Data interaction method and related products thereof

Abstract

The invention discloses a data interaction method and related products thereof, and the method comprises the steps: carrying out the interpretation of obtained source data, and generating basic interpretation data; understanding the basic interpretation data by using the data understanding large model, and generating application understanding data expressing potential contents of the basic interpretation data; and in response to a user demand, processing the application understanding data and / or the basic interpretation data by using the data processing logic generated by the demand analysis large model to generate a data processing result. According to the scheme, the basic interpretation data can be understood by utilizing the large language model, and the application understanding data containing the potential content of the data is generated. The understanding data is used as the application link layer to establish a deepened link between the basic data layer and the demand response layer, so that accurate response to the user demand is realized, and the effective processing and generation speed of the data for the user demand is improved.

Description

Technical Field

[0001] This solution relates to the field of data processing technology. More specifically, it relates to a data interaction method based on a large model and related products. Background Technology

[0002] Large Language Models (LLMs), often shortened to "large models," typically refer to deep learning models with a large number of parameters in the field of artificial intelligence. With the development and advancement of natural language processing and artificial intelligence technologies, large models have been applied in various industries. Leveraging their powerful text understanding and analysis capabilities, large models can effectively process and analyze massive amounts of complex data, providing strong support for data-driven information development.

[0003] Data is the carrier of information; it is an objectively existing symbolic record. Through analysis and processing, a wealth of readable information can be presented for people's reference and use. As the amount of data that needs to be transmitted in people's daily lives and work gradually increases, the difficulty of data processing also increases.

[0004] Traditional data interaction involves setting pointers to data, storing it in formatted tables or stacks, and then retrieving the data via pointers when processing is needed. This traditional method is only suitable for small datasets (below the hundreds of millions). With the dramatic increase in data volume (reaching billions or even tens of billions), traditional methods can no longer meet the current demands for efficient data interaction and processing. Summary of the Invention

[0005] The purpose of this invention is to provide a model-based data interaction method and related products to solve the problems of poor data accuracy and low efficiency of interactive data processing in existing interaction processes.

[0006] To achieve the above objectives, the following technical solution is adopted: Firstly, this solution provides a data interaction method, the steps of which include: The acquired source data is interpreted to generate basic interpretation data; By using large-scale data understanding models to understand basic interpretation data, application understanding data is generated that expresses the potential content of the basic interpretation data. In response to user needs, the system utilizes data processing logic generated from the large model of needs analysis to process application-understanding data and / or basic interpretation data, generating data processing results.

[0007] In some instances, the step of interpreting the acquired source data to generate basic interpretation data includes: Obtain source data; Use feature information to identify models or interpret large models, and identify key data in the source data; Key data is labeled to obtain multiple key pieces of information as the basis for interpretation.

[0008] In some instances, the steps of using a large data understanding model to understand basic interpretive data and generate application-understanding data that expresses the potential content of the basic interpretive data include: By leveraging large-scale data understanding models, combined with practical experience and / or technical guidelines in the application domain, we can perform understanding and mining on basic interpretive data to generate application understanding data that contains the underlying logic and / or potential information of the data.

[0009] In some instances, the steps of responding to user needs and utilizing data processing logic generated from a large demand analysis model to process application-understanding data and / or basic interpretation data to generate data processing results include: In response to user needs, the application understands the data by using a large-scale requirement analysis model to generate data operation execution requirements; Based on the data operation execution requirements, the basic understanding data is analyzed using the requirement analysis model to generate data processing logic; Using the aforementioned data processing logic, the application-understanding data and / or basic interpretation data are processed to generate data processing results for answering user needs.

[0010] Secondly, this solution provides a cloud server, which includes: The first intelligent agent interprets the acquired source data and generates basic interpretation data; The second intelligent agent uses a large data understanding model to understand the basic interpretation data and generate application understanding data that expresses the potential content of the basic interpretation data. The third intelligent agent responds to user needs and uses data processing logic generated by the large model of demand analysis to process application-understanding data and / or basic interpretation data, generating data processing results.

[0011] In some instances, the first agent is used to perform the following steps: Obtain source data; Use feature information to identify models or interpret large models, and identify key data in the source data; Key data is labeled to obtain multiple key pieces of information as the basis for interpretation.

[0012] In some instances, the second agent is used to perform the following steps: By leveraging large-scale data understanding models, combined with practical experience and / or technical guidelines in the application domain, we can perform understanding and mining on basic interpretive data to generate application understanding data that contains the underlying logic and / or potential information of the data.

[0013] In some instances, the third agent is used to perform the following steps: In response to user needs, the application understands the data by using a large-scale requirement analysis model to generate data operation execution requirements; Based on the data operation execution requirements, the basic understanding data is analyzed using the requirement analysis model to generate data processing logic; Using the aforementioned data processing logic, the application-understanding data and / or basic interpretation data are processed to generate data processing results for answering user needs.

[0014] Thirdly, this solution provides a data interaction device, comprising: a memory, one or more processors; the memory and processors are connected via a communication bus; the processors are configured to execute instructions in the memory; the memory stores instructions for performing the steps of the methods described in any of the preceding claims.

[0015] Fourthly, this solution provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method as described in any of the preceding claims.

[0016] The beneficial effects of this invention are as follows: This solution leverages a large language model to understand basic interpretive data and generate application-understanding data containing potential data content. By using this application-understanding data as an application linking layer, it establishes a deeper connection between the underlying basic data layer and the user demand response layer, enabling precise responses to user needs and improving the speed of effective data processing and generation tailored to user requirements. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram illustrating an example of the data interaction method described in this scheme is shown. Figure 2 This diagram illustrates an example of generating basic interpretation data as described in this scheme. Figure 3A schematic diagram illustrating an example of generating the data processing result described in this scheme; Figure 4 This diagram illustrates the interaction process between user requirements and underlying data as described in this solution. Figure 5 A schematic diagram of the cloud server described in this solution is shown; Figure 6 A schematic diagram of the data processing device described in this solution is shown. Detailed Implementation

[0019] To make the present invention, its technical solutions, and advantages clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0020] Analysis and research into existing data interaction problems reveal that with the development of science and technology, the amount of data generated daily is increasing, and the frequency of data interaction is also rising. Traditional data interaction requires preprocessing and compiling the collected data before storing it in a formatted database. This storage process is essentially a simple storage of formatted data and cannot provide in-depth management based on the potential content expressed by the data. When a user needs the data processing results, the system first searches for pointer tags to index the data based on keywords in the user's requirements, and then retrieves the required data from the database using these pointer tags. Then, the data undergoes complex processing to obtain the results. This approach suffers from slow response times and long processing times when managing large-scale (hundreds of millions) or ultra-large-scale (tens of billions) data sets, failing to meet users' demands for accuracy and efficiency in data interaction.

[0021] Therefore, this solution aims to provide a data interaction method based on a large model and related products. This solution leverages a large language model to understand the potential content of data and generate descriptive information that expresses this potential content. Through this descriptive information, a deeper connection is established between the underlying data and user needs, enabling precise responses to user demands and improving the speed of effective data processing tailored to those needs.

[0022] The following section will describe in detail a data interaction method proposed in this solution, with reference to the accompanying drawings. Specifically, as shown in the attached figures... Figure 1 As shown, the data interaction method includes: S1. Interpret the obtained source data to generate basic interpretation data; S2. Utilize large-scale data understanding models to understand basic interpretation data and generate application understanding data that expresses the potential content of the basic interpretation data; S3. Respond to user needs and utilize the data processing logic generated by the large model of needs analysis to process application-understood data and / or basic interpretation data, and generate data processing results.

[0023] Source data refers to data directly derived from actual business activities, observation processes, inspection processes, and other application scenarios within the domain in which this method is applied. It is typically unprocessed raw data or data that has undergone preliminary processing. Source data can be single data entries or datasets in various forms, such as numbers, text, images, videos, and audio. Source data contains a wealth of detailed information, with data volumes reaching tens of millions (1GB) or hundreds of millions (10GB) or more. Source data is fundamental in data interaction; a deep understanding and mining of the potential content within the source data enables faster and more accurate data processing. This solution utilizes a cloud server as the carrier for the data interaction method. The cloud server can acquire source data either by retrieving it from the data acquisition device or the local database where the source data resides, after user authorization, or by having the user upload the source data to the cloud server via a terminal device. Depending on the importance and volume of the source data, real-time, scheduled, or periodic transmission methods can be used. During source data transmission, environmental information such as the data acquisition time, location, user identity, and acquisition device number can also be uploaded to the cloud server, making the source of the information clearer.

[0024] After obtaining the source data, it is first interpreted to generate basic interpretation data. Specifically, such as... Figure 2 As shown, the basic steps for interpreting data may include: S101, Obtain source data; S102. Use feature information to identify models or interpret large models to identify key data in source data; S103. Annotate key data to obtain multiple key information as basic interpretation data.

[0025] In this solution, a feature information recognition model or a large-scale data interpretation model is used to extract key data such as key features, keywords, key values, and key image regions of interest to users from the source data. Then, using annotation methods such as arrows, highlighting, and bounding boxes, the key data is annotated to obtain multiple pieces of key information. This key information is then used as the foundation for interpretation.

[0026] Depending on the application, users' objects of interest can be numbers, text, images, video frames, audio clips, and so on. For example, in the medical field, users might be interested in textual descriptions of lesions in diagnostic reports; or medical images of key areas such as physiological tissue segmentation, key points, vascular midlines, and lesion signs. In the computer field, users might be interested in a segment of binary encoded information. In physical education, users might be interested in the image features of an athlete's posture during exercise.

[0027] Basic interpretive data is the underlying data in the data exchange process with users. It not only contains important information about the object of interest but also expresses a great deal of potential content. This potential content needs to be understood using a large language model. Specifically, a large data understanding model can be used, combined with practical experience in the application domain and / or technical guidelines, to understand and mine the basic interpretive data, generating application understanding data that includes potential logic, potential information, and other potential content.

[0028] In some instances, applied understanding data can be the latent information of an object of interest described in textual form. This latent information can reveal important details in the basic interpretation data that can only be analyzed or inferred through specialized knowledge. For example, this method is applied to the healthcare field, with the source data being CT medical image sequences of the liver. By interpreting the source data using a feature information recognition model or a large-scale data interpretation model, basic interpretation data such as abnormal tissue areas in the medical images can be obtained. Ordinary patients can only distinguish the liver and surrounding normal tissue areas from images, using the brightness and darkness of the images to distinguish the boundaries of various physiological tissues. However, ordinary patients are not sensitive to abnormal areas in medical images. Even patients with some medical experience can only identify the approximate location of abnormal areas in medical images. In this case, a large-scale data interpretation model can be used to understand and mine the basic interpretation data, obtaining applied understanding data such as changes, trends, and implicit features of the object of interest to the patient.

[0029] In some instances, application-understanding data can be the underlying logic of data about objects of interest described in text form. This underlying logic reveals relationships or logical connections between various basic interpretation data that can only be identified through extensive reasoning or practice. For example, this method is applied to the education field, where source data includes scanned test papers, test score statistics, student answer times, types of solutions, and knowledge points covered in the questions. Typically, test answer times are closely related to question difficulty. However, besides difficulty affecting answer times and scores, are there other factors that influence these factors? Teachers need to review numerous test papers, combine this information with various other sources, and rely on their teaching experience to determine these factors. In this case, feature recognition models or large-scale data interpretation models can be used to interpret the source data, generating basic interpretation data such as answer times, test scores, and knowledge point types. Then, using large-scale data interpretation models, combined with teachers' practical experience and knowledge of knowledge point difficulty types, the basic interpretation data is analyzed and mined to generate application-understanding data containing underlying data relationships. Application-understanding data can demonstrate: 1. The score on the exam is actually related to the degree of coverage of knowledge points. Usually, students have a higher grasp of important knowledge points and a higher accuracy rate in answering them, while students have a lower grasp of less important knowledge points, resulting in a lower accuracy rate in answering them, which affects the score on the exam.

[0030] 2. The time allotted for answering the questions actually depends on the types of solutions available for each question. A question may have multiple solutions, but some solutions are simpler and require less time, while others are more complex and require more time, thus resulting in different time allotted for answering the questions.

[0031] Understanding data through these applications, which contain underlying logical relationships, allows users to clearly and quickly grasp the underlying logic reflected in the data.

[0032] In some instances, application-understanding data can be statistical analysis information of key data within an object of interest, presented in a non-textual format. This statistical analysis information can visually represent the potential condition of the object of interest through text, images, or other means. For example, this method is applied to the medical field, where the source data is a sequence of medical images. Feature information recognition models or large-scale data interpretation models can be used to interpret the source data, generating key data such as physiological tissue segmentation, key points, vascular centerlines, and lesion signs. Then, the key data is labeled to generate basic interpretation data. Finally, using a large-scale data understanding model, combined with clinical experience and medical guidelines, statistical analysis is performed on the basic interpretation data to generate application-understanding data such as statistical information and trend analysis of the object of interest.

[0033] It should be noted that the examples of potential content described above are non-limiting. Those skilled in the art should understand that the aforementioned examples of potential content should be considered exemplary illustrations of this solution. The potential content described above can be adjusted during the training of a large model according to the application domain and the user's desired objects of interest, so that the trained large model can adapt to the application domain and user needs.

[0034] In this solution, at least one of the large-scale data interpretation model and the large-scale data understanding model can be pre-trained on a large-scale general dataset based on deep learning. Then, using a smaller, specialized dataset and user-defined or user-defined target data, the pre-trained large model is adapted to a specific task or domain. User feedback on the large model's output guides further optimization, ultimately resulting in a large model for the corresponding application domain. The specialized dataset can be practical experience, technical guidelines, etc., within the application domain. In some examples, the large-scale data interpretation model can be based on deep learning neural networks or Transformer models. It can be pre-trained using general medical reports and general medical images datasets. Then, using clinical experience, medical guidelines, expert-annotated medical images, doctor's diagnostic reports, etc., as specialized datasets, and user-defined or user-defined features, keywords, key images, etc., as target data, the pre-trained large model can acquire the ability to process medical-related tasks. Feedback from professional doctors on the large model's output is then used to optimize the model, ultimately resulting in a large-scale data interpretation model capable of interpreting source data in the medical field. The large-scale data understanding model can be based on deep learning neural networks. This approach utilizes the latent logic and information inherent in general medical data for pre-training. Then, it leverages the latent content existing between clinical experience, expert-annotated medical images, and physician diagnostic reports—practice data—as a professional dataset. Simultaneously, it uses user-defined or user-defined disease and abnormality data as target data, enabling the pre-trained large-scale model to acquire the ability to understand relevant latent content in the medical field. Feedback from doctors, experts, and other medical professionals is then used to optimize the large-scale model's output, ultimately resulting in a data understanding model capable of interpreting basic data. Furthermore, the feature information recognition model described in this solution can be constructed using a deep learning neural network as a foundation. Key regions such as physiological tissue segmentation, key points, vascular centerlines, and lesion signs marked in medical images are used as input. Through multiple training iterations, a feature information recognition model capable of interpreting key information from the source data is obtained.

[0035] The data interaction method described in this solution is applicable to various fields such as information statistical analysis, biology, engineering and manufacturing, energy, materials, environment, aerospace, communications, medicine, and education. When training the large-scale information interpretation model, providing a large-scale general dataset and professional knowledge relevant to the corresponding field is sufficient to complete the training.

[0036] As users' daily lives and work pace accelerate, a large amount of new data is generated. Therefore, it is necessary to update the source data to meet users' needs for accurate and timely analysis results. Depending on the importance of the data or user needs, cloud servers can be configured to poll new source data in real-time, on a scheduled basis, or periodically. The cloud server obtains new source data as update data through the polling mechanism. Then, a large-scale data interpretation model is used to interpret the updated data, generating basic interpretation data. At this point, the basic interpretation data of the source data and the basic interpretation data of the updated data can be merged to form updated basic interpretation data. The large-scale data understanding model is then used to further analyze and mine the updated basic interpretation data, generating updated application understanding data. This updated application understanding data allows users to more intuitively understand the changes and development trends of objects of interest.

[0037] In some instances, the source data is a patient's physical examination data, including medical imaging data of the kidney area. A large-scale data interpretation model is used to interpret and describe the source data, obtaining basic interpretation data. This basic interpretation data is labeled with relevant characteristics of the patient's ureteral stones. For example, the number of ureteral stones is single; the stone size is 4mm*3mm. The large-scale data understanding model is then used to analyze and understand this basic interpretation data, yielding applied understanding data. This applied understanding data analyzes the risk level and causes of the patient's ureteral stones. For example, a stone size not exceeding 5mm is considered a mild case. The cause of the stones might be due to prolonged sitting or insufficient fluid intake. Several months later, the patient undergoes a follow-up examination, obtaining updated data. After the cloud server recurs and retrieves the updated data, the large-scale data interpretation model is used to interpret it, obtaining updated basic interpretation data. This updated basic interpretation data is labeled with updated information regarding the relevant characteristics of the patient's ureteral stones. For example, the patient had a single ureteral stone measuring 4mm x 5mm; and two stones in the lower calyx of the right kidney, with the largest stone measuring 8mm in diameter. The updated baseline interpretation data and the source data were merged to form merged baseline interpretation data. Then, a large-scale data understanding model was used to perform interpretation mining on the merged baseline interpretation data to generate updated application understanding data. This updated application understanding data provided a new understanding and analysis of the risk level and causes of the patient's ureteral stones. For example, the patient's ureteral stone size increased, and a large new stone formed in the lower calyx of the right kidney, indicating a severe kidney stone condition with risks of urinary tract obstruction, infection, secondary hematuria, hydronephrosis, and potentially causing severe pain. The significant increase in the number and size of stones over several months may be related to prolonged sitting and insufficient water intake, and may also be related to high levels of calcium oxalate, salt, or excessive protein intake in the patient's long-term diet.

[0038] The core purpose of acquiring basic interpretation data and application understanding data is to better serve user needs. Therefore, this solution responds to user requests by using application understanding data in the application linking layer as a link to filter the necessary basic interpretation data from the basic data layer. Then, in the request response layer, the application understanding data and / or basic interpretation data are processed to generate data processing results that answer user requests. Specifically, as follows... Figure 3 As shown, in response to user needs, the data processing logic generated from the large model of needs analysis processes the application-understanding data and / or basic interpretation data to generate data processing results. The steps include: S301. Respond to user needs by using the large-scale requirement analysis model to analyze the application's understanding data and generate data operation execution requirements. S302. Based on the data operation execution requirements, use the requirement analysis model to analyze the basic understanding data and generate data processing logic. S303. Using the data processing logic, process the application understanding data and / or basic interpretation data to generate data processing results.

[0039] In this solution, users can input their needs via voice or text. Based on these needs, a demand parsing model, combined with the application domain's knowledge system, is used to analyze the application understanding data and generate data operation execution requirements. In some examples, users need to understand the causes and treatments of kidney stones. "Causes and treatments of kidney stones" can be used as the user's demand, triggering the analysis of the application understanding data using the demand parsing model. During this analysis, data such as the patient's current risk level of kidney stones and inferred causes can be obtained. However, simply providing this data as a processing result is insufficient to fully address the user's needs. More processing results are needed, such as different types of kidney stones and their impact on the body; lifestyle details to consider beyond the main causes of kidney stones; treatment recommendations and rehabilitation plans. Therefore, generating data operation execution requirements triggers further processing of the basic understanding data, leading to more comprehensive data processing results. Based on the data operation execution requirements, the basic understanding data is analyzed using a large-scale requirement analysis model to generate data processing logic that meets the user's needs. In some instances, the data processing logic includes a general processing flow of data search, data filtering, and data processing steps. In this case, the search, filtering, and processing of application understanding data and basic interpretation data are executed sequentially according to the general processing flow. If the data processing steps include multiple sub-processing steps that process basic interpretation data, then these sub-processing steps can be used to process the corresponding basic interpretation data during the execution of the data processing steps. The application understanding data and / or basic interpretation data are analyzed and processed according to the data processing logic to obtain multiple processing results, which are then sent to the user. If the user adds a data integration requirement, then after executing the data processing logic of the search, data filtering, and data processing steps, it is necessary to continue with the step of integrating multiple processing results to obtain an integrated processing result, which is then sent to the user. Therefore, the "clumsy method" interaction can be based on user needs and executed precisely according to the data processing logic to obtain processing results that answer the user's requirements. Data processing results used to answer user needs can include: textual descriptions such as analysis results, guidance, and implementation suggestions; graphic information such as data charts, trend graphs, and presentation slides; and dynamic information such as audio and video.

[0040] In this approach, the large-scale demand parsing model can be based on deep learning, pre-trained using a large-scale, general-purpose demand parsing dataset and a set of data processing logic corresponding to the demands. Then, it is trained using a smaller dataset specific to the application domain, enabling the large model to adapt to demand parsing tasks in that domain. Finally, user feedback on the model's output guides further optimization, ultimately resulting in the final large-scale demand parsing model.

[0041] This solution leverages application-understanding data to deeply link underlying data with user needs. Through data processing logic generated from the parsing of user needs, it processes basic interpreted data and / or application-understanding data to produce data processing results, thereby achieving a precise response to user needs and improving the speed of effective data processing and generation tailored to user requirements.

[0042] Example 1 In this example, as Figure 4 As shown, taking medical data interaction in the medical field as an example, the data interaction method will be further explained.

[0043] A patient, experiencing heart discomfort, underwent cardiac magnetic resonance imaging (CMR) of the heart region on the doctor's recommendation, obtaining a dataset of cardiac MRI images. Based on the image data, the doctor provided a structured report diagnosing conditions such as myocardial ischemia, hypertension, and arrhythmia.

[0044] With the patient's authorization, the hospital uploads the cardiac MRI image dataset and structured report as source data to the cloud server. The cloud server uses a large-scale data interpretation model to interpret the source data, obtaining basic interpretation data. This basic interpretation data forms the foundational data layer in the interaction method described in this example, containing multiple key data points related to the patient's heart organ status and disease characteristics.

[0045] By leveraging a large-scale data understanding model to analyze and mine basic interpretive data, application understanding data is generated. This application understanding data serves as crucial data for the application linking layer, establishing a deep connection between the basic data layer and the demand response layer. Application understanding data can contain potential content that expresses the basic interpretive data. For example, application understanding data might include: "...the patient has problems such as myocardial ischemia and hypertension, mainly caused by coronary atherosclerosis. Based on the patient's MRI image data analysis, the patient may experience symptoms such as angina pectoris and fatigue, and has a high risk of coronary heart disease...". Potential content from application understanding data could also include: "...the patient's MRI image data shows abnormalities in myocardial structure, indicating a risk of cardiomyopathy, requiring periodic follow-up examinations...".

[0046] Six months later, the patient underwent a routine full-body checkup, obtaining a dataset of scan images including echocardiogram, neck Doppler ultrasound, and lung CT scans. Based on the image data, the doctor provided a structured report with the diagnostic results for each examined area. With the patient's authorization, the hospital uploaded the scan image dataset and structured report as new source data to a cloud server. The cloud server retrieved new source data through a polling mechanism as updated data, and used a large-scale data interpretation model to interpret the updated data, generating basic interpretation data. This basic interpretation data included multiple key data points regarding the patient's heart-related organ status and disease characteristics, neck-related organ status and disease characteristics, and lung-related organ status and disease characteristics, among others. At this point, the basic interpretation data of the updated data and the basic interpretation data of the source data were merged to form merged basic interpretation data. Then, the large-scale data understanding model was used to perform interpretation mining on the merged basic interpretation data to generate updated application understanding data. The updated application understanding data reveals the following potential findings: "...the patient's MRI images show abnormalities in myocardial structure. Echocardiography shows left atrial wall thickening, indicating a risk of cardiomyopathy,...". The updated application understanding data also includes: "...carotid artery stenosis of less than 30% may cause stroke or transient ischemic attack. Carotid artery stenosis can lead to carotid atherosclerosis. The patient also has myocardial ischemia and hypertension, and is at risk of coronary atherosclerosis. Therefore, the patient has a higher risk of coronary artery disease...".

[0047] In this example, the core purpose of acquiring basic interpretation data and application understanding data is to accelerate data analysis and processing at the demand response layer based on user needs. Specifically, the user submits a request via voice or text: "How is my heart condition? How can I maintain my heart health?" The cloud server, based on the user's request, utilizes a demand parsing model combined with medical knowledge to analyze the application understanding data and generate data operation execution requirements. Through analyzing the application understanding data, information such as the user's risk level, causes, and potential complications related to heart disease can be obtained. It is also necessary to provide data processing results such as medical treatment and medication recommendations, and lifestyle and health planning suggestions. Therefore, data operation execution requirements are generated to trigger further processing operations on the application understanding data and / or basic interpretation data.

[0048] Based on the data operation execution requirements, a large-scale requirement analysis model is used to analyze the basic understanding data and generate data processing logic that meets the user's needs. This data processing logic includes data search, data filtering, and data analysis steps. Following this logic, the cloud server searches for medical data directly or indirectly related to the heart, based on the application's understanding of the data. For example, data related to the heart and neck. Then, it filters medical data related to heart conditions, such as data on abnormal heart organ status, heart disease data, and neck lesion data. Finally, a medical data integration and processing model, pre-trained based on data integration and processing technology, integrates and processes the application's understanding data and / or the filtered basic interpretation data, generating data processing results presented in various formats, including text, charts, graphic reports, PowerPoint presentations, audio, and video. These results provide users with information on their current heart health status, disease risk level, disease progression trends, medical treatment recommendations, and lifestyle health planning. For example, the data processing results might include: "...The patient has coronary atherosclerosis and carotid atherosclerosis, which can easily affect blood supply to the brain, causing headaches, dizziness, and even a risk of mild coma. It is recommended that the patient maintain a healthy diet, moderate exercise, and a good mental state. At the same time, regular monitoring of blood pressure, blood lipids, and blood sugar levels is necessary to promptly identify and manage potential risk factors..."

[0049] As demonstrated in Example 1, application-understanding data can more clearly and accurately reveal the potentially important information within medical data. It enables the creation of a deeper link between underlying data and user needs. When a user submits a request, application-understanding data can more accurately and quickly locate key data information in the source data, achieving a precise response to the user's needs and improving the speed of effective data processing for user requests.

[0050] Example 2 In this example, the interaction of learning materials in the field of education will be used as an example to further illustrate the data interaction method.

[0051] Plant protection organizations use terminal devices to upload images, text descriptions, pictures, and other relevant teaching materials about flowers as source data to a cloud server. The cloud server then uses a large-scale data interpretation model to analyze the source data and generate basic interpretation data. This basic interpretation data includes: the flower's name, variety, characteristics, historical origin, knowledge system of the flower variety, and related information. For example, basic interpretation data for peonies might include: Flower name: Peony, also known as: National Beauty and Heavenly Fragrance, Luoyang Flower, Tang Flower, etc.

[0052] Flower varieties: Plants of the genus Paeonia in the family Paeoniaceae.

[0053] Types of flower petals: single-petaled, semi-double-petaled, and double-petaled.

[0054] Related videos of peonies: In video number 001, the area at coordinates (x1, y1) and (x2, y2) is a sunflower-type peony, filmed at Luoyang International Peony Garden. In video number 002, the area at coordinates (x3, y3) and (x4, y5) is a pavilion-type peony, filmed at the Crested Ibis Pear Garden Peony Garden.

[0055] Basic data for interpreting peony flowers may also include: A knowledge system encompassing the biological, cultural, and economic value of peonies.

[0056] Information about the history of peonies, information about related figures, information about related cultural relics, etc.

[0057] Using large-scale data understanding models, we can analyze and mine basic interpretive data to generate application understanding data. Application understanding data can contain potential content that expresses the basic interpretive data. For example, application understanding data may include, but is not limited to, the following: Botanical characteristics: Peony is a deciduous shrub, usually 1-2 meters tall, with large, fragrant flowers in a variety of colors, including red, pink, white, and yellow. It typically blooms in May. Peony petals are mostly double, with a beautiful shape, and are often used in horticulture and decoration.

[0058] Historical Characteristics: Records indicate that peonies were first discovered in the Qinling and Bashan Mountains. Hanzhong was the earliest place to cultivate peonies; as early as the Spring and Autumn and Warring States periods, there were records of peonies in the Book of Songs. During the Qin and Han dynasties, peonies were already used in medicine, so their aesthetic and medicinal value was recognized very early on. Large-scale cultivation gradually developed during the Sui dynasty. It flourished during the Tang and Song dynasties, with many famous poets leaving behind timeless verses about peonies…

[0059] Cultural Symbolism: In Chinese culture, the peony is regarded as a symbol of wealth and prosperity, frequently appearing in poetry, paintings, and handicrafts. It has many other names, such as "National Beauty," "Flower of Wealth and Honor," and "Luoyang Flower," reflecting its important position in history and culture. The peony's beauty and fragrance have made it a subject of many works of art. … The peony is loved not only for its beauty but also for its profound cultural connotations and symbolic meaning. It occupies an important place in traditional Chinese culture, representing people's yearning and pursuit of a better life. Whether in home gardening or in cultural arts, the peony plays an indispensable role.

[0060] Popular varieties to admire: Yao Huang: Originating in Luoyang during the Song Dynasty, its flowers are initially pale yellow, turning golden yellow when in full bloom. With full, rounded flowers and a delicate fragrance, it is hailed as the "King of Flowers." Wei Zi: Originating in Luoyang during the Five Dynasties period, its flowers are purplish-red, lotus-shaped or crown-shaped, with a long blooming period, abundant flowers, and full blooms, earning it the title of "Queen of Flowers." Zhao Fen: Originating in Heze during the Qing Dynasty, its flowers are pink. The plant is vigorous, produces a large number of flowers, and is a multi-flowered variety with a pleasant fragrance. Er Qiao: Originating in Yinli Garden during the Yuanfeng era of the Song Dynasty, originally called "Luoyang Jin," it is quite unique that it can produce both purplish-red and pinkish-white flowers on the same plant or branch. Through the dedicated research of botanical experts, two new peony varieties, "Jingyu" and "Wenhai," have been successfully cultivated. The experts used terminal devices to upload images, textual descriptions, and pictures of the new peony varieties, along with other relevant teaching materials, to a cloud server as new source data. The cloud server uses a polling mechanism to retrieve the new source data as update data and employs a large-scale data interpretation model to interpret this updated data, generating foundational interpretation data. This foundational interpretation data contains several key data points related to the two new peony varieties, "Jingyu" and "Wenhai." At this point, the foundational interpretation data from the updated data and the source data is merged to form merged foundational interpretation data. Then, the large-scale data understanding model is used to analyze and mine this merged foundational interpretation data, generating updated application understanding data. This updated application understanding data includes descriptions of the new peony varieties, as well as information on their kinship with traditional peonies, flower shape evolution patterns, morphological characteristics, and cultivation correlations.

[0061] In this example, the core purpose of acquiring basic interpretation data and application understanding data is to accelerate data analysis and processing based on user needs. Specifically, a student has a strong interest in learning about flowers and wants a detailed explanation of peony-related knowledge. In this case, the user can express their need via voice or text: "An explanation of peony flowers and related knowledge." Based on the user's needs, the cloud server uses a demand parsing model, combined with the language description logic of the education field, to parse the application understanding data and generate data operation execution requirements. Then, based on the data operation execution requirements, the demand parsing model is used to parse the basic understanding data, generating data processing logic that meets the user's needs. This data processing logic includes: data retrieval steps, data integration and analysis steps, and data expansion steps.

[0062] The cloud server, following data processing logic and based on application understanding, retrieves fundamental interpretive data directly or indirectly related to peonies. This includes the peony's knowledge system, flower shape evolution patterns, historical and cultural information, and popular viewing information. Then, using a medical data integration and organization model pre-trained based on data integration and organization technology, the server integrates and organizes the fundamental interpretive data (knowledge system, flower shape evolution patterns, historical and cultural information) and application understanding data, generating explanatory materials presented in various formats such as text, charts, graphic reports, PowerPoint presentations, audio, and video. Finally, popular viewing information, viewing location navigation, and cultivation methods are integrated as extended information into the explanatory materials, and the level of detail is adjusted to generate data processing results that meet user needs.

[0063] As demonstrated in Example 2, applying data understanding can more clearly and accurately reveal the important potential information within teaching data. It enables the construction of a deeper link between underlying teaching data and user needs. When users raise requests, applying data understanding can more accurately and quickly locate key data in the source data, achieving a precise response to the corresponding user needs. Furthermore, applying data understanding can improve the speed of effective data processing for user needs, while also making the generated learning materials richer and more comprehensive.

[0064] Example 3 In this example, the data interaction method is further explained using the data interaction in the field of information statistical analysis for analyzing hidden pathogenic factors of diseases.

[0065] Healthcare institutions upload daily patient examination and diagnosis data, patient visit process data, medical resource operation data, doctor treatment data, and patient behavior during treatment, etc., as source data to a cloud server. The cloud server uses a large-scale data interpretation model to interpret the source data, obtaining basic interpretation data. Basic interpretation data includes several key data points. For example, these key data points may include: patient diagnosis reports, electrocardiogram characteristics, lesion characteristics in CT images, etc. They may also include: patient traffic, equipment usage frequency, drug consumption, medical staff deployment, medical supply consumption, patient examination processes, and patient hospitalization processes on the day the institution operates. Furthermore, they may include: doctor medication dosage, medication time, patient recovery status, and behavior during nursing care for patients with a particular condition. Finally, they may include: the number of patients seeking treatment for a specific disease at different time periods; treatment methods for a specific disease for different income groups; and the treatment effects of different methods on a specific disease.

[0066] Using a large-scale data understanding model, we can analyze and mine basic interpretation data to generate application understanding data. Application understanding data can contain potential content that expresses the basic interpretation data. For example, application understanding data could include: "...From January to March, there were many patients seeking medical attention for cerebrovascular diseases. Patients underwent brain MRI (magnetic resonance imaging) examinations as advised by their doctors. Brain CT images revealed localized intracranial blood flow obstruction. Cerebrovascular diseases are mostly acute in onset, with patients over 40 years of age. Many acute cerebrovascular diseases are caused by ischemia and are usually treated with medications to improve collateral circulation in the brain, without the need for hospitalization. Acute cerebrovascular diseases caused by hemorrhage are less common and usually require emergency treatment by medical personnel, along with inpatient medication..." For example, data interpretation could include: "...A significant number of patients sought medical attention for chronic obstructive pulmonary disease (COPD) between October and December. These patients underwent chest CT scans as advised by their doctors. The chest CT images revealed respiratory infections in most patients. Among lung disease patients, those with vascular disease often experienced acute onset, with an age of onset exceeding 40 years. Many COPD cases were caused by bronchitis, and these patients typically received medication to alleviate their condition under medical supervision, without requiring hospitalization for observation. While fewer COPD cases were caused by emphysema, these patients are highly susceptible to developing pulmonary heart disease or respiratory failure, and doctors usually recommend hospitalization for medication treatment..." In this example, a disease control research institution needs to study the impact of seasons on certain diseases. The user can then submit a request via voice or text: "Some diseases have seen a significant increase in outpatient visits in the fourth quarter." The cloud server, based on the user's request, uses a demand parsing model to analyze the application understanding data and generate data operation execution requirements. This analysis reveals important information about outpatient visits and diseases, such as a higher number of patients seeking medical attention for chronic obstructive pulmonary disease (COPD) between October and December, with more patients receiving medication and fewer requiring inpatient medication. However, this information is insufficient to fully address the user's needs. Further information is needed, including factors contributing to COPD and control recommendations; recommendations on healthcare worker involvement and medication preparation; and prevention and treatment suggestions. This generates data operation execution requirements for further processing of the basic interpretation data and / or application understanding data.

[0067] Based on the data operation execution requirements, a large-scale requirement analysis model is used to analyze the basic understanding data and generate data processing logic that meets the user's needs. This data processing logic includes data search, data analysis, and data integration steps. Following this logic, the cloud server, based on the application understanding data, retrieves basic interpretation data directly or indirectly related to chronic obstructive pulmonary disease (COPD). Then, using a medical data integration and analysis model pre-trained based on data integration and analysis technology, the basic interpretation data and application understanding data are combined to integrate and analyze the associated factors of COPD, generating data processing results related to these factors. Finally, based on environmental conditions and medical records from October to December, the system provides recommendations for COPD prevention and treatment, medical staff deployment plans, and drug pre-stocking plans, among other data processing results.

[0068] Finally, through the data integration step, the data processing results are integrated and optimized, and presented using text, charts, graphic reports, presentation slides, audio, video, and other methods to generate data processing results that meet user needs.

[0069] Based on the above implementation methods of data interaction, such as Figure 5 As shown, this solution further provides a cloud server 401 comprising: a first intelligent agent 402, a second intelligent agent 403, and a third intelligent agent 404. The first intelligent agent 402 uses a large-scale data interpretation model to interpret the source data and generate basic interpretation data. The second intelligent agent 403 uses a large-scale data understanding model to understand and mine the basic interpretation data, generating application understanding data that can express the potential content of the basic interpretation data.

[0070] The first intelligent agent 302 can utilize feature information recognition models or large-scale data interpretation models, combined with application domain knowledge such as practical experience and technical guidelines, to identify key information and key regions of interest in the source data. Then, it annotates the key data to obtain multiple key pieces of information as basic interpretation data. The first intelligent agent 302 also has the ability to poll newly generated source data. It can poll new source data uploaded after user authorization. The first intelligent agent 302 uses the polled new source data as updated data. It then uses the large-scale data interpretation model to interpret the updated data, generating basic interpretation data. Finally, it merges the basic interpretation data of the updated data and the basic interpretation data of the source data to generate merged basic interpretation data.

[0071] The first intelligent agent 302 can utilize a large-scale data understanding model, combined with practical experience and / or technical guidelines in the application domain, to perform understanding and mining on basic interpretation data, generating application understanding data containing the underlying logic and / or potential information of the data. This application understanding data can be text-based or non-text-based data such as images, audio, video, and statistical charts. It's important to note that when there are no data updates, existing basic interpretation data can be used for understanding and mining. However, if updated data is available, the large-scale data understanding model must be used to perform understanding and mining on the merged basic interpretation data to generate updated application understanding data. This updated application understanding data allows users to more intuitively understand the changes and development trends of objects of interest.

[0072] The third intelligent agent 304 can respond to user requests by using a large-scale request parsing model to analyze application understanding data and generate data operation execution requirements. Then, based on these data operation execution requirements, it uses the large-scale request parsing model to analyze basic understanding data and generate data processing logic. Finally, it uses this data processing logic to process the application understanding data and / or basic interpretation data, generating data processing results.

[0073] This solution leverages large language models to understand basic interpretive data, generating application-understanding data that contains the underlying content of the data. By using application-understanding data, a deeper link is established between the underlying data and user needs, enabling precise responses to user demands and improving the speed of effective data processing tailored to user needs.

[0074] Based on the above-described data interaction method implementation, this solution further provides a computer-readable storage medium. This computer-readable storage medium is used to implement the program product of the above-described pipe connection method. It may be a portable compact disc read-only memory (CD-ROM) and include program code, and can run on a device, such as a personal computer. However, the program product of this solution is not limited to this. In this document, the readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0075] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0076] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0077] Program code contained on a computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0078] Program code for performing the operations of this solution can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0079] Based on the above-described implementation of the data interaction method, this solution further provides a data interaction device. For example... Figure 6 As shown, the data interaction device is an example and should not impose any limitations on the functionality and scope of use of this embodiment.

[0080] like Figure 6As shown, the data interaction device 501 is presented in the form of a general-purpose computing device. The components of the data interaction device 501 may include, but are not limited to: at least one storage unit 502, at least one processing unit 503, a display unit 504, and a bus 505 for connecting different system components.

[0081] The storage unit 502 stores program code that can be executed by the processing unit 503, causing the processing unit 503 to perform the steps of the various exemplary embodiments described in the pipeline connection method above. For example, the processing unit 503 can perform the steps of the data interaction method described above.

[0082] Storage unit 502 may include volatile storage units, such as random access memory (RAM) and / or cache storage units, and may further include read-only memory (ROM).

[0083] Storage unit 502 may also include programs / utilities with program modules, such program modules including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0084] Bus 505 may include a data bus, an address bus, and a control bus.

[0085] The data interaction device 501 can also communicate with one or more external devices 507 (e.g., keyboards, pointing devices, Bluetooth devices, etc.) via the input / output (I / O) interface 406. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the data interaction device 501, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0086] Obviously, the above embodiments of the present invention are examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all the implementation methods here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A data interaction method, characterized in that, The steps of this method include: The acquired source data is interpreted to generate basic interpretation data; By using large-scale data understanding models to understand basic interpretation data, application understanding data is generated that expresses the potential content of the basic interpretation data. In response to user needs, the system utilizes data processing logic generated from the large model of needs analysis to process application-understanding data and / or basic interpretation data, generating data processing results.

2. The data interaction method according to claim 1, characterized in that, The steps of interpreting the acquired source data to generate basic interpretation data include: Obtain source data; Use feature information to identify models or interpret large models, and identify key data in the source data; Key data is labeled to obtain multiple key pieces of information as the basis for interpretation.

3. The data interaction method according to claim 1, characterized in that, The steps involved in using a large data understanding model to understand basic interpretation data and generate application-level data that expresses the potential content of the basic interpretation data include: By leveraging large-scale data understanding models, combined with practical experience and / or technical guidelines in the application domain, we can perform understanding and mining on basic interpretive data to generate application understanding data that contains the underlying logic and / or potential information of the data.

4. The data interaction method according to claim 1, characterized in that, The steps for responding to user needs and utilizing the data processing logic generated by the large-scale demand analysis model to process application-understanding data and / or basic interpretation data to generate data processing results include: In response to user needs, the application understands the data by using a large-scale requirement analysis model to generate data operation execution requirements; Based on the data operation execution requirements, the basic understanding data is analyzed using the requirement analysis model to generate data processing logic; Using the aforementioned data processing logic, the application-understanding data and / or basic interpretation data are processed to generate data processing results for answering user needs.

5. A cloud server, characterized in that, The cloud server includes: The first intelligent agent interprets the acquired source data and generates basic interpretation data; The second intelligent agent uses a large data understanding model to understand the basic interpretation data and generate application understanding data that expresses the potential content of the basic interpretation data. The third intelligent agent responds to user needs and uses data processing logic generated by the large model of demand analysis to process application-understanding data and / or basic interpretation data, generating data processing results.

6. The cloud server according to claim 5, characterized in that, The first intelligent agent is used to perform the following steps: Obtain source data; Use feature information to identify models or interpret large models, and identify key data in the source data; Key data is labeled to obtain multiple key pieces of information as the basis for interpretation.

7. The cloud server according to claim 5, characterized in that, The second agent is used to perform the following steps: By leveraging large-scale data understanding models, combined with practical experience and / or technical guidelines in the application domain, we can perform understanding and mining on basic interpretive data to generate application understanding data that contains the underlying logic and / or potential information of the data.

8. The cloud server according to claim 5, characterized in that, The third agent is used to perform the following steps: In response to user needs, the application understands the data by using a large-scale requirement analysis model to generate data operation execution requirements; Based on the data operation execution requirements, the basic understanding data is analyzed using the requirement analysis model to generate data processing logic; Using the aforementioned data processing logic, the application-understanding data and / or basic interpretation data are processed to generate data processing results for answering user needs.

9. A data interaction device, characterized in that, include: Memory, one or more processors; The memory and processor are connected via a communication bus; the processor is configured to execute instructions from the memory. The memory stores instructions for performing each step of the method as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 4.

Images