment of treatment plans for respiratory diseases, thereby promoting the optimization of patient treatment effects and the realization of personalized medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a schematic diagram of the execution flow of the artificial intelligence-based respiratory disease treatment feedback data processing method provided in an embodiment of the present application.

[0012] Figure 2 This is a schematic diagram of the hardware architecture of the artificial intelligence-based respiratory disease treatment feedback data processing system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The present application will be described in detail below with reference to the accompanying drawings. Figure 1This is a flow chart of an artificial intelligence-based respiratory disease treatment feedback data processing method provided by an embodiment of the present application. The artificial intelligence-based respiratory disease treatment feedback data processing method is introduced in detail below.

[0014] Step S110: Obtain a first treatment feedback matching model and a second treatment feedback matching model. The first treatment feedback matching model and the second treatment feedback matching model are configured to determine semantic relevance of feedback content between respiratory disease treatment feedback data. The model decision error of the first treatment feedback matching model is greater than the model decision error of the second treatment feedback matching model, and the model decision performance of the first treatment feedback matching model is greater than the model decision performance of the second treatment feedback matching model.

[0015] In this embodiment, in order to improve the accurate matching capability of treatment feedback, this embodiment develops and optimizes two treatment feedback matching models: a first treatment feedback matching model (model A) and a second treatment feedback matching model (model B).

[0016] In detail, the server first connects to the data warehouse and retrieves the pre-trained versions of Model A and Model B. Both Model A and Model B are designed to evaluate the semantic relevance of feedback content between respiratory disease treatment feedback data, but they differ in design and performance.

[0017] Model A (First Treatment Feedback Matching Model): Utilizing a deep learning architecture, this model uses a multi-layer neural network to extract semantic features from text data and then calculate the correlations between respiratory disease treatment feedback data. Although Model A may incur significant errors in its decision-making process (for example, inaccurate feedback assessments for certain complex cases), it offers high decision-making performance and can rapidly respond to and process large amounts of data.

[0018] Model B (Second Treatment Feedback Matching Model): Utilizing an ensemble learning approach, it combines the predictions of multiple basic models (such as support vector machines and decision trees) and uses a voting mechanism to arrive at a final correlation estimate. Model B has a relatively small decision error, but is slightly less efficient than Model A when processing large amounts of data. After acquiring both models, the server loads them into memory for subsequent iterative parameter interactive training.

[0019] Step S120: Acquire multiple respiratory disease treatment feedback data training sequences, each of which includes target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data. Counterexample respiratory disease treatment feedback data in different respiratory disease treatment feedback data training sequences are associated with different training costs, each of which represents a complexity index for distinguishing between the target respiratory disease treatment feedback data and the corresponding counterexample respiratory disease treatment feedback data.

[0020] In order to train and optimize the treatment feedback matching model, the server needs to extract actual treatment feedback data from the treatment feedback database as training samples.

[0021] Specifically, the server first connects to a treatment feedback database containing thousands of treatment feedback records from respiratory disease patients over the past few years. Each treatment feedback record details the patient's diagnosis, treatment plan, and treatment effect feedback.

[0022] The server selects a certain amount of target respiratory disease treatment feedback data from the database randomly or according to a specific strategy (such as by disease type, treatment time period, etc.). These target respiratory disease treatment feedback data will be used as positive examples in each training sequence for model learning and evaluation.

[0023] For each target respiratory disease treatment feedback data set, the server needs to find corresponding counterexample respiratory disease treatment feedback data, i.e., records that differ significantly from the target respiratory disease treatment feedback data in treatment content or feedback semantics. To generate high-quality counterexample respiratory disease treatment feedback data, Model A (a model with relatively good initial performance) is used to perform a preliminary evaluation of each record in the treatment feedback database, calculating the semantic relevance of the feedback content between each record and the target respiratory disease treatment feedback data set. The server then sorts the records from high to low based on the relevance score, and selects the top X records with the lowest scores as the counterexample respiratory disease treatment feedback data set. Here, X is a preset positive integer representing the number of counterexamples corresponding to each target respiratory disease treatment feedback data set.

[0024] A training cost can be assigned to each counterexample respiratory disease treatment feedback data set. The training cost is dynamically calculated based on factors such as the degree of semantic difference between the counterexample respiratory disease treatment feedback data and the target respiratory disease treatment feedback data, and the complexity of the counterexample respiratory disease treatment feedback data. The greater the difference and the higher the complexity, the higher the training cost.

[0025] The server combines each target respiratory disease treatment feedback data and its corresponding counterexample respiratory disease treatment feedback data into a training sequence. Each training sequence contains a clear target (positive example) and the interference items (counterexamples) to be distinguished, as well as the corresponding training cost information.

[0026] Step S130, determining the number of training iterations of each of the respiratory disease treatment feedback data training sequences, wherein the number of training iterations of any of the respiratory disease treatment feedback data training sequences is positively correlated with the training cost of the counterexample respiratory disease treatment feedback data in the corresponding respiratory disease treatment feedback data training sequence.

[0027] In this embodiment, during the iterative parameter interactive training process, different training sequences need to be allocated different numbers of training iterations to achieve the best training effect due to the different complexity and training costs of their counter-example respiratory disease treatment feedback data.

[0028] Specifically, the server traverses all training sequences and determines the number of training iterations based on the training cost of the counterexample respiratory disease treatment feedback data in each training sequence. Specifically, for each training sequence, the server first calculates the average training cost of all its counterexample respiratory disease treatment feedback data. Based on the average training cost, the server assigns a different number of training iterations to each training sequence. The higher the cost, the more iterations are required to ensure that the model can fully learn and distinguish complex counterexample respiratory disease treatment feedback data. Finally, the server sorts all training sequences in ascending order of the number of iterations, preparing for subsequent iterative parameter interactive training.

[0029] Step S140 , performing iterative parameter interactive training based on the plurality of respiratory disease treatment feedback data training sequences and the second treatment feedback matching model using the first treatment feedback matching model in ascending order of training iteration numbers, to generate a target treatment feedback matching model.

[0030] In this embodiment, after determining the number of training iterations for each training sequence, the server begins to execute an iterative parameter interactive training process, aiming to combine the efficiency of model A and the accuracy of model B to generate a target treatment feedback matching model with better performance.

[0031] In detail, the server first initializes an empty target-treatment-feedback matching model (Model C), which will be gradually constructed based on the interactive training results of Model A and Model B.

[0032] The server then selects a training sequence with the fewest training iterations and uses Model A to perform a preliminary evaluation of the target respiratory disease treatment feedback data and the counterexample respiratory disease treatment feedback data in that sequence, generating the semantic relevance of the first feedback content. Simultaneously, the server performs the same evaluation using Model B to generate the semantic relevance of the second feedback content.

[0033] Next, the server compares the evaluation results of Model A and Model B and calculates the error between them. Based on this error information, the server optimizes the weight and bias parameters of Model A to make it closer to the accurate evaluation results of Model B. The optimized Model A (or some of its parameters) is then used to update the target Model C.

[0034] The server processes the remaining training sequences in ascending order of training iteration number. The initial evaluation, error calculation, and parameter optimization steps are repeated for each sequence. As the iterations progress, the target model C gradually combines the efficient processing capabilities of model A with the accurate evaluation capabilities of model B.

[0035] After multiple rounds of iterative training, the server evaluates the performance of the target model C. If the model performance meets the preset standard or shows a clear convergence trend, training is stopped; otherwise, the training strategy is adjusted based on the latest evaluation results, and iterative training continues until the requirements are met.

[0036] Once the target model C is trained and verified, the server deploys it into the actual treatment feedback matching system. Doctors and patients can use this system to quickly and accurately assess the semantic relevance of treatment feedback, providing strong support for adjusting treatment plans for respiratory diseases.

[0037] Based on the above steps, by combining the advantages of the two initial treatment feedback matching models, a more optimized target treatment feedback matching model was generated. The application of this target treatment feedback matching model will significantly improve the accuracy and efficiency of treatment feedback matching in medical institutions.

[0038] Based on the above steps, the embodiment of the present application significantly improves the evaluation accuracy of the semantic relevance of feedback content between respiratory disease treatment feedback data by combining the efficient decision-making performance of the first treatment feedback matching model with the low decision-making error characteristics of the second treatment feedback matching model. By introducing a training cost mechanism and dynamically adjusting the number of iterations of each training sequence accordingly, the model's ability to discriminate complex feedback data is effectively enhanced, especially when dealing with counterexample data that are significantly different from the target respiratory disease treatment feedback data and difficult to distinguish. As a result, not only is the convergence speed of the model accelerated, but the advantages of the two models are also integrated through an iterative parameter interaction training strategy. The target treatment feedback matching model ultimately generated significantly reduces decision-making errors while maintaining efficient processing capabilities, providing more accurate and reliable feedback data support for the adjustment of treatment plans for respiratory diseases, thereby promoting the optimization of patient treatment effects and the realization of personalized medical services.

[0039] In a possible implementation, step S120 includes: Step S121 : obtaining target respiratory disease treatment feedback data and a treatment feedback database, wherein the treatment feedback database includes a plurality of treatment feedback samples.

[0040] In this embodiment, the server first connects to the medical institution's data warehouse and retrieves the latest target respiratory disease treatment feedback data. This target respiratory disease treatment feedback data is typically patient treatment feedback records collected over a recent period of time, including detailed diagnosis information, treatment plans, treatment effect feedback, etc.

[0041] The server is also connected to a treatment feedback database that stores treatment feedback samples from tens of thousands of respiratory disease patients over the past few years. These treatment feedback samples serve as a potential source of counterexample respiratory disease treatment feedback data for training and optimizing the treatment feedback matching model.

[0042] Step S122: Invoking a target treatment feedback matching model to estimate the semantic relevance of feedback content between the target respiratory disease treatment feedback data and each treatment feedback sample in the treatment feedback database, thereby generating a target feedback content semantic relevance corresponding to each treatment feedback sample. The target feedback content semantic relevance corresponding to any treatment feedback sample is positively correlated with the training cost of the corresponding treatment feedback sample.

[0043] In this embodiment, the server selects an initial target treatment feedback matching model (which can be Model A or Model B; in this example, Model A is assumed to be used as the initial estimation model). The server then traverses each treatment feedback sample in the treatment feedback database and uses Model A to estimate the semantic relevance of the feedback content between each treatment feedback sample and the target respiratory disease treatment feedback data.

[0044] Specifically, the server loads Model A into memory. For each treatment feedback sample in the treatment feedback database, the server extracts its textual content and passes it to Model A for processing. Model A uses a multi-layer neural network to extract the sample's semantic features, compares them with the semantic features of the target respiratory disease treatment feedback data, and calculates a semantic relevance score for the feedback content. The server records the relevance score for each treatment feedback sample and uses it as the "target feedback content semantic relevance" for that sample.

[0045] Step S123 , extracting X treatment feedback samples from the treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data according to descending order of semantic relevance of the target feedback content corresponding to each treatment feedback sample, where X is a positive integer.

[0046] After calculating the correlation score between each treatment feedback sample and the target respiratory disease treatment feedback data, the server needs to assign a training cost to each treatment feedback sample. The training cost is used to represent the complexity of the model in distinguishing target samples from counterexamples.

[0047] The server initially determines the sample's complexity and difficulty of differentiation based on the relevance score. Generally speaking, samples with lower relevance scores (i.e., greater differences from the target sample) have higher training costs. The server may also consider other factors, such as the sample's text length, the number of technical terms it contains, and the degree of semantic ambiguity, to dynamically adjust the training cost. Ultimately, each treatment feedback sample is assigned a specific training cost, which is used in subsequent training sequence construction and iteration allocation.

[0048] The server extracts a certain number of samples from the treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data according to the descending order of the semantic relevance of the target feedback content corresponding to each treatment feedback sample.

[0049] Specifically, the server first sorts the samples in the treatment feedback database by relevance score, from low to high. Then, the server selects the top X treatment feedback samples with the lowest scores as counterexample respiratory disease treatment feedback data. Here, X is a preset positive integer representing the number of counterexamples corresponding to each target sample. If the number of samples in the treatment feedback database is much larger than X, the server simply selects the top X treatment feedback samples. If the number of samples is less than X, the server selects all available samples.

[0050] Step S124 : configuring a plurality of respiratory disease treatment feedback data training sequences based on the target respiratory disease treatment feedback data and the extracted X counter-example respiratory disease treatment feedback data.

[0051] In this embodiment, the server combines each target respiratory disease treatment feedback data and its corresponding X counterexample respiratory disease treatment feedback data into a training sequence. These training sequences will be used in the subsequent iterative parameter interactive training process.

[0052] Specifically, the server labels each target respiratory disease treatment feedback data set as a positive example. It then combines the previously extracted X negative respiratory disease treatment feedback data sets with the positive example data to form a training sequence. Each training sequence contains a clear target (positive example) and the distractors (negative examples) to be distinguished, as well as the training cost information corresponding to each negative respiratory disease treatment feedback data set. The server stores all training sequences for subsequent iterative training.

[0053] Through the above steps, the server retrieves multiple training sequences from the treatment feedback database, including target respiratory disease treatment feedback data and corresponding counterexample respiratory disease treatment feedback data. These training sequences will be used in the subsequent iterative parameter interactive training process to optimize the performance of the treatment feedback matching model.

[0054] In a possible implementation, when the target treatment feedback matching model is the first treatment feedback matching model, step S124 includes: Step S1241 , sorting the extracted X counter-example respiratory disease treatment feedback data according to the descending order of semantic relevance of target feedback content corresponding to the X counter-example respiratory disease treatment feedback data, and generating a target respiratory disease treatment feedback data distribution.

[0055] In this example, at a medical institution, a server is using a first treatment feedback matching model (Model A) as the initial target treatment feedback matching model to build and optimize a model for respiratory disease treatment feedback matching. This scenario details how, when the target model is Model A, the server configures multiple training sequences based on target respiratory disease treatment feedback data and extracted counterexample respiratory disease treatment feedback data.

[0056] Specifically, based on the preliminary evaluation of Model A, the server has extracted X counterexample respiratory disease treatment feedback data from the treatment feedback database that are least correlated with the target respiratory disease treatment feedback data. Next, the server needs to rank these counterexample respiratory disease treatment feedback data based on their semantic relevance to the target feedback content (generated by Model A).

[0057] The server iterates over these X counterexample respiratory disease treatment feedback data and obtains the semantic relevance score of the target feedback content corresponding to each counterexample respiratory disease treatment feedback data. These counterexample respiratory disease treatment feedback data are sorted from low to high based on the score (i.e., from weak to strong relevance). Lower scores indicate greater divergence from the target data and a greater challenge for model training. After sorting, the server stores these counterexample respiratory disease treatment feedback data in the new order to generate a distribution of the target respiratory disease treatment feedback data. This distribution reflects the difficulty of these counterexample respiratory disease treatment feedback data from the perspective of Model A.

[0058] Step S1242: Estimate the feedback content semantic relevance between the target respiratory disease treatment feedback data and each counterexample respiratory disease treatment feedback data through the second treatment feedback matching model, and generate candidate feedback content semantic relevance corresponding to each counterexample respiratory disease treatment feedback data.

[0059] Step S1243 , sorting the X counter-example respiratory disease treatment feedback data according to the descending order of the semantic relevance of the candidate feedback content corresponding to each counter-example respiratory disease treatment feedback data, to generate a candidate respiratory disease treatment feedback data distribution.

[0060] Step S1244: Perform multiple rounds of data sampling in the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution, respectively, in a reverse sampling direction. A respiratory disease treatment feedback data training sequence includes Y counterexample respiratory disease treatment feedback data, where Y is a positive integer. One round of data sampling is used to extract Z counterexample respiratory disease treatment feedback data from each of the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution, where Z∈[Y / 2,X].

[0061] Step S1245: After the mth round of data sampling is completed, based on the Z counter-example respiratory disease treatment feedback data extracted from the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution in this round, a list of counter-example respiratory disease treatment feedback data corresponding to the mth training iteration number is configured.

[0062] Step S1246 , configuring the respiratory disease treatment feedback data training sequence of the mth training iteration according to the target respiratory disease treatment feedback data and the counterexample respiratory disease treatment feedback data list configured in this round, where m is a positive integer.

[0063] To introduce diversity and potentially discover features that Model A failed to capture, the server next uses a second treatment feedback matching model (Model B) to re-evaluate the feedback content semantic relevance between the target respiratory disease treatment feedback data and each counterexample respiratory disease treatment feedback data.

[0064] Specifically, the server loads Model B into memory. For each ranked counterexample respiratory disease treatment feedback data set, the server uses Model B to calculate a semantic relevance score between the candidate feedback content and the target respiratory disease treatment feedback data set. Based on these scores, the server again ranks the counterexample respiratory disease treatment feedback data sets, this time based on the evaluation results of Model B. After the ranking is complete, a distribution of candidate respiratory disease treatment feedback data sets is generated.

[0065] Next, the server now has two sorted distributions of counterexample respiratory disease treatment feedback data: one is the target distribution based on model A, and the other is the candidate distribution based on model B. To combine the advantages of the two models, the server will alternately perform reverse sampling in these two distributions to construct multiple training sequences.

[0066] Specifically, the server sets the number Y of counterexample respiratory disease treatment feedback data that each training sequence should contain, and determines the number Z of sampling per round, where the range of Z is [Y / 2, X).

[0067] In each round of sampling, the server first selects Z counter-example respiratory disease treatment feedback data from the target distribution, and then selects Z counter-example respiratory disease treatment feedback data from the candidate distribution. This reverse sampling method ensures that the training sequence contains both the samples that model A considers most challenging and the samples that model B may disagree with.

[0068] When Z is greater than Y / 2, the server randomly selects a portion of data from each distribution to ensure that each training sequence ultimately contains Y counterexample respiratory disease treatment feedback data. For example, if Y = 6 and Z = 5, the server may select 3 from the target distribution, 3 from the candidate distribution, and randomly discard one, or select according to other strategies.

[0069] After each round of sampling, the server creates a list of counterexample respiratory disease treatment feedback data corresponding to the number of training iterations extracted from the counterexample respiratory disease treatment feedback data. The server then combines this list with the target respiratory disease treatment feedback data to generate a complete training sequence.

[0070] Specifically, after the mth round of sampling (m is a positive integer), the server has Z counterexample respiratory disease treatment feedback data extracted from the two distributions. The server compiles this data into a list, labeled as the counterexample respiratory disease treatment feedback data list corresponding to the mth training iteration.

[0071] The server then combines this list with the target respiratory disease treatment feedback data (as positive examples) to form a training sequence containing Y+1 samples (Y negative examples + 1 positive example).

[0072] The server stores this training sequence and prepares it for use in subsequent iterative parameter interactive training processes. As m increases, the server will repeat the above sampling and configuration process until all required training sequences are generated.

[0073] Through the above steps, the server is configured with multiple respiratory disease treatment feedback data training sequences, which combine the evaluation results of model A and model B, aiming to optimize the final target treatment feedback matching model through iterative parameter interaction training.

[0074] In a possible implementation, when Z is greater than Y / 2, step S1245 includes: Step S1245 - 1 : randomly selecting Q counter-example respiratory disease treatment feedback data from the Z counter-example respiratory disease treatment feedback data extracted from the target respiratory disease treatment feedback data distribution in this round.

[0075] Step S1245-2: Randomly select W counter-example respiratory disease treatment feedback data from the Z counter-example respiratory disease treatment feedback data extracted from the candidate respiratory disease treatment feedback data distribution in this round, where Q and W are both positive integers, and Q+W=Y.

[0076] Step S1245-3: configuring a list of counter-example respiratory disease treatment feedback data corresponding to the mth training iteration number based on the Q counter-example respiratory disease treatment feedback data and the W counter-example respiratory disease treatment feedback data.

[0077] When constructing a training sequence for respiratory disease treatment feedback data, the server configures a list of counterexample data based on the results of each round of sampling. If the number of counterexample data, Z, extracted from both the target and candidate respiratory disease treatment feedback data distributions exceeds half the number of counterexample data required for the training sequence (i.e., Z > Y / 2), the server must adopt a specific strategy for selecting the counterexample data to be used in the training sequence.

[0078] In this example, the server has completed this round of sampling from the target respiratory disease treatment feedback data distribution and extracted Z counterexample respiratory disease treatment feedback data. Because Z is greater than Y / 2, the server needs to randomly select a portion of this data to ensure that the total number of counterexample data in the final training sequence is Y.

[0079] Specifically, the server first calculates the number of counterexamples, Q, that need to be drawn from the target distribution. Since Q + W = Y, and given Z (the total number of counterexamples drawn from the target distribution) and Y (the total number of counterexamples required for the training sequence), the server can calculate Q = Y - W. (Although W is not yet determined, the server can assume a reasonable starting value or range and adjust it later.) However, in practice, since Z is already given and greater than Y / 2, a more straightforward approach is to determine a Q value such that Q ≤ Z and is as close to Y / 2 as possible (to maintain a balance between the contributions of the two distributions), while also ensuring that Q + W = Y.

[0080] The server uses a random number generator or similar algorithm to randomly select Q data from the Z counterexample data extracted from the target distribution. These selected data will be used for subsequent training sequences.

[0081] Similar to extracting data from the target distribution, the server also needs to randomly select a portion of negative examples from the candidate respiratory disease treatment feedback data distribution. However, this time the selection is based on the previously determined Q value and the total number of negative examples Y required for the training sequence.

[0082] In detail, once the Q value is determined, the server can calculate the number of counterexample data W that needs to be extracted from the candidate distribution, that is, W = Y - Q.

[0083] The server again uses a random number generator or similar algorithm to randomly select W data from the Z counterexample data extracted from the candidate distribution. These selected data, together with the Q data extracted from the target distribution, will constitute the final list of counterexample data used for training sequences.

[0084] After randomly selecting Q and W negative examples from the two distributions, respectively, the server combines these data to form a list of negative respiratory disease treatment feedback data corresponding to the mth training iteration. Specifically, the server merges the Q negative examples from the target distribution and the W negative examples from the candidate distribution into a single list. This list now contains Y negative examples, the exact number required for the training sequence. The server configures this list, along with the target respiratory disease treatment feedback data (as positive examples), into a complete training sequence for the mth training iteration. Once the training sequence is ready, the server can use it in subsequent iterative parameter interactive training processes.

[0085] Through the above steps, when Z is greater than Y / 2, the server configures a list of counterexample respiratory disease treatment feedback data corresponding to the mth training iteration, ensuring the diversity and balance of counterexample data in the training sequence.

[0086] In a possible implementation, the target respiratory disease treatment feedback data is configured with positive respiratory disease treatment feedback data.

[0087] Step S123 includes: Step S1231: Optimize the treatment feedback database based on the positive respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data to generate an optimized treatment feedback database. The data optimization includes removing treatment feedback samples in the treatment feedback database that are identical to the positive respiratory disease treatment feedback data.

[0088] Step S1232 , extracting X treatment feedback samples from the optimized treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data according to descending order of semantic relevance of the target feedback content corresponding to each treatment feedback sample.

[0089] When constructing a training sequence for respiratory disease treatment feedback data, the server needs to ensure that there are clear differences between the target respiratory disease treatment feedback data (including positive feedback) and its corresponding negative example data so that the model can effectively learn to distinguish them. To achieve this, the server first optimizes the treatment feedback database to remove samples that are duplicated or overly similar to the target positive example data, and then extracts negative example data from this optimized database.

[0090] Before extracting negative data, the server has already identified the target respiratory disease treatment feedback data and configured corresponding positive respiratory disease treatment feedback data for it. Positive data is usually positive feedback on the treatment effect of the target disease and is used as positive examples for model learning during the training process.

[0091] The server then performs data optimization on the entire treatment feedback database to prepare for extracting negative examples. The main purpose of optimization is to eliminate samples that are identical or highly similar to the target positive examples, as these samples are not valuable for distinguishing positive examples from negative examples.

[0092] Specifically, the server first loads the target respiratory disease treatment feedback data and its corresponding positive example data into memory. The server then iterates over each sample in the treatment feedback database and calculates a similarity score between each sample and the target positive example data using a text similarity algorithm (such as cosine similarity or Jaccard similarity). For samples whose similarity scores exceed a preset threshold, the server marks them as "duplicate" or "too similar" and removes them from the database. These removed samples are no longer included in the subsequent negative example data extraction process.

[0093] After the above processing, the server generates an optimized treatment feedback database, in which there are obvious differences between the samples and the target positive data.

[0094] After optimizing the treatment feedback database, the server can now extract negative examples corresponding to the target respiratory disease treatment feedback data. Negative examples are typically negative feedback on treatment efficacy or feedback related to other diseases, and serve as negative examples for model learning during the training process. Specifically, the server uses the previously selected target treatment feedback matching model (e.g., Model A) to evaluate the semantic relevance of the feedback content between each sample in the optimized database and the target respiratory disease treatment feedback data. The server sorts the samples in the database in descending order of relevance scores. Samples with lower scores indicate greater differences from the target data and are more likely to be selected as negative examples. The server then extracts the top X samples from the sorted database as negative respiratory disease treatment feedback data. Here, X is a preset positive integer representing the number of negative examples corresponding to each target data point. These extracted negative examples, along with the target respiratory disease treatment feedback data (including positive feedback), form part of the training sequence for subsequent iterative parameter interaction training.

[0095] Through the above steps, the server is configured with positive and negative examples of the target respiratory disease treatment feedback data. These negative examples will serve as an important part of the training sequence, helping the model learn how to accurately distinguish between good and bad treatment effects, thereby improving the accuracy of respiratory disease treatment feedback matching.

[0096] In a possible implementation, a respiratory disease treatment feedback data training sequence with a training iteration number of m is represented as an m-th respiratory disease treatment feedback data training sequence, where m is a positive integer.

[0097] The method of performing the mth round of parameter interactive training based on the mth respiratory disease treatment feedback data training sequence and the second treatment feedback matching model using the first treatment feedback matching model includes: Step A110 , estimating feedback content semantic relevance between target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence through the first treatment feedback matching model, and generating first feedback content semantic relevance.

[0098] Step A120 , estimating the feedback content semantic relevance between the target respiratory disease treatment feedback data and the counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence through the second treatment feedback matching model, and generating a second feedback content semantic relevance.

[0099] Step A130: Optimizing the weight information and bias information of the first treatment feedback matching model based on the error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content.

[0100] In this example, during the construction and optimization of a respiratory disease treatment feedback matching model, the server uses iterative parameter interactive training to combine the strengths of the first treatment feedback matching model (Model A) and the second treatment feedback matching model (Model B), gradually improving model performance. This scenario details how the server utilizes these two models to perform the mth round of parameter interactive training.

[0101] Before starting the mth round of parameter interactive training, the server has prepared a training sequence of respiratory disease treatment feedback data with m training iterations, denoted as the mth training sequence. This sequence contains the target respiratory disease treatment feedback data (positive example) and the corresponding negative example respiratory disease treatment feedback data.

[0102] The server first loads the first treatment feedback matching model (model A) into memory and prepares to use it to evaluate the semantic relevance of the feedback content between the target data and the counterexample data in the mth training sequence.

[0103] Specifically, the server iterates over each counterexample in the mth training sequence. For each counterexample, the server inputs it, along with the target data, into Model A. Model A extracts semantic features from the text data through its multi-layer neural network structure and calculates a semantic relevance score for the feedback content between the target data and the counterexample. The server records the first semantic relevance score for each counterexample.

[0104] Next, the server loads the second treatment feedback matching model (Model B) into memory and uses it to re-evaluate the relevance between the target data and the counterexample data in the mth training sequence to generate a second set of relevance scores. In detail, similar to using Model A, the server iterates over each counterexample data in the mth training sequence. For each counterexample data, the server inputs it, along with the target data, into Model B. Model B calculates the feedback content semantic relevance score between the target data and the counterexample data using its ensemble learning method (combining multiple base models). The server records the second feedback content semantic relevance score corresponding to each counterexample data.

[0105] After obtaining two sets of relevance scores (the first and second sets), the server optimizes the weights and biases of the first treatment feedback matching model (Model A) based on the error between these two sets of scores. Specifically, the server calculates the error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content for each counterexample data point. The error can be the absolute difference, squared difference, or other form of difference metric. For all counterexample data points in the entire training sequence, the server aggregates these error values and evaluates the performance of Model A based on the sum or average of the errors. Based on this error information, the server uses backpropagation or other optimization algorithms to adjust the weights and bias parameters of Model A. The goal is to reduce the difference between the predictions of Model A (the first set of relevance scores) and the predictions of Model B (the second set of relevance scores). After multiple rounds of iterative adjustments, the server evaluates whether the performance of Model A has improved. If the improvement is significant, training continues with the next round. If the improvement is not significant or the preset training round limit has been reached, training is terminated and the optimized Model A is saved as the target treatment feedback matching model.

[0106] Through the above steps, the server completes the mth round of parameter interactive training, further optimizing the performance of the first treatment feedback matching model (Model A). This process will be repeated multiple times until the model performance reaches a satisfactory level.

[0107] In a possible embodiment, the counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence includes at least two, and each counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence corresponds to a first feedback content semantic association and a second feedback content semantic association.

[0108] Step A130 includes: Step A131, polling each counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence, and assigning a training importance factor to the corresponding counter-example respiratory disease treatment feedback data based on the training cost of the k-th counter-example respiratory disease treatment feedback data polled in this round, wherein the training importance factor is positively correlated with the training cost, and k is a positive integer and is not greater than the number of counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence.

[0109] Step A132: determining a parameter interaction training error parameter corresponding to the k-th counter-example respiratory disease treatment feedback data based on an error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content corresponding to the k-th counter-example respiratory disease treatment feedback data.

[0110] Step A133, after all the counter-example respiratory disease treatment feedback data are polled, the training error parameters of the first treatment feedback matching model are calculated based on the training importance factors of the counter-example respiratory disease treatment feedback data and the corresponding parameter interaction training error parameters.

[0111] Step A134: Optimize the weight information and bias information of the first treatment feedback matching model according to the optimization goal of minimizing the training error parameter.

[0112] During the iterative parameter interactive training of the respiratory disease treatment feedback matching model, the server needs to carefully process the negative examples in each training sequence to more effectively optimize the model's weights and biases. This scenario details how the server optimizes the first treatment feedback matching model (Model A) based on the error between the semantic relevance of the first and second feedback content, as well as the training cost of the negative examples.

[0113] In the mth respiratory disease treatment feedback data training sequence, there are at least two counterexample respiratory disease treatment feedback data. The server has calculated the first and second feedback content semantic relevance scores between these counterexample data and the target data using model A and model B, respectively.

[0114] The server starts polling each counterexample data in the mth training sequence. For each counterexample data polled (kth), the server assigns a training importance factor to it based on its training cost.

[0115] Specifically, the server maintains a record containing the training cost and an initially empty training importance factor field for each counterexample. During polling, the server reads the training cost of the kth counterexample and calculates its training importance factor based on pre-defined rules (such as linear or logarithmic mapping). The training importance factor is positively correlated with the training cost: a higher training cost indicates a larger training importance factor. The server stores the calculated training importance factor in the record for the corresponding counterexample.

[0116] For each counterexample data item (kth) polled, the server calculates the error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content to determine the error contribution of the counterexample data item during the parameter interaction training process. Specifically, the server reads the semantic relevance scores of the first and second feedback contents corresponding to the kth counterexample data item. The error between the two is calculated, which can be in the form of absolute error, squared error, etc. The calculated error value is stored as the parameter interaction training error parameter corresponding to the kth counterexample data item.

[0117] After all counterexample data are polled and assigned training importance factors, and the parameter interaction training error parameters are determined, the server calculates the overall training error parameters for the first therapeutic feedback matching model (Model A). Specifically, the server iterates through all records of counterexample data, reading the training importance factors and parameter interaction training error parameters for each counterexample data. The overall training error parameters are calculated using a weighted summation approach. Specifically, the parameter interaction training error parameters for each counterexample data are multiplied by their training importance factors, and all products are summed to obtain the total. This total is the overall training error parameter for Model A during the mth round of parameter interaction training.

[0118] Finally, the server adjusts the weights and biases of Model A based on the optimization goal of minimizing the overall training error parameter. Specifically, the server uses backpropagation or other optimization algorithms to optimize the parameters of Model A. During backpropagation, the overall training error parameter serves as the input to the loss function. The algorithm updates the weights and biases of Model A using gradient descent or other optimization strategies to reduce the overall training error parameter. After multiple rounds of iterative optimization, the server evaluates whether the performance of Model A has improved. If the improvement is significant, training continues with the next round. If the improvement is not significant or the preset training conditions (such as the maximum number of iterations, minimum error threshold, etc.) have been met, training stops and the optimized Model A is saved.

[0119] Through these steps, the server optimized the weights and biases of the first treatment feedback matching model (Model A) based on the error between the semantic relevance of the first and second feedback pieces and the training cost of the counterexample data. This process is the core of iterative parameter interaction training and is crucial for improving model performance.

[0120] In a possible implementation, before performing iterative parameter interactive training, the method further includes: Step B110: Obtain a first sample sequence, wherein the first sample sequence includes at least one first combined sample set, each first combined sample set including: target respiratory disease treatment feedback data, a first positive example corresponding to the target respiratory disease treatment feedback data, and a first negative example corresponding to the target respiratory disease treatment feedback data. The target respiratory disease treatment feedback data in different first combined sample sets may be different.

[0121] Step B120: performing feature comparison learning on the first treatment feedback matching model based on the first sample sequence.

[0122] In this embodiment, before iterative parameter interaction training, the server first performs feature comparison learning on the first treatment feedback matching model (Model A) to improve its initial performance. This step aims to help the model better learn how to distinguish semantic relevance features between different feedback contents by comparing the target respiratory disease treatment feedback data with its corresponding positive and negative examples.

[0123] The server first connects to a data warehouse or pre-prepared dataset to obtain the first sample sequence, which contains a series of carefully selected combined sample sets, each of which is built around a specific target respiratory disease treatment feedback data.

[0124] Specifically, the server reads the first sample sequence file or database table and loads it into memory. Each first combined sample set includes at least three parts: a target respiratory disease treatment feedback data set, first positive feedback data corresponding to the target data set, and first negative feedback data corresponding to the target data set. The target respiratory disease treatment feedback data sets in different combined sample sets are unique, ensuring data diversity and representativeness.

[0125] After receiving the first sample sequence, the server begins using these samples to perform feature comparison learning on the first treatment feedback matching model (Model A). This process primarily focuses on how the model distinguishes semantic differences between the target data and its positive and negative examples. Specifically, the server first preprocesses each feedback data point in the sample sequence, including text cleaning, word segmentation, and stop word removal, to prepare for subsequent feature extraction. Next, leveraging the existing architecture of Model A (e.g., a multi-layer neural network), the server extracts features from each feedback data point. These features typically represent the semantics of the text data and reflect the inherent correlations between the data. For each first combined sample set, the server performs feature comparisons between the target data, the positive example data, and the negative example data. These comparisons may include Euclidean distances or cosine similarities between feature vectors, aiming to quantify the semantic differences between the data. Based on the feature comparison results, the server calculates the model's loss (e.g., cross-entropy loss) for distinguishing between positive and negative examples. Then, using the backpropagation algorithm, the server adjusts the weights and bias parameters of Model A to minimize this loss and improve the model's discriminative ability.

[0126] The above process will be repeated for multiple iterations until the model's performance on the validation set reaches a stable state or the preset stopping condition is met. Each iteration will update the model parameters based on the new loss value, gradually optimizing the model's performance.

[0127] After completing feature comparison learning, the server needs to evaluate whether Model A's performance has improved. This is typically accomplished by testing the model on an independent validation set. Specifically, the server prepares a validation set containing target respiratory disease treatment feedback data and its positive and negative examples. Model A is used to make predictions on the samples in the validation set, and metrics such as the model's precision, recall, and F1 score for distinguishing positive and negative examples are calculated. These metrics are compared with the model's performance before feature comparison learning to evaluate the learning outcomes.

[0128] In one possible embodiment, the first treatment feedback matching model extracts semantic feature data from each of the two respiratory disease treatment feedback data, and uses the semantic feature matching degree between the two extracted semantic feature data as the semantic relevance of the feedback content between the two respiratory disease treatment feedback data. The second treatment feedback matching model fuses the two respiratory disease treatment feedback data to obtain fused respiratory disease treatment feedback data, estimates the semantic feature data of the fused respiratory disease treatment feedback data, and generates the semantic relevance of the feedback content between the two respiratory disease treatment feedback data based on the extracted semantic feature data.

[0129] In this embodiment, in the task of matching treatment feedback for respiratory diseases, the server uses two different models to evaluate the semantic relevance between feedback data: the first treatment feedback matching model (Model A) and the second treatment feedback matching model (Model B). These two models use different strategies when processing feedback data, but share the common goal of quantifying the similarity or relevance between feedback content.

[0130] The first treatment feedback matching model (Model A) extracts semantic feature data from two respiratory disease treatment feedback data sets and calculates the matching degree between these feature data to assess the semantic relevance of the feedback content between the two sets. Specifically, the server passes two respiratory disease treatment feedback data sets (denoted as Data A and Data B) as input to Model A. Model A's internal multi-layer neural network structure processes the input data, extracting semantic features layer by layer through components such as convolutional layers, pooling layers, and fully connected layers. These features are high-dimensional vectors that capture key information from the text data. After extracting the semantic feature vectors of Data A and Data B, Model A calculates the similarity or matching degree between the two vectors. This is typically achieved using methods such as cosine similarity and Euclidean distance, aiming to quantify the semantic proximity between the two feedback data sets. Based on the calculated feature matching degree, Model A outputs a relevance score, which indicates the strength of the semantic relevance between the feedback content of Data A and Data B.

[0131] Unlike Model A, the second treatment feedback matching model (Model B) employs a fusion strategy to assess the semantic relevance of feedback content between two respiratory disease treatment feedback datasets. Similarly, the server passes two respiratory disease treatment feedback datasets (Data A and Data B) as input to Model B. Model B first fuses Data A and Data B. This fusion can be performed through simple concatenation and weighted averaging, or through more complex attention mechanisms or recurrent neural network structures. The goal of this fusion is to generate a fused respiratory disease treatment feedback dataset that incorporates the combined information of both datasets. Next, Model B extracts semantic features from this fused dataset. This process is similar to that of Model A, employing a multi-layer neural network architecture to extract high-dimensional semantic feature vectors layer by layer. Because the fused dataset already incorporates the combined information of Data A and Data B, Model B can directly assess the semantic relevance of feedback content between the two original datasets based on the semantic features of this fused dataset. This is typically achieved by calculating the similarity or difference between the fused dataset's feature vector and a baseline vector (such as a zero vector or a task-specific prior knowledge vector). However, in practical applications, model B may directly output a score representing the strength of the association without explicitly calculating the similarity with the baseline vector. Ultimately, model B outputs a relevance score that reflects the strength of the semantic relevance of the feedback content of data A and data B from a fusion perspective.

[0132] In one possible implementation, the method further includes: Step S150 : obtaining a query sequence for a respiratory disease treatment feedback matching task, wherein the query sequence includes a plurality of reference treatment feedback data.

[0133] Step S160 : calling the target treatment feedback matching model to extract each reference treatment feedback data in the query sequence, and generating semantic feature data of each reference treatment feedback data.

[0134] Step S170 : When the respiratory disease treatment feedback data to be matched is obtained, the target treatment feedback matching model is used to extract the respiratory disease treatment feedback data to be matched, and semantic feature data of the respiratory disease treatment feedback data to be matched is generated.

[0135] Step S180 , respectively calculating the semantic feature matching degree between the semantic feature data of each reference treatment feedback data and the semantic feature data of the respiratory disease treatment feedback data to be matched, as the feedback content semantic relevance between the corresponding reference treatment feedback data and the respiratory disease treatment feedback data to be matched.

[0136] Step S190: Extract at least one reference treatment feedback data from the query sequence as reference treatment feedback data based on descending order of feedback content semantic relevance and based on the feedback content semantic relevance between each reference treatment feedback data and the respiratory disease treatment feedback data to be matched; and generate a matching result for the respiratory disease treatment feedback data to be matched based on the extracted at least one reference treatment feedback data.

[0137] In this example, within a medical institution, the server is responsible for matching respiratory disease treatment feedback, helping doctors quickly find historical treatment feedback data that is similar or related to the feedback data to be matched. This process involves multiple steps, including obtaining a query sequence, extracting semantic feature data, calculating the semantic feature matching degree, and generating matching results.

[0138] First, the server receives a request for a respiratory disease treatment feedback matching task. To respond to this request, the server first needs to retrieve a query sequence from the database. This query sequence contains multiple reference treatment feedback data sets, which are patient treatment feedback records collected over a period of time.

[0139] Specifically, the server retrieves relevant records from the treatment feedback database based on the parameters in the request (such as disease type, treatment time period, etc.), and organizes the retrieved records into a query sequence, which contains multiple reference treatment feedback data.

[0140] After receiving the query sequence, the server invokes the target treatment feedback matching model (optimized through iterative parameter interactive training) to process each reference treatment feedback data point in the sequence to extract its semantic feature data. Specifically, the server loads the target treatment feedback matching model into memory. It then iterates through each reference treatment feedback data point in the query sequence and inputs its text content into the target treatment feedback matching model. The target treatment feedback matching model processes this input data using its internal multi-layer neural network structure, extracting high-dimensional semantic feature vectors. The server stores the semantic feature vector corresponding to each reference treatment feedback data point for subsequent use.

[0141] At a certain point in time, the server obtains a new set of respiratory disease treatment feedback data to be matched. This set of respiratory disease treatment feedback data to be matched may be the latest treatment feedback record from the current patient.

[0142] Specifically, the server receives and stores the respiratory disease treatment feedback data to be matched, processes the respiratory disease treatment feedback data to be matched using the same target treatment feedback matching model, and extracts semantic feature data (semantic feature vector) thereof.

[0143] After extracting the semantic features of all reference treatment feedback data and respiratory disease treatment feedback data to be matched in the query sequence, the server begins to calculate the semantic feature matching between them to evaluate the semantic relevance between the feedback contents. In detail, the server traverses each reference treatment feedback data in the query sequence. For each reference treatment feedback data, the similarity or matching degree between its semantic feature vector and the semantic feature vector of the respiratory disease treatment feedback data to be matched is calculated. This can be achieved through methods such as cosine similarity and Euclidean distance. The calculated matching degree is stored as the feedback content semantic relevance score between the corresponding reference treatment feedback data and the respiratory disease treatment feedback data to be matched.

[0144] After evaluating the semantic relevance of feedback content between all reference treatment feedback data and the respiratory disease treatment feedback data to be matched, the server extracts at least one reference treatment feedback data from the query sequence as final reference treatment feedback data according to descending order of relevance strength, and generates a matching result.

[0145] Specifically, the server sorts the reference treatment feedback data in the query sequence based on the semantic relevance scores of the feedback content. The server extracts the reference treatment feedback data from the sorted sequence according to pre-defined rules (e.g., selecting the top N data with the highest scores). The extracted reference treatment feedback data is compiled into a matching result report, which may include detailed information about the reference treatment feedback data and the strength of its relevance to the respiratory disease treatment feedback data to be matched. The server then sends the matching result report to the requesting party (e.g., a physician's workstation) for further analysis and use.

[0146] In a possible implementation, step S190 includes: Step S191 , fusing the extracted reference treatment feedback data with the respiratory disease treatment feedback data to be matched, to generate fused data corresponding to the reference treatment feedback data.

[0147] Step S192 , calling the second treatment feedback matching model respectively to estimate the feedback content semantic relevance between the corresponding reference treatment feedback data and the respiratory disease treatment feedback data to be matched based on the fusion data corresponding to the reference treatment feedback data.

[0148] Step S193: According to the descending order of the semantic relevance of the feedback content and the semantic relevance of each feedback content extracted by the second treatment feedback matching model, the first T reference treatment feedback data are extracted from the at least one reference treatment feedback data as the target treatment feedback data, where T is a positive integer.

[0149] Step S194: generating a matching result of the respiratory disease treatment feedback data to be matched based on the extracted T target treatment feedback data.

[0150] In this embodiment, after extracting and initially evaluating the reference treatment feedback data, the server further processes this data to generate accurate matching results for the respiratory disease treatment feedback data to be matched. This process involves steps such as data fusion, re-evaluating semantic relevance using a second treatment feedback matching model, and extracting target treatment feedback data based on the strength of the relevance.

[0151] First, the server fuses each extracted reference treatment feedback data with the respiratory disease treatment feedback data to be matched to generate fused data containing comprehensive information of the two.

[0152] Specifically, for each selected reference treatment feedback data in the query sequence, the server merges its textual content with the textual content of the respiratory disease treatment feedback data to be matched. This merging can be done through simple concatenation or through intelligent fusion based on specific rules or algorithms, such as using an attention mechanism to highlight important information. The fused data not only incorporates the characteristics of the reference treatment feedback data, but also the characteristics of the respiratory disease treatment feedback data to be matched, providing a more comprehensive perspective for subsequent relevance assessment.

[0153] Next, the server calls the second treatment feedback matching model (model B) to process the fused data to re-evaluate the semantic relevance of the feedback content between each reference treatment feedback data and the respiratory disease treatment feedback data to be matched.

[0154] Specifically, the server loads Model B into memory. It then iterates over all fused data (each fused data point corresponds to a combination of reference treatment feedback data and the respiratory disease treatment feedback data to be matched). Each fused data point is input into Model B, which, using its unique fusion strategy and semantic feature extraction capabilities, generates a semantic feature vector for the fused data point. Based on these semantic feature vectors, Model B calculates the semantic relevance score between the fused data point and the respiratory disease treatment feedback data to be matched (from the fusion perspective).

[0155] After obtaining the semantic relevance scores of all fused data, the server extracts the first T reference treatment feedback data as target treatment feedback data according to the descending order of the scores.

[0156] Specifically, the server sorts all fused data by their semantic relevance scores. Based on a preset T value (a positive integer), the server extracts the reference treatment feedback data corresponding to the top T fused data with the highest scores from this sorted list. These extracted data are considered the target treatment feedback data that is semantically closest to or most relevant to the respiratory disease treatment feedback data to be matched.

[0157] Finally, the server generates matching results for the respiratory disease treatment feedback data to be matched based on the extracted T target treatment feedback data. Specifically, the server compiles information about these T target treatment feedback data, including their original text content and their relevance scores to the respiratory disease treatment feedback data to be matched. This information is organized into a matching result report, which may also include a brief analysis or recommendations for the target data. The server then sends the matching result report to the requesting party (such as a doctor's workstation or patient terminal) for further analysis, decision-making, or feedback.

[0158] Figure 2FIG. 1 shows the hardware structure of the artificial intelligence-based respiratory disease treatment feedback data processing system 100 for implementing the artificial intelligence-based respiratory disease treatment feedback data processing method provided in an embodiment of the present application. Figure 2 As shown, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0159] In one possible design, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can be a single server or a server group. The server group can be centralized or distributed (for example, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can be a distributed system). In some embodiments, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can be local or remote. For example, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the artificial intelligence-based respiratory disease treatment feedback data processing system 100 can be implemented on an artificial intelligence-based respiratory disease treatment feedback data processing system. By way of example only, the artificial intelligence-based respiratory disease treatment feedback data processing system can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, or the like, or any combination thereof.

[0160] The machine-readable storage medium 120 can store data and / or instructions. In some embodiments, the machine-readable storage medium 120 can store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 can store data and / or instructions used by the artificial intelligence-based respiratory disease treatment feedback data processing system 100 to execute or use to complete the exemplary methods described in this application.

[0161] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the artificial intelligence-based respiratory disease treatment feedback data processing method of the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0162] The specific implementation process of the processor 110 can be found in the various method embodiments executed by the above-mentioned artificial intelligence-based respiratory disease treatment feedback data processing system 100. The implementation principles and technical effects are similar and will not be repeated here in this embodiment.

[0163] In addition, an embodiment of the present application also provides a readable storage medium, which has computer-executable instructions preset in the readable storage medium. When the processor executes the computer-executable instructions, the above-mentioned artificial intelligence-based respiratory disease treatment feedback data processing method is implemented.

[0164] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of this application, multiple features are sometimes merged into one embodiment, figure or description thereof.

Claims

1. A method for processing respiratory disease treatment feedback data based on artificial intelligence, characterized in that: The method comprises: Obtaining a first treatment feedback matching model and a second treatment feedback matching model, wherein the first treatment feedback matching model and the second treatment feedback matching model are configured to determine feedback content semantic relevance between respiratory disease treatment feedback data; a model decision error of the first treatment feedback matching model is greater than a model decision error of the second treatment feedback matching model, and a model decision performance of the first treatment feedback matching model is greater than a model decision performance of the second treatment feedback matching model; Acquire multiple respiratory disease treatment feedback data training sequences, each of the respiratory disease treatment feedback data training sequences including target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data; the counterexample respiratory disease treatment feedback data in different respiratory disease treatment feedback data training sequences correspond to different training costs, the training costs being used to represent a complexity index for distinguishing the target respiratory disease treatment feedback data from the corresponding counterexample respiratory disease treatment feedback data; Determining the number of training iterations for each of the respiratory disease treatment feedback data training sequences, wherein the number of training iterations for any of the respiratory disease treatment feedback data training sequences is positively correlated with the training cost of counterexample respiratory disease treatment feedback data in the corresponding respiratory disease treatment feedback data training sequence; According to the ascending order of the number of training iterations, the first treatment feedback matching model is used to perform iterative parameter interactive training based on the multiple respiratory disease treatment feedback data training sequences and the second treatment feedback matching model to generate a target treatment feedback matching model.

2. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 1, characterized in that: The step of obtaining a plurality of respiratory disease treatment feedback data training sequences includes: Acquiring target respiratory disease treatment feedback data and a treatment feedback database, wherein the treatment feedback database includes a plurality of treatment feedback samples; Invoking a target treatment feedback matching model to estimate the feedback content semantic relevance between the target respiratory disease treatment feedback data and each treatment feedback sample in the treatment feedback database, and generating a target feedback content semantic relevance corresponding to each treatment feedback sample; wherein the target feedback content semantic relevance corresponding to any treatment feedback sample is positively correlated with the training cost of the corresponding treatment feedback sample; Extracting X treatment feedback samples from the treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data according to descending order of semantic relevance of target feedback content corresponding to each treatment feedback sample, where X is a positive integer; A plurality of respiratory disease treatment feedback data training sequences are configured based on the target respiratory disease treatment feedback data and the extracted X counter-example respiratory disease treatment feedback data.

3. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 2, characterized in that: When the target treatment feedback matching model is the first treatment feedback matching model, configuring multiple respiratory disease treatment feedback data training sequences based on the target respiratory disease treatment feedback data and the extracted X counter-example respiratory disease treatment feedback data includes: Sorting the extracted X counter-example respiratory disease treatment feedback data according to descending order of semantic relevance of target feedback content corresponding to the X counter-example respiratory disease treatment feedback data to generate a target respiratory disease treatment feedback data distribution; estimating feedback content semantic relevance between the target respiratory disease treatment feedback data and each counterexample respiratory disease treatment feedback data by using the second treatment feedback matching model, and generating candidate feedback content semantic relevance corresponding to each counterexample respiratory disease treatment feedback data; sorting the X counterexample respiratory disease treatment feedback data according to descending order of semantic relevance of candidate feedback content corresponding to each counterexample respiratory disease treatment feedback data to generate a candidate respiratory disease treatment feedback data distribution; Performing multiple rounds of data sampling in the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution according to a reverse sampling direction; wherein a respiratory disease treatment feedback data training sequence includes Y counter-example respiratory disease treatment feedback data, where Y is a positive integer, and one round of data sampling is used to extract Z counter-example respiratory disease treatment feedback data from the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution, where Z∈[Y / 2,X]; After the mth round of data sampling is completed, based on the Z counter-example respiratory disease treatment feedback data extracted from the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution in this round, a counter-example respiratory disease treatment feedback data list corresponding to the mth training iteration number is configured; According to the target respiratory disease treatment feedback data and the counterexample respiratory disease treatment feedback data list configured in this round, the respiratory disease treatment feedback data training sequence of the mth training iteration is configured, where m is a positive integer.

4. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 3, characterized in that: When Z is greater than Y / 2, the Z counter-example respiratory disease treatment feedback data extracted from the target respiratory disease treatment feedback data distribution and the candidate respiratory disease treatment feedback data distribution in this round are used to configure a counter-example respiratory disease treatment feedback data list corresponding to the mth training iteration number, including: Randomly selecting Q counter-example respiratory disease treatment feedback data from the Z counter-example respiratory disease treatment feedback data extracted from the target respiratory disease treatment feedback data distribution in this round; Randomly select W counter-example respiratory disease treatment feedback data from the Z counter-example respiratory disease treatment feedback data extracted from the candidate respiratory disease treatment feedback data distribution in this round; where Q and W are both positive integers, and Q+W=Y; According to the Q counter-example respiratory disease treatment feedback data and the W counter-example respiratory disease treatment feedback data, a counter-example respiratory disease treatment feedback data list corresponding to the m-th training iteration number is configured.

5. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to any one of claims 2 to 4, characterized in that: The target respiratory disease treatment feedback data is configured with positive respiratory disease treatment feedback data; Extracting X treatment feedback samples from the treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data according to descending order of semantic relevance of target feedback content corresponding to each treatment feedback sample includes: Optimizing the treatment feedback database based on the positive respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data to generate an optimized treatment feedback database; wherein the data optimization includes: removing treatment feedback samples in the treatment feedback database that are identical to the positive respiratory disease treatment feedback data; According to the descending order of the semantic relevance of the target feedback content corresponding to each treatment feedback sample, X treatment feedback samples are extracted from the optimized treatment feedback database as counterexample respiratory disease treatment feedback data corresponding to the target respiratory disease treatment feedback data.

6. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 1, characterized in that: The respiratory disease treatment feedback data training sequence with a training iteration number of m is expressed as the m-th respiratory disease treatment feedback data training sequence, where m is a positive integer; The method of performing the mth round of parameter interactive training based on the mth respiratory disease treatment feedback data training sequence and the second treatment feedback matching model using the first treatment feedback matching model includes: estimating feedback content semantic relevance between target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence using the first treatment feedback matching model to generate a first feedback content semantic relevance; estimating feedback content semantic relevance between target respiratory disease treatment feedback data and counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence using the second treatment feedback matching model to generate a second feedback content semantic relevance; Optimizing weight information and bias information of the first treatment feedback matching model based on an error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content.

7. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 6, characterized in that: The counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence includes at least two counter-example respiratory disease treatment feedback data, and each counter-example respiratory disease treatment feedback data in the m-th respiratory disease treatment feedback data training sequence corresponds to a first feedback content semantic relevance and a second feedback content semantic relevance; Optimizing the weight information and bias information of the first treatment feedback matching model based on the error between the semantic relevance of the first feedback content and the semantic relevance of the second feedback content includes: polling each counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence, and assigning a training importance factor to the corresponding counterexample respiratory disease treatment feedback data according to the training cost of the kth counterexample respiratory disease treatment feedback data in this round of polling, wherein the training importance factor is positively correlated with the training cost, and k is a positive integer and is not greater than the number of counterexample respiratory disease treatment feedback data in the mth respiratory disease treatment feedback data training sequence; determining a parameter interaction training error parameter corresponding to the k-th counterexample respiratory disease treatment feedback data based on an error between a first feedback content semantic relevance and a second feedback content semantic relevance corresponding to the k-th counterexample respiratory disease treatment feedback data; After all the counter-example respiratory disease treatment feedback data are polled, calculating the training error parameter of the first treatment feedback matching model according to the training importance factor of each counter-example respiratory disease treatment feedback data and the corresponding parameter interaction training error parameter; The weight information and bias information of the first treatment feedback matching model are optimized according to the optimization goal of minimizing the training error parameter.

8. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 1, characterized in that: The method further comprises: Obtaining a query sequence retrieved for a respiratory disease treatment feedback matching task, wherein the query sequence includes a plurality of reference treatment feedback data; calling a target treatment feedback matching model to extract each reference treatment feedback data in the query sequence to generate semantic feature data of each reference treatment feedback data; When the respiratory disease treatment feedback data to be matched is obtained, the target treatment feedback matching model is used to extract the respiratory disease treatment feedback data to be matched, and semantic feature data of the respiratory disease treatment feedback data to be matched is generated; respectively calculating the semantic feature matching degree between the semantic feature data of each reference treatment feedback data and the semantic feature data of the respiratory disease treatment feedback data to be matched, as the feedback content semantic relevance between the corresponding reference treatment feedback data and the respiratory disease treatment feedback data to be matched; extracting at least one reference treatment feedback data from the query sequence as reference treatment feedback data according to a descending order of feedback content semantic relevance and according to feedback content semantic relevance between each reference treatment feedback data and the respiratory disease treatment feedback data to be matched; A matching result of the respiratory disease treatment feedback data to be matched is generated based on the extracted at least one reference treatment feedback data.

9. The method for processing respiratory disease treatment feedback data based on artificial intelligence according to claim 8, characterized in that: Generating a matching result of the respiratory disease treatment feedback data to be matched based on the extracted at least one reference treatment feedback data includes: fusing the extracted reference treatment feedback data with the respiratory disease treatment feedback data to be matched, respectively, to generate fused data corresponding to the reference treatment feedback data; respectively calling the second treatment feedback matching model to estimate the feedback content semantic relevance between the corresponding reference treatment feedback data and the respiratory disease treatment feedback data to be matched based on the fusion data corresponding to the reference treatment feedback data; Extracting, in descending order of semantic relevance of the feedback content and according to the semantic relevance of each feedback content extracted by the second treatment feedback matching model, first T reference treatment feedback data from the extracted at least one reference treatment feedback data as target treatment feedback data, where T is a positive integer; A matching result of the respiratory disease treatment feedback data to be matched is generated based on the extracted T target treatment feedback data.

10. A respiratory disease treatment feedback data processing system based on artificial intelligence, characterized in that: The artificial intelligence-based respiratory disease treatment feedback data processing system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the artificial intelligence-based respiratory disease treatment feedback data processing method described in any one of claims 1 to 9 above.

Images