Method for constructing crop planting service model of knowledge processing organization based on fragmentation

Abstract

A kind of crop planting service model construction method based on fragmented knowledge processing organization, method includes: collecting agricultural knowledge information for crop planting knowledge;Based on the extraction of key words in agricultural knowledge information and semantic division, so that each agricultural knowledge information generates corresponding one text data;For the text data processed, classification is carried out in turn from two characteristic dimensions of crop category and crop planting problem;Crop planting problem and its solution process are expressed into reasoning logic, core network and correlation table are constructed;According to the core network, all the keywords converted into standard vocabulary description are composed into directed graph network;According to directed graph network and correlation table, decision calculation based on directed graph network for decision-making crop planting is given;Based on directed graph network and decision calculation, the crop planting service model of agricultural fragmented knowledge is given.

Description

Technical Field

[0001] This invention relates to the field of swarm intelligence computing technology, and in particular to a method for constructing a crop planting service model based on fragmented knowledge processing organization. Background Technology

[0002] Collective intelligence originates in nature and exists within various biological clusters, regardless of size, demonstrating its effectiveness. This intelligence is not only reflected in the cooperation and competition among individuals but also in the entire group's ability to adapt to the environment and solve problems. With the development of human society, modern collective intelligence is constantly evolving, from initial cooperation and competition between social groups to the sharing and collaboration of public spaces, and then to the widespread connectivity and high interactivity of the internet age, the scope for the application of collective intelligence is becoming increasingly broad.

[0003] Leveraging collective intelligence in decision-making is a common application, as it improves the accuracy and effectiveness of decisions by aggregating the opinions and knowledge of multiple individuals. Various methods exist for collective decision-making, including reasoning-based, similarity-metric-based, and relationship-based approaches, all aiming to enhance consensus and accuracy. However, when decision items or data are inherently chaotic, traditional methods struggle to address semantic ambiguity and incompleteness. Simple machine learning or deep learning methods may also fail to effectively solve the problem because they require high-quality data and complete information, which chaotic data often lacks. Therefore, it is necessary to explore collective decision-making methods better suited for handling chaotic data.

[0004] The aforementioned data situation can also be described as fragmented knowledge, characterized by its multi-source nature, social nature, disorder, incompleteness, and abundance of redundancy and metaphorical elements. In the internet age, this fragmented knowledge is widespread. Particularly in the agricultural planting field, the differences in how agricultural planting knowledge is expressed across different regions, coupled with the colloquial expressions used by many farmers, result in a significant amount of fragmented descriptions of relevant knowledge. To facilitate better processing and utilization of this agricultural planting knowledge by agricultural practitioners, there is an urgent need for a technology and invention to provide a communication platform for experience sharing, Q&A consultations, and other interactive activities. Such a platform can help integrate fragmented agricultural planting knowledge, improve its accessibility and usability, and thus promote the development and progress of the planting industry.

[0005] The information disclosed in the background section is only for enhancing the understanding of the background of this invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the problems existing in current technologies, this invention proposes a method for constructing a crop planting service model based on fragmented knowledge processing. This model allows users to consult on agriculture-related issues and engage in knowledge-based question-and-answer sessions with other users by uploading text and images. The model itself possesses the basic ability to judge the correctness of input knowledge and can add new knowledge or delete redundant knowledge, updating its knowledge in real time to adapt to constantly changing user needs.

[0007] This invention is achieved through the following technical solution:

[0008] A method for constructing a crop planting service model based on fragmented knowledge processing includes:

[0009] Step 1: Collect agricultural knowledge information for crop cultivation, including text and images;

[0010] Step 2: Extract keywords based on the agricultural knowledge information and perform semantic segmentation so that each piece of agricultural knowledge information generates a corresponding text data;

[0011] Step 3: Convert all keywords in the text data into keywords with standardized vocabulary descriptions according to their meanings;

[0012] Step 4: Classify the processed text data sequentially from two feature dimensions: crop category and crop planting issues;

[0013] Step 5: Express the crop planting problem and its solution process as reasoning logic, construct the core network and correlation table. The correlation table is used to evaluate the relationship between crop categories and crop planting problems, as well as the relationship between crop planting problems and problem-solving measures.

[0014] Step 6: Based on the core network, construct a directed graph network from all keywords that have been converted into standard vocabulary descriptions;

[0015] Step 7: Based on the directed graph network and the correlation table, provide a decision calculation for crop planting based on the directed graph network, which consists of reading the correlation table and formulas.

[0016] Step 8: Based on the directed graph network and decision computation, a crop planting service model for fragmented agricultural knowledge is given.

[0017] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 2, the semantic segmentation and extraction of text data are performed using the term frequency-inverse text frequency index method. During the extraction process, six keyword categories are determined: crop category, crop planting problem, crop growth environment, problem occurrence time, crop growth status, and problem handling measures. Keywords are selected by combining the frequency of keyword occurrence in the text and the number of texts containing the keyword in all texts, using the TF-IDF value of the term frequency-inverse text frequency index method.

[0018] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 5, the reasoning logic gives the relationship between the problem and the answer. First, the type of crop is determined. Then, by observing the state of the crop and combining the current environmental conditions and the time when the problem occurs, the problem of the crop is judged, and corresponding problem handling measures are given based on the problem of the crop.

[0019] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 5, the core network takes the crop category as the center of the network. Around the center, keywords related to possible crop planting problems and problem-solving measures related to the crop category are logically organized to create a core network for each crop category. The directed graph network is formed based on the core network plus keywords with three characteristics: crop growth status, crop growth environment, and problem occurrence time. In the network structure of the directed graph network, each node represents a specific instance of a certain keyword, and the connection between nodes represents a thinking process or logical association, indicating the reasoning path from one keyword to another.

[0020] In the method for constructing a crop planting service model based on fragmented knowledge processing, the correlation table provides a quantitative way to measure the correlation between different factors.

[0021] In the method for constructing a crop planting service model based on fragmented knowledge processing organization, the reasoning logic includes the logic of judging crop problems by observing the state of crops, from crop growth status to crop planting problems, and the logic of selecting treatment methods based on problems in the crop planting process, from crop planting problems to problem treatment measures. The reasoning logic first infers possible answers from the organizational structure using existing information, and then filters the most likely results through a correlation table.

[0022] In the method for constructing a crop planting service model based on fragmented knowledge processing, step 3 includes:

[0023] The Word2vec model is used to determine whether words from an industry-standard vocabulary set are synonyms with other words.

[0024] Word vectors are obtained from standard words using Word2vec;

[0025] For each other word in the non-industry standard knowledge, after obtaining the word vector using Word2vec, the similarity between the standard word and the other words is calculated. If the similarity is greater than a threshold, it is judged as a synonym; otherwise, it is not a synonym.

[0026] If a word is identified as a synonym, all other words under that feature are deleted; otherwise, no action is taken.

[0027] The generation process of the core network in the method for constructing a crop planting service model based on fragmented knowledge processing organization is as follows:

[0028] Based on the classification results of crop categories, we obtain many sets of data. The crop category characteristics of each set of data are consistent. Therefore, for each set of data, we obtain a core network, as follows:

[0029] The first step is to construct a tree. Following the inherent logic, a tree is built based on three features: crop category, crop planting problem, and problem-solving measures. The crop category of this data set is used as the root node; the crop planting problem is used as the second-level child node, and these are connected to the root node. Then, for each second-level child node, the problem-solving measures are used as the third-level child node, and these are connected to their corresponding second-level child nodes. All data under the same second-level child node have the same crop category and crop planting problem, thus establishing a tree structure based on these three features.

[0030] The second step is to add branch information. For two adjacent nodes N1 and N2, assuming N1 is the node closer to the root node, the branch N1N2 represents the thought process from N1 to N2. This process starts with considering a crop category, its growing environment, growth status, and the time the problem occurred, leading to the conclusion that it is a crop cultivation problem. It then considers the appropriate problem-solving measures for a given crop cultivation problem, presented as a probability, calculated using the following formula:

[0031]

[0032] In the method for constructing a crop planting service model based on fragmented knowledge processing, the creation process of two correlation tables is as follows:

[0033] A table showing the correlation between "crop category" and "crop planting issues" was created.

[0034] "Crop growing environment," "problem occurrence time," and "crop growing status" are represented by word sets E, T, and S, respectively, containing o, m, and n keywords. These sets contain all words related to the corresponding features in the dataset. Let a represent a crop, a ∈ A, where A is the crop set; b represent a crop planting problem, b ∈ B, where B is the crop planting problem set; c represent a problem-solving measure, c ∈ C, where C is the problem-solving measure set; e ∈ E, t ∈ T, and s ∈ S.

[0035] For crop category a, one possible crop planting problem is b. Given a certain e, t, and s, the influence of e, t, and s on "crop category a and its crop planting problem b" are as follows:

[0036]

[0037]

[0038]

[0039] Where, ω E ω T ω S The weights of these three features are all less than 1, depending on their influence on the "crop category" to the "crop planting problem." Generally, "crop growth status" has the greatest influence, followed by "crop growth environment," and "problem occurrence time" has the least influence. a,b Let n be the number of data entries where "crop category" is a and "crop planting problem" is b. e,a,b In this case, the number of data entries containing the word 'e' is n. t,a,b and n s,a,b Similarly, for a single crop, the impact of three characteristics is calculated for different crop cultivation problems, and a correlation table between the "crop" and the "crop cultivation problem" is established using the following method.

[0040] The first step is to calculate the probability P based on the branches between the "crop category" node a and the corresponding "crop planting problem" node b, calculated during the core network generation process. ab Determine P according to the formula below. ab The distance D between "crop planting issues" and "crop categories" ab ,

[0041]

[0042] D ab =-lnP ab ,

[0043] The second step involves focusing on each crop planting problem. For each crop planting problem, the W value for each keyword is calculated using the formula mentioned above. Then, the distance between each keyword and the crop planting problem is determined using the formula below.

[0044] d = -lnW,

[0045] Determine the relationship between each keyword and the crop planting problem.

[0046] The third step is to create a table, with each keyword as a column and each crop planting question as a row. For a given crop category 'a', fill in all the distance values ​​to build a relevance table.

[0047] Creation of a correlation table between "Crop Planting Issues" and "Issuance Measures":

[0048] "Crop growing environment," "problem occurrence time," and "crop growing status" are represented by word sets E, T, and S, respectively, containing o, m, and n keywords. These sets contain all words related to the corresponding features in the dataset. Let a represent a crop, a ∈ A, where A is the crop set; b represent a crop planting problem, b ∈ B, where B is the crop planting problem set; c represent a problem-solving measure, c ∈ C, where C is the problem-solving measure set; e ∈ E, t ∈ T, and s ∈ S.

[0049] For a problem-solving measure c, and given a, t, s, then the magnitudes of the influence of e, t, s on "crop planting problem b with problem-solving measure c" are respectively:

[0050]

[0051]

[0052]

[0053] Meanwhile, the magnitude of the influence of crop category a on "crop planting problem b and its problem-solving measure is c" is:

[0054]

[0055] Where, ω E ω T ω S ω represents the weights of three features: "crop growth status," "crop growth environment," and "problem occurrence time," all less than 1. The weights depend on the magnitude of the influence of these three features on the transition from "crop planting problem" to "problem-solving measures." Generally, "crop growth status" has the greatest weight, followed by "crop growth environment," and "problem occurrence time" has the smallest. AThis is the weight of "crop category" when calculating this effect; it is greater than ω. E ω T ω s Small, n b,c Let n be the number of data entries where "crop planting problem" is b and "problem-solving measure" is c. e,b,c In this case, the number of data entries containing the word 'e' is n. t,b,c n s,b,c and n a,b,c Similarly, for a single crop, the impact of four characteristics was calculated for different crop cultivation issues.

[0056] Establish a correlation table between the "crop planting issues" and "issue handling measures" using the following method;

[0057] The first step is to calculate the probability P of the branches between the nodes of the "crop planting problem" b and the corresponding "problem-solving measures" b, calculated during the core network generation process. bc Determine P according to the formula below. bc The distance D between "crop planting issues" and "crop categories" bc ,

[0058]

[0059] D bc =-lnP bc ,

[0060] The second step involves focusing on each problem-solving measure. For each problem-solving measure, calculate the W value for each keyword using the formula mentioned above, and then determine the distance between each keyword and the crop planting problem using the formula below.

[0061] d = -lnW,

[0062] Determine the relationship between each keyword and the problem-solving measures, given the identified problem-solving measures.

[0063] The third step is to create a table. For each keyword, add a "crop category" column, and for each problem-solving measure, add a row. Then, for a specific crop planting problem b, fill in all the distance values ​​to build a relevance table.

[0064] In the method for constructing a crop planting service model based on fragmented knowledge processing, the decision-making calculation from "crop growth status" to "crop planting problem" is as follows:

[0065] The first step is to find the corresponding keywords for "crop category," "crop growing environment," "problem occurrence time," and "crop growth status" in the organizational structure for each data point, and mark the corresponding second-level child nodes. Then, iterate through the marked second-level child nodes in the organizational structure to find all third-level child nodes connected to any of them. After traversing all the nodes, a series of third-level child nodes are obtained, which correspond to the keywords "crop planting problem."

[0066] The second step is to list the obtained keywords for "crop planting problems," and for each keyword, list the keywords corresponding to the second-level child nodes that are related to it, and categorize them according to the characteristics of "crop growth environment," "time of problem occurrence," and "crop growth status."

[0067] The third step is to calculate the relation number for each keyword "crop planting problem" b under the specific keywords "crop growth environment", "problem occurrence time", and "crop growth status" for that crop category a, using the following formula:

[0068]

[0069] Here, ω is a weight representing the "credibility" of the existing data structure; n is the total number of keywords in the data, including "crop growth environment," "problem occurrence time," and "crop growth status," which is the same number as the number of keywords represented by all child nodes in the second level of the organizational structure; and ∑d is the sum of the values ​​read from the correlation table between "crop category" and "crop planting problem" after matching all the keywords related to the crop planting problem.

[0070] Fourth step, retrieve all obtained H values. b The smallest 1 to 2 values ​​correspond to the crop planting problem calculated under that data.

[0071] In the method for constructing a crop planting service model based on fragmented knowledge processing, the decision-making calculation from "crop planting problem" to "problem-solving measures" is as follows:

[0072] The first step is to find the corresponding keywords "crop growth status", "crop growth environment", "problem occurrence time" and "crop planting problem" in the organizational structure for a data set, and mark the corresponding second and third level child nodes. Then, traverse the marked third level child nodes in the organizational structure to find all the fourth level child nodes connected to them. After traversing, a series of fourth level child nodes are obtained, which correspond to the keywords "problem handling measures".

[0073] The second step is to calculate the relation coefficient for each obtained "problem-handling measure" keyword c, under the keywords "crop growth environment," "problem occurrence time," and "crop growth status" marked in crop category a, when the crop planting problem is b. The formula is as follows:

[0074]

[0075] Where ω is a weight representing the "credibility" of the existing data structure; n is the total number of keywords in the data, namely "crop growth environment," "problem occurrence time," and "crop growth status," which is the same number as the number of keywords represented by all child nodes in the second level of the organizational structure; and ∑d is the sum of the values ​​read from the correlation table between "crop planting problem" and "problem handling measure" after matching the keywords related to the problem handling measure.

[0076] The third step is to retrieve all the obtained H values. c The smallest 1 to 2 values ​​correspond to the problem-solving measures calculated for that data.

[0077] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 8, based on the directed graph network and decision computation, a crop planting service model based on fragmented agricultural knowledge is provided. Based on the directed graph structure, an input module, an output module, a user interaction module, and a right / wrong judgment module are added to construct the service model. On this basis, service functions are provided and the implementation logic of each function is provided.

[0078] The service model has the following functions:

[0079] Querying crop planting issues under current conditions: The model can determine whether the crop is growing normally or is diseased based on the input crop category and crop growth environment conditions;

[0080] How to handle problems encountered during queries: When users encounter problems with crops, such as crop diseases or abnormal growth, the model can provide corresponding problem-solving measures and suggestions;

[0081] Consultation on correct methods and information for crop cultivation: Users can search for correct methods and information about crop cultivation, such as when cultivation problems occur, growing conditions, etc.

[0082] Upload text and images to seek answers: Users can upload text and images to seek answers or explanations from the model for specific questions, or to seek answers from other users;

[0083] The ability to judge right and wrong by combining knowledge from multiple sources: The model can combine knowledge from multiple sources to judge the input information and confirm its correctness;

[0084] Adding new knowledge and removing redundant knowledge on the existing structure: The model can continuously improve its knowledge base based on the existing knowledge structure, improve its performance and accuracy, and adapt to the needs of more different scenarios.

[0085] Logic implementation of some functions:

[0086] Consult the correct methods and information.

[0087] Its function is achieved by combining two decisions, namely, from "crop growth status" to "crop planting problem" and from "crop planting problem" to "problem handling measures", or one of them. When agricultural workers have such problems, after determining the "crop type", "crop growth environment", "problem occurrence time" and "crop growth status", the answer to the problem is obtained through a directed graph structure.

[0088] Upload your own text and images.

[0089] After the data is uploaded, it becomes new data. After step 2, it is transformed into a standard format, and then the next operation is carried out according to the problem.

[0090] Judge whether it is true or false.

[0091] First, treat "crop planting problem" and "problem-solving measures" as unknowns, then perform calculations. Compare the results obtained through the directed graph structure with the given results. If they are the same or the former basically contains the latter, then it is correct. It should be noted that since the results given by the organizational structure are obtained in the last step by taking the 1 to 2 results with the lowest relation coefficients from multiple results, the following rule should be added to the judgment criteria for correctness: if the given result appears among all possible results before the organizational structure determines the final result, then it is partially correct; if it does not appear, then it is incorrect.

[0092] Learn and add new knowledge.

[0093] The first step is to determine if the data is correct. If it is correct, continue; otherwise, terminate. This determination can be made by an expert or by the model itself.

[0094] The second step is to arrange the new keywords in the new data. If the new keywords already exist in the original structure, they remain unchanged; otherwise, they are arranged according to their respective characteristics at the corresponding levels of the organizational structure, serving as new nodes.

[0095] The third step is to add connections between the layers according to the logic of the new data. That is, after the first step, traverse all nodes. If the nodes and corresponding keywords of adjacent layers appear at the same time in the new data and there is no connection, then add a connection.

[0096] Remove outdated or redundant knowledge.

[0097] The first step is to determine if the data entry has the aforementioned problems. If so, continue; otherwise, terminate. This determination should be made by an expert, or if the model's result is found to be inconsistent with the determined correct result.

[0098] The second step is to determine which two features are causing the problem and to list all the keywords corresponding to those features in that data entry.

[0099] The third step is to mark the corresponding nodes for these keywords in the structure.

[0100] The fourth step is to delete all the connections between the marked nodes. This step may delete some correct logic at the same time. Therefore, after the deletion operation, it is recommended to add a correct piece of knowledge corresponding to the deleted knowledge.

[0101] Compared with the prior art, the present invention has the following advantages:

[0102] This invention constructs a fragmented knowledge network structure, distinct from previous single-knowledge graph structures, and establishes a decision-making method upon it. This design enables the invention to better handle the large amounts of unstructured text and image information present in the agricultural field, extracting useful knowledge and organizing it into an easily understandable and usable knowledge structure. Simultaneously, this decision-making method can adapt to the judgment of uncertain and fuzzy agricultural knowledge, improving the accuracy and reliability of decision-making.

[0103] Meanwhile, this invention presents a novel processing framework and organizational form for handling fragmented knowledge, exhibiting good versatility. This method requires only unified data processing and is lightweight, without needing to guarantee coupling between steps. Certain specific implementation methods within each step can be varied according to data volume and project requirements. Therefore, this invention possesses strong adaptability and flexibility.

[0104] In addition, the agricultural service model presented in this invention has the following features: This service model possesses basic learning intelligence, human intelligence, and collective decision-making capabilities, and can uniformly integrate experience gained through deep learning and human experience gained in production and daily life. This allows the model to more comprehensively utilize various knowledge and information, improving the accuracy and efficiency of decision-making. By combining deep learning and human intelligence, the model can better understand and handle complex problems, adapting to different scenarios and needs. Simultaneously, the introduction of collective decision-making capabilities enables the model to fully utilize collective wisdom, integrating the opinions and suggestions of multiple individuals, and improving the reliability and effectiveness of decision-making. Attached Figure Description

[0105] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0106] In the attached diagram:

[0107] Figure 1 This diagram outlines the processing flow of fragmented agricultural knowledge, from steps 1 to 6. The knowledge acquisition and information extraction sections represent steps 1 and 2, the network and table generation section represents step 5, the arrows in the middle represent steps 3 and 4, and the final knowledge organization section represents step 6.

[0108] Figure 2 The diagram shows a sample of the completed core network construction in step 5. In this diagram, assuming there are n types of "crop planting problems" within this set of data (i.e., the same "crop category"), and we simply use numbers to represent the "crop planting problems" and "problem-solving measures," the final tree structure diagram of these data points is shown in the figure.

[0109] Figure 3 The diagram presents the final directed graph network structure of step 6, which represents the organizational structure of fragmented agricultural knowledge. This step uses two sets of data from two different crops, forming a tree structure as the initial state. Figure 3 'a' represents the initial state; Figure 3 b. Disconnect the connection between the root node and the second-level nodes; Figure 3 c. Insert relevant keywords; Figure 3 d. Add all the connections based on the data; Figure 3 e-fusion of repeated keywords; Figure 3 f merges the two trees and then merges the repeated keywords as the final state;

[0110] Figure 4 This is a schematic diagram of the basic functions and internal framework of the agricultural service model given in step 8 of this invention. It includes the core part, namely the fragmented agricultural knowledge organization structure and the addition and deletion operations performed on it, as well as other parts, and the user interaction part, namely the consultation, uploading and question-and-answer operations.

[0111] The present invention will be further explained below with reference to the accompanying drawings and embodiments. Detailed Implementation

[0112] Specific embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While specific embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0113] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0114] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0115] like Figures 1 to 4 As shown, the method for constructing a crop planting service model based on fragmented knowledge processing organizations includes the following steps:

[0116] Step 1: Collect agricultural knowledge information for crop cultivation, including text and images;

[0117] Step 2: Extract keywords based on the agricultural knowledge information and perform semantic segmentation so that each piece of agricultural knowledge information generates a corresponding text data;

[0118] Step 3: Convert all keywords in the text data into keywords with standardized vocabulary descriptions according to their meanings;

[0119] Step 4: Classify the processed text data sequentially from two feature dimensions: crop category and crop planting issues;

[0120] Step 5: Express the crop planting problem and its solution process as reasoning logic, construct the core network and correlation table. The correlation table is used to evaluate the relationship between crop categories and crop planting problems, as well as the relationship between crop planting problems and problem-solving measures.

[0121] Step 6: Based on the core network, construct a directed graph network from all keywords that have been converted into standard vocabulary descriptions;

[0122] Step 7: Based on the directed graph network and the correlation table, provide a decision calculation for crop planting based on the directed graph network, which consists of reading the correlation table and formulas.

[0123] Step 8: Based on the directed graph network and decision computation, a crop planting service model for fragmented agricultural knowledge is given.

[0124] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 2, the semantic segmentation and extraction of text data are performed using the TF-IDF (Term Frequency – Inverse Document Frequency) method. During the extraction process, six keyword categories are determined: crop category, crop planting problem, crop growth environment, problem occurrence time, crop growth status, and problem handling measures. The TF-IDF value of the TF-IDF method is used to filter keywords by combining the frequency of keyword occurrence in the text and the number of texts in which the keyword appears in all texts.

[0125] In the method for constructing a crop planting service model based on fragmented knowledge processing, step 5 involves reasoning logic that establishes the relationship between questions and answers. First, the type of crop is determined. Then, by observing the crop's condition and considering current environmental conditions and the timing of the problem's occurrence, a judgment is made regarding the crop's problems. Based on these problems, corresponding solutions are provided.

[0126] In the method for constructing a crop planting service model based on fragmented knowledge processing, in step 5, the core network takes the crop category as the center of the network. Around the center, keywords related to possible crop planting problems and problem-solving measures related to the crop category are logically organized to create a core network for each crop category. The directed graph network is formed based on the core network plus keywords with three characteristics: crop growth status, crop growth environment, and problem occurrence time. In the network structure of the directed graph network, each node represents a specific instance of a certain keyword, and the connection between nodes represents a thinking process or logical association, indicating the reasoning path from one keyword to another.

[0127] In the method for constructing a crop planting service model based on fragmented knowledge processing, the correlation table provides a quantitative way to measure the correlation between different factors.

[0128] In the method for constructing a crop planting service model based on fragmented knowledge processing organization, the reasoning logic includes the logic of judging crop problems by observing the state of crops, from crop growth status to crop planting problems, and the logic of selecting treatment methods based on problems in the crop planting process, from crop planting problems to problem treatment measures. The reasoning logic first infers possible answers from the organizational structure using existing information, and then filters the most likely results through a correlation table.

[0129] In the method for constructing a crop planting service model based on fragmented knowledge processing, step 3 includes:

[0130] The Word2vec (Word to Vector) model is used to determine whether words from an industry-standard vocabulary set are synonyms with other words.

[0131] Word vectors are obtained from standard words using Word2vec;

[0132] For each other word in the non-industry standard knowledge, after obtaining the word vector using Word2vec, the similarity between the standard word and the other words is calculated. If the similarity is greater than a threshold, it is judged as a synonym; otherwise, it is not a synonym.

[0133] If a word is identified as a synonym, all other words under that feature are deleted; otherwise, no action is taken.

[0134] The generation process of the core network in the method for constructing a crop planting service model based on fragmented knowledge processing organization is as follows:

[0135] Based on the classification results of crop categories, we obtain many sets of data. The crop category characteristics of each set of data are consistent. Therefore, for each set of data, we obtain a core network, as follows:

[0136] The first step is to construct a tree. Following the inherent logic, a tree is built based on three features: crop category, crop planting problem, and problem-solving measures. The crop category of this data set is used as the root node; the crop planting problem is used as the second-level child node, and these are connected to the root node. Then, for each second-level child node, the problem-solving measures are used as the third-level child node, and these are connected to their corresponding second-level child nodes. All data under the same second-level child node have the same crop category and crop planting problem, thus establishing a tree structure based on these three features.

[0137] The second step is to add branch information. For two adjacent nodes N1 and N2, assuming N1 is the node closer to the root node, the branch N1N2 represents the thought process from N1 to N2. This process starts with considering a crop category, its growing environment, growth status, and the time the problem occurred, leading to the conclusion that it is a crop cultivation problem. It then considers the appropriate problem-solving measures for a given crop cultivation problem, presented as a probability, calculated using the following formula:

[0138]

[0139] The process of generating the relevance table in the method for constructing a crop planting service model based on fragmented knowledge processing organization:

[0140] A table showing the correlation between "crop category" and "crop planting issues" was created.

[0141] "Crop growing environment," "problem occurrence time," and "crop growing status" are represented by word sets E, T, and S, respectively, containing o, m, and n keywords. These sets contain all words related to the corresponding features in the dataset. Let a represent a crop, a ∈ A, where A is the crop set; b represent a crop planting problem, b ∈ B, where B is the crop planting problem set; c represent a problem-solving measure, c ∈ C, where C is the problem-solving measure set; e ∈ E, t ∈ T, and s ∈ S.

[0142] For crop category a, one possible crop planting problem is b. Given a certain e, t, and s, the influence of e, t, and s on "crop category a and its crop planting problem b" are as follows:

[0143]

[0144]

[0145]

[0146] Where, ω E ω T ω S The weights of these three features are all less than 1, depending on their influence on the "crop category" to the "crop planting problem." Generally, "crop growth status" has the greatest influence, followed by "crop growth environment," and "problem occurrence time" has the least influence. a，b Let n be the number of data entries where "crop category" is a and "crop planting problem" is b. e，a，b In this case, the number of data entries containing the word 'e' is n. t，a，b and n s，a，bSimilarly, for a single crop, the impact of three characteristics is calculated for different crop cultivation problems, and a correlation table between the "crop" and the "crop cultivation problem" is established using the following method.

[0147] The first step is to determine P based on the probability Pab calculated from the branches between the "crop category" node a and the corresponding "crop planting problem" node b, as determined during the core network generation process, using the formula below. ab The distance D between "crop planting issues" and "crop categories" ab ,

[0148]

[0149] D ab =-lnP ab ,

[0150] The second step involves focusing on each crop planting problem. For each crop planting problem, the W value for each keyword is calculated using the formula mentioned above. Then, the distance between each keyword and the crop planting problem is determined using the formula below.

[0151] d = -lnW,

[0152] Determine the relationship between each keyword and the crop planting problem.

[0153] The third step is to create a table, with each keyword as a column and each crop planting question as a row. For a given crop category 'a', fill in all the distance values ​​to build a relevance table.

[0154] Creation of a correlation table between "Crop Planting Issues" and "Issuance Measures":

[0155] "Crop growing environment," "problem occurrence time," and "crop growing status" are represented by word sets E, T, and S, respectively, containing o, m, and n keywords. These sets contain all words related to the corresponding features in the dataset. Let a represent a crop, a ∈ A, where A is the crop set; b represent a crop planting problem, b ∈ B, where B is the crop planting problem set; c represent a problem-solving measure, c ∈ C, where C is the problem-solving measure set; e ∈ E, t ∈ T, and s ∈ S.

[0156] For a problem-solving measure c, and given e, t, and s, the magnitudes of the influence of e, t, and s on "crop planting problem b with problem-solving measure c" are respectively:

[0157]

[0158]

[0159]

[0160] Meanwhile, the magnitude of the influence of crop category a on "crop planting problem b and its problem-solving measure is c" is:

[0161]

[0162] Where, ω E ω T ω S ω represents the weights of three features: "crop growth status," "crop growth environment," and "problem occurrence time," all less than 1. The weights depend on the magnitude of the influence of these three features on the transition from "crop planting problem" to "problem-solving measures." Generally, "crop growth status" has the greatest weight, followed by "crop growth environment," and "problem occurrence time" has the smallest. A This is the weight of "crop category" when calculating this effect; it is greater than ω. E ω T ω s Much smaller. b,c Let n be the number of data entries where "crop planting problem" is b and "problem-solving measure" is c. e,b,c In this case, the number of data entries containing the word 'e' is n. t,b,c n s,b,c and n a,b,c Similarly, for a single crop, the impact of four characteristics was calculated for different crop cultivation issues.

[0163] Create a correlation table between the "crop planting issues" and "problem-handling measures" using the following method.

[0164] The first step is to calculate the probability P of the branches between the nodes of the "crop planting problem" b and the corresponding "problem-solving measures" b, calculated during the core network generation process. bc Determine P according to the formula below. bc The distance D between "crop planting issues" and "crop categories" bc ,

[0165]

[0166] D bc =-lnP bc ,

[0167] The second step involves focusing on each problem-solving measure. For each problem-solving measure, calculate the W value for each keyword using the formula mentioned above, and then determine the distance between each keyword and the crop planting problem using the formula below.

[0168] d = -lnW,

[0169] Determine the relationship between each keyword and the problem-solving measures, given the identified problem-solving measures.

[0170] The third step is to create a table. For each keyword, add a "crop category" column, and for each problem-solving measure, add a row. Then, for a specific crop planting problem b, fill in all the distance values ​​to build a relevance table.

[0171] In the method for constructing a crop planting service model based on fragmented knowledge processing, the decision-making calculation from "crop growth status" to "crop planting problem" is as follows:

[0172] The first step is to find the corresponding keywords for "crop category," "crop growing environment," "problem occurrence time," and "crop growth status" in the organizational structure for each data point, and mark the corresponding second-level child nodes. Then, iterate through the marked second-level child nodes in the organizational structure to find all third-level child nodes connected to any of them. After traversing all the nodes, a series of third-level child nodes are obtained, which correspond to the keywords "crop planting problem."

[0173] The second step is to list the obtained keywords for "crop planting problems," and for each keyword, list the keywords corresponding to the second-level child nodes that are related to it, and categorize them according to the characteristics of "crop growth environment," "time of problem occurrence," and "crop growth status."

[0174] The third step is to calculate the relation number for each keyword "crop planting problem" b under the specific keywords "crop growth environment", "problem occurrence time", and "crop growth status" for that crop category a, using the following formula:

[0175]

[0176] Here, ω is a weight representing the "credibility" of the existing data structure; n is the total number of keywords in the data, including "crop growth environment," "problem occurrence time," and "crop growth status," which is the same number as the number of keywords represented by all child nodes in the second level of the organizational structure; and ∑d is the sum of the values ​​read from the correlation table between "crop category" and "crop planting problem" after matching all the keywords related to the crop planting problem.

[0177] Fourth step, retrieve all obtained H values. b The smallest 1 to 2 values ​​correspond to the crop planting problem calculated under that data.

[0178] In the method for constructing a crop planting service model based on fragmented knowledge processing, the decision-making calculation from "crop planting problem" to "problem-solving measures" is as follows:

[0179] The first step is to find the corresponding keywords "crop growth status", "crop growth environment", "problem occurrence time" and "crop planting problem" in the organizational structure for each data point, and mark the corresponding second and third level child nodes. Then, traverse the marked third level child nodes in the organizational structure to find all the fourth level child nodes connected to them. After traversing, a series of fourth level child nodes are obtained, which correspond to the keywords "problem handling measures".

[0180] The second step is to calculate the relation coefficient for each obtained "problem-handling measure" keyword c, under the keywords "crop growth environment," "problem occurrence time," and "crop growth status" marked in crop category a, when the crop planting problem is b. The formula is as follows:

[0181]

[0182] Where ω is a weight representing the "credibility" of the existing data structure; n is the total number of keywords in the data, namely "crop growth environment," "problem occurrence time," and "crop growth status," which is the same number as the number of keywords represented by all child nodes in the second level of the organizational structure; and ∑d is the sum of the values ​​read from the correlation table between "crop planting problem" and "problem handling measure" after matching the keywords related to the problem handling measure.

[0183] The third step is to retrieve all the obtained H values. c The smallest 1 to 2 values ​​correspond to the problem-solving measures calculated for that data.

[0184] In the proposed method for constructing a crop planting service model based on fragmented knowledge processing, a service model and some of its functions are presented, based on previously proposed methods and frameworks.

[0185] The core of this service model is the directed graph structure obtained in step 6. By adding input, output, user interaction, and correctness / error judgment modules to this structure, the basic service model can be constructed. The model's functionality, based on these components, primarily includes the following functions:

[0186] Querying crop planting issues under current conditions: The model can determine whether the crop is growing normally or is diseased based on the input crop category and crop growth environment conditions;

[0187] How to handle problems encountered during queries: When users encounter problems with crops, such as crop diseases or abnormal growth, the model can provide corresponding problem-solving measures and suggestions;

[0188] Consultation on correct methods and information for crop cultivation: Users can search for correct methods and information about crop cultivation, such as when cultivation problems occur, growing conditions, etc.

[0189] Upload text and images to seek responses: Users can upload text and images to seek responses or explanations from the model for specific questions, or to seek responses from other users.

[0190] The ability to judge right and wrong by combining knowledge from multiple sources: The model can combine knowledge from multiple sources to judge the input information and confirm its correctness;

[0191] Adding new knowledge and removing redundant knowledge on the existing structure: The model can continuously improve its knowledge base based on the existing knowledge structure, improve its performance and accuracy, and adapt to the needs of more different scenarios.

[0192] The method for constructing a crop planting service model based on fragmented knowledge processing organization provides partial functional implementation of the service model.

[0193] Consult the correct methods and information.

[0194] This function is implemented by combining two decisions: from "crop growth status" to "crop planting problem" and from "crop planting problem" to "problem-solving measures," or one of them. When agricultural workers have this type of problem, after determining the "crop category," "crop growth environment," "problem occurrence time," and "crop growth status," the answer to the problem is obtained through a directed graph structure.

[0195] Upload your own text and images.

[0196] After the data is uploaded, it becomes new data. After step 2, it is transformed into a standard format, and then the next operation is carried out according to the problem.

[0197] Knowledge services

[0198] This function is unrelated to structure and computation; it is a form of communication between two users.

[0199] Judge whether it is true or false.

[0200] This function is implemented by first treating "crop planting problem" and "problem-solving measures" as unknowns, then performing calculations, and comparing the results obtained through the directed graph structure with the given results. If they are the same or the former basically contains the latter, then it is correct. It is important to note that since the results given by the organizational structure are obtained in the final step by selecting the 1-2 results with the lowest relation coefficients from multiple results, the following rule should be added to the judgment of correctness: if the given result appears among all possible results before the organizational structure determines the final result, then it is partially correct; if it does not appear, then it is incorrect.

[0201] Learn and add new knowledge.

[0202] The first step is to determine if the data is correct. If correct, continue; otherwise, terminate. This determination can be made by an expert or by the model itself.

[0203] The second step is to arrange the new keywords in the new data. If the new keywords already exist in the original structure, they remain unchanged; otherwise, they are arranged according to their respective characteristics at the corresponding levels of the organizational structure, serving as new nodes.

[0204] The third step is to add connections between the layers based on the logic of the new data. That is, after the first step, traverse all nodes, and if adjacent layers' nodes and corresponding keywords appear simultaneously in the new data, and no connection exists, then add a connection.

[0205] Remove outdated or redundant knowledge.

[0206] The first step is to determine if the data entry has the aforementioned problems. If so, continue; otherwise, terminate. This determination should be made by an expert, or if the model's result is found to be inconsistent with the determined correct result.

[0207] The second step is to determine which two features are causing the problem and to list all the keywords corresponding to those features in the data.

[0208] The third step is to mark the corresponding nodes for these keywords in the structure.

[0209] The fourth step is to delete all the connections between the marked nodes. This step may delete some correct logic at the same time, so it is recommended to add a corresponding correct piece of knowledge after the deletion.

[0210] In one embodiment, first according to Figures 1 to 3 The content is explained in detail in steps 1 to 6.

[0211] Step 1: Knowledge Acquisition

[0212] Knowledge collection correspondence Figure 1 The green part in the "Knowledge Acquisition, Information Extraction" box.

[0213] The sources of knowledge primarily include industry-standard knowledge (statements, descriptions, and vocabulary, etc.) and fragmented experiences and descriptions uploaded by users on the internet. Specifically, this includes some searchable open-source standard datasets, standard agricultural terminology descriptions, standard descriptive statements and explanations from official internet agricultural platforms, text descriptions from various questioners and answerers on internet agricultural Q&A platforms, and other descriptions. By combining industry-standard knowledge with fragmented experiences and descriptions, significant deviations from accurate knowledge can be prevented when all data is fragmented.

[0214] Knowledge can take the form of images and text. Textual knowledge can appear alone, while image-based knowledge is often accompanied by a short text description. Simply compile the textual knowledge and label the corresponding images with their corresponding image numbers.

[0215] The knowledge primarily encompasses the agricultural subfield of crop cultivation, potentially including crop types, descriptions of various crop conditions, crop growth environments and events, current crop disasters and pathologies, and methods for addressing these conditions. If this knowledge originates from industry standards, it will serve as benchmark knowledge for the data fusion process. If it comes from fragmented experiences and descriptions uploaded by users online, it will exhibit fragmented characteristics, including disorder, incomplete information, metaphors, and redundancy, which will be addressed in the data fusion process.

[0216] Step 2: Information Extraction

[0217] Information extraction correspondence Figure 1 The light blue and yellow parts in the "Knowledge Acquisition, Information Extraction" box.

[0218] The ultimate goal of information extraction is to reduce a piece of knowledge to the form of a list of data as shown below. Specifically, a piece of textual knowledge becomes the following form after information extraction:

[0219]

[0220] Except for the numbering, which is inherited from the original knowledge, the other six items are features of new data, all presented in the form of words and phrases. Among them, the "crop category" feature contains a unique word, "crop planting problem" and "problem handling measures" contain 1 to 2 words, the number of words for "crop growth status" and "crop growth environment" needs to be set according to the text length, and the "problem occurrence time" feature uses fixed information representing "month", such as "January", "February", etc.

[0221] After information extraction, an image data point becomes the following format:

[0222]

[0223] If there is a correspondence between image knowledge and text knowledge, then merging the two under the corresponding features will result in a single piece of data obtained after this step.

[0224] In other words, step 2 transforms the "text + image" or "text" knowledge into a dataset with six features, each composed of words or phrases. Then, if the "crop planting question" feature in a dataset contains multiple words, it is decomposed into multiple datasets. The decomposed datasets retain the same remaining features as the original dataset, and each "crop planting question" contains one word from the original dataset. Similarly, when "problem-solving measures" contain multiple words, the processing method is the same. The processed data is then added to the dataset, and the original data is deleted.

[0225] The following section introduces the text knowledge processing methods. In step 2, the text knowledge information extraction method is the TD-IDF method, which calculates two weights for a keyword to determine its importance to the text, ultimately selecting the keyword with the higher importance. The specific formula for this method is as follows:

[0226]

[0227]

[0228] tf-idf i,j =tf i,j ×idf i ,

[0229] This formula represents the tf, idf, and tf-idf values ​​of the word numbered i in I within the j-th text knowledge. Here, a represents the number of text knowledge items, j = 1, 2, 3, ..., a, j are text numbers, there exists a defined word set I, which is the complete set of possible keywords required for a TF-IDF task, k is the number of words in I, i = 1, 2, 3, ..., k, i is the number of each word in I, n... i，j m represents the number of times the word numbered i appears in the j-th knowledge item. i It is the number of texts containing the word numbered i, which appears in all textual knowledge.

[0230] Therefore, in this invention, after the text knowledge is organized, it is necessary to determine the set of all possible keywords required for the task. A total of five word sets are needed: crop category word set, crop planting problem word set, problem handling measure word set, crop growth status word set, and crop growth environment word set. This is also the focus of the entire invention, because these word sets need to include most of the terms related to crop planting from industry standards and the internet, encompassing both standardization and fragmentation. After processing the text knowledge using each of the five word sets, and setting the final number of keywords to be selected for each word set, the words representing the five characteristics of each piece of text data can be obtained. For the time of problem occurrence, it can be directly extracted from the timestamp of the problem occurrence in the data source and converted into "month". At this point, the text knowledge processing is complete.

[0231] The following is a brief explanation of image knowledge processing methods.

[0232] In step 2, information extraction from image knowledge can be equivalent to using various image understanding methods. In this respect, the chosen method is not limited to any one specific approach; it can be selected based on the effectiveness of different methods.

[0233] The extracted information mainly includes crop category, crop planting problem, crop growth status, and crop growth environment. This information is consistent with the feature names of the final data. Crop category primarily refers to the name of the crop contained in the image. This is clarified within a defined range using object detection algorithms such as Mask R-CNN (Mask Region-based Convolutional Neural Network) and the YOLO (YouOnly Look Once) series. For crop planting problems, if the image carries a label indicating a crop planting problem, image classification models such as CNN (Convolutional Neural Network) and VGG (Visual Geometry Group) can be used to clarify the problem. For crop growth status, visual language pre-training models such as CLIP (Contrastive Language-Image Pre-training) can be used to train the matching of the detected portion (i.e., the specific crop) with the status description. Then, based on the pre-trained model, the crop growth status is generated. For the crop growth environment, visual language pre-training models such as CLIP can be used to train the matching of the image with its environmental description. Then, for each image, a description of the environment faced by the crop in the image is obtained.

[0234] As can be seen from step 2, this invention requires, before formally processing fragmented image knowledge, obtaining a good pre-trained model using an open-source dataset for the chosen method, and then using this model for image knowledge processing. Alternatively, images can be manually labeled directly, skipping the pre-training step. Then, different keywords are obtained according to different training models, and these keywords are combined into a prescribed form to complete the extraction of image knowledge information.

[0235] Furthermore, after the data transformation is complete, the original knowledge is not deleted, but stored for use in some functions of the service model designed in step 8.

[0236] The table below provides a clear explanation of each feature in this invention.

[0237]

[0238]

[0239] Step 3: Knowledge Integration

[0240] The primary task of Step 3 is to perform NLP (Natural Language Processing) synonym processing to address the ambiguity in the expression and meaning of fragmented knowledge. Because agricultural knowledge on the internet comes from diverse sources, different expressions may correspond to the same meaning, or the same expression may have different meanings. This ambiguity can negatively impact subsequent knowledge processing and organization. Therefore, the purpose of Step 3 is to unify the expression of the data obtained in Step 2 and eliminate ambiguity.

[0241] Previously, all the original knowledge had been transformed into the structured data set in step 2. Therefore, this step mainly involves merging synonyms and processing ambiguous words for all keywords under each feature. The industry standard word set will be used as the benchmark, and other words with similar expressions will be deleted.

[0242] The Word2vec (Word to Vector) model will be used to determine whether words from an industry-standard vocabulary set are synonyms with other words. The main steps are to identify words from industry-standard knowledge for each keyword under each feature in all the data, and then perform the following operations on each of these words:

[0243] The first step is to use Word2vec to obtain the word vector for this word (hereinafter referred to as the standard word);

[0244] The second step is to obtain the word vector for each other word in non-industry standard knowledge (hereinafter referred to as other words) using Word2vec, and then calculate the similarity between the standard word and other words. If the similarity is greater than the threshold, it is judged as a synonym; otherwise, it is not a synonym.

[0245] The third step is to determine if a word is a synonym, and then delete all other words under that feature; otherwise, no action is taken.

[0246] This process will handle the entire dataset, ensuring that each word in the dataset is essentially an industry-standard description, and greatly reducing redundancy and complexity. For those words that are not processed and are not in the industry standard, they will have their own meanings. They are knowledge that is not included in the industry standard knowledge, and they are fragmented knowledge with metaphorical and disordered characteristics, which will be reflected in the dataset.

[0247] Step 4: Data Classification

[0248] This step will classify the data processed in step 3 according to two feature dimensions: "crop category" and "crop planting problem". Classification based on "crop category" is to prepare for the core network in step 5, while classification based on "crop planting problem" will play a role in the construction of the relevance table in step 5.

[0249] This step does not require the intervention of an NLP model. When classifying by "crop category", simply group data with consistent "crop category" characteristics together. The same applies to classification based on "crop planting issues".

[0250] Step 5: Generate Network and Tables

[0251] Corresponding generation of network and table Figure 1 The section circled in red, "Network and Table Generation," is the core of the entire processing framework. Based on the results of steps 3 and 4, a core network centered on "crop category" and a relevance table focusing on "crop planting issues" and "issue handling measures" will be generated.

[0252] Looking at the entire knowledge of crop cultivation, we can find that its inherent logic is a process from determining the crop category, observing the crop's growth status, considering environmental factors and the timing of the problem's occurrence, to identifying the crop cultivation problem and proposing solutions based on all available information. The environment and the timing of the problem's occurrence play a supporting role in this process. Therefore, the generation of networks and tables must also adhere to this inherent logic.

[0253] The generation process of the core network is described below.

[0254] Based on the classification results by "crop category" in step 4, we obtain many sets of data. Each set of data shares the same "crop category" characteristic, thus resulting in a core network for each set. The method is as follows:

[0255] The first step is to construct a tree. Logically, a tree can be built based on three features: "Crop Category," "Crop Planting Problem," and "Problem-Solving Measures." The "Crop Category" of this data set is used as the root node. Next, "Crop Planting Problem" is used as a second-level child node, connected to the root node. Then, for each second-level child node, "Problem-Solving Measures" is used as a third-level child node, connected to its corresponding second-level child node. All data under the same second-level child node share the same "Crop Category" and "Crop Planting Problem." At this point, a tree structure based on these three features has been established. This tree structure can now simply represent the hierarchical and dependency relationships between agricultural knowledge items. Each node represents a specific feature or attribute, and the path from the root node to the leaf nodes already forms a rudimentary form of agricultural knowledge entries.

[0256] The second step is to add branch information. For two adjacent nodes N1 and N2, assuming N1 is the node closer to the root node, the branch N1N2 represents the thought process from N1 to N2, specifically manifested in two aspects: starting with a crop category (including crop growth environment, crop growth status, and the time the problem occurred), the process of determining what crop the problem is; and starting with a crop planting problem, the process of determining what problem-solving measures should be used to solve it. It is presented in probabilities, representing the likelihood of obtaining each answer after encountering these situations. The specific calculation formula is as follows:

[0257]

[0258] Figure 2 This diagram illustrates the final tree structure constructed from a set of data. It assumes there are *n* types of "crop planting problems" in this dataset, and simply uses numbers to represent "crop planting problems" and "problem-solving measures." Each path from the root node to a leaf node represents a combination of features: "crop category," "crop planting problem," and "problem-solving measure." Each combination contains all data in the dataset that share the same feature combination.

[0259] Regarding the addition of the remaining three features: Logically, it's reasonable to determine crop planting problems based on the crop growth environment, the time the problem occurred, and the crop's condition at that time. Therefore, these features should be placed between "crop category" and "crop planting problem," and they will also influence the relationship between "crop planting problem" and "problem-solving measures." However, according to step 2, the number of keywords for "crop growth environment" and "crop growth status" is unknown. The reason is simple: for the same crop category and the same crop planting problem, different people will produce different descriptions, even after step 3. This is a result of "fragmentation," so limiting the number of keywords for these two features will affect the final decision. Once incorporated into the tree structure, adding so many keywords is clearly inappropriate. Therefore, another way is needed to represent the "thinking process" between "crop category" and "crop planting problem," and between "crop planting problem" and "problem-solving measures." As for "problem occurrence time," although it is simple, it is insufficient to be a condition for independently judging "crop planting problem." Therefore, it should be considered together with "crop growth environment" and "crop growth status." The method for adding these three features to the tree structure will be described in detail in step 6.

[0260] The process of generating the relevance table is described below.

[0261] Ultimately, this invention aims to provide a decision-making method centered on the organization of fragmented knowledge about crop cultivation. This correlation network serves as a preprocessing step for the decision-making method. It primarily addresses the correlation between "crop category" and "crop cultivation problem," the correlation between "crop cultivation problem" and "problem-solving measures," and other correlations.

[0262] In this generation process, the symbols are defined as follows: "Crop growth environment," "problem occurrence time," and "crop growth state" each have word sets E, T, and S, containing o, m, and n keywords respectively. These sets contain all words under the corresponding features in the dataset. Let a represent a crop, a∈A, where A is the crop set; b represent a crop planting problem, b∈B, where B is the crop planting problem set; c represent a problem-solving measure, c∈C, where C is the problem-solving measure set; e∈E, t∈T, and s∈S.

[0263] 1. Crop categories and crop cultivation issues

[0264] There is a relationship between "crop category" and "crop planting problem," and this relationship largely depends on "crop growing environment," "problem occurrence time," and "crop growth status." To determine the correlation between "crop category" and "crop planting problem," in addition to the probability generated in the core network, the influence of "crop growing environment," "problem occurrence time," and "crop growth status" should also be considered. This influence can essentially be expressed as "how much influence different crop growing environments, problem occurrence times, and crop growth status keywords can have on the combination of a specific crop category and crop planting problem, that is, how much influence they have on the relationship between that crop category and crop planting problem." This influence is represented by distance; the greater the distance, the smaller the influence. It can be calculated in the following way.

[0265] Based on the classification results of "crop category" in step 4, we get many sets of data. Each set of data has the same "crop category" characteristic. Therefore, for each set of data, we will get a correlation table between "crop category" and "crop condition".

[0266] Suppose that for crop category a, one possible crop planting problem is b, and there are certain values ​​e, t, and s. Then, the magnitudes of the influence of e, t, and s on "crop category a and its crop planting problem b" are as follows:

[0267]

[0268]

[0269]

[0270] Where, ω E ω T ω S , where is the weight of each of the three features, all less than 1, depending on the magnitude of their influence on the transition from "crop category" to "crop condition problem". Generally, "crop growth status" has the greatest influence, followed by "crop growth environment", and "problem occurrence time" has the least influence. a，b Let n be the number of data entries where "crop category" is a and "crop planting problem" is b. e，a，b In this case, the number of data entries containing the word 'e' is n. t，a，b and n s，a，b Similarly.

[0271] For a crop, we can calculate the impact of three characteristics on different crop planting problems. At this time, we can build a correlation table between the "crop" and the "crop condition" according to the following method.

[0272] The first step is to calculate the probability P based on the branches between the "crop category" node a and the corresponding "crop planting problem" node b, calculated during the core network generation process. ab Determine the distance between "Crop Planting Problem" and "Crop Category" using the formula below.

[0273] D ab =-lnP ab

[0274] This initially establishes the relationship between crop categories and the planting issues of each crop; the closer the relationship, the stronger the connection.

[0275] The second step involves focusing on each crop planting problem and performing the following operations: For a crop planting problem, calculate the W value for each keyword using the formula above, and determine the distance between each keyword and the crop planting problem using the formula below.

[0276] d = -lnW

[0277] This allows us to determine the relationship between various keywords and the crop planting problem.

[0278] The third step is to create a table. Using each keyword as a column and each crop planting question as a row, for a given crop category 'a', all possible crop planting questions are b1, b2, ..., b1. The relevance table based on this is as follows:

[0279]

[0280] This table clearly shows the relationship between crops and crop conditions, influenced by three characteristics. The smaller the value in the table, the stronger the relationship. The number of such tables corresponds to the number of "crop" types in the total data.

[0281] 2. Crop planting problems and solutions

[0282] Clearly, there is a strong correlation between "crop planting problems" and "problem-handling measures." The crop itself can also significantly influence the final treatment method. In addition, "crop growing environment," "time of problem occurrence," and "crop growth status" also have subtle influences on the treatment method. Logically, it is obvious and correct that the "problem-handling measures" are jointly determined by the other five characteristics. The relationship between each characteristic and the "problem-handling measures" can be calculated as follows, using distance to represent the relationship; the greater the distance, the weaker the relationship.

[0283] Based on the classification results of "crop planting problems" in step 4, we obtain many sets of data. Each set of data shares the same "crop planting problem" characteristics. Therefore, for each set of data, we can obtain a correlation table between "crop planting problems" and "problem-solving measures." The table is constructed as follows for each set of data, i.e., for each specific "crop planting problem" b:

[0284] For a problem-solving measure c, and given e, t, and s, the magnitudes of the influence of e, t, and s on "crop planting problem b with problem-solving measure c" are respectively:

[0285]

[0286]

[0287]

[0288] It is largely consistent with the calculation of the relationship between "crop category" and "crop planting problem".

[0289] Similarly, the magnitude of the influence of crop category a on "crop planting problem b and its problem-solving measure is c" is:

[0290]

[0291] ω A This is the weight of "crop category" when calculating this effect; it is greater than ω. E ω T ω S They are all much smaller. b，c Let n be the number of data entries where "crop planting problem" is b and "problem-solving measure" is c. e，b，c In this case, the number of data entries containing the word 'e' is n. t，b，c n s，b，c and n a，b，c Similarly.

[0292] Similarly, the distances from each keyword in "Crop Category," "Crop Growing Environment," "Time of Problem Occurrence," and "Crop Growing Status" to "Problem Handling Measures," and the distance from "Crop Planting Problem" to "Problem Handling Measures," are also calculated using the following two formulas:

[0293] d = -lnW

[0294] D = -lnP

[0295] At this point, P is the value of the branch between the "crop planting problem" node and the corresponding "problem handling measures" node calculated during the core network generation process.

[0296] Finally, following the method used to create the correlation table for "Crop Category" and "Crop Planting Issues," the "Crop Planting Issues" category was changed to "Issuance Measures," and a "Crop Category" column was added. The table was then filled in according to the calculated data, resulting in this correlation table between "Crop Planting Issues" and each "Issuance Measures." The number of such tables corresponds to the number of types of "Crop Planting Issues" in the total data.

[0297] At this point, the core network and correlation table are complete, and they will play important roles in steps 6 and 7, respectively.

[0298] Step 6: Knowledge Organization

[0299] Follow these steps to build the final fragmented agricultural knowledge organization structure from the core network in step 5. At this point, the number of trees in the core network corresponds to the number of crop categories, which is the number of different attributes of the feature "crop category".

[0300] The first step is to disconnect the root node in each tree from the child nodes of the second level, that is, to disconnect the connection between "crop category" and "crop planting problem", and to make "crop planting problem" a child node of the third level.

[0301] The second step is to fill in all the keywords "crop growth environment", "problem occurrence time" and "crop growth status" in the corresponding dataset between "crop category" and "crop planting problem" as new second-level child nodes.

[0302] The third step is to connect the root node and all child nodes of the second level. For each tree and its corresponding dataset, traverse each data point in the dataset and perform the following operations on each data point: list all keywords for "crop growth environment", "problem occurrence time", and "crop growth status", find these keywords in the second-level child nodes and label them, and at the same time find the keyword "crop planting problem" for this data point in the third-level child nodes and label it. Then, for each labeled keyword in the second level, connect it with the root node and the labeled keywords in the third level.

[0303] The fourth step is to merge the same words in the child nodes of the fourth layer. At this point, the child nodes of the fourth layer are keywords for "problem handling measures". Merge the same keywords together and inherit all the original connections.

[0304] The fifth step is to combine all the trees processed in steps one through four, and then merge all the child nodes from steps two, three, and four, following the method used in step four. The result of this step is equivalent to the following: for any child node, any two keywords corresponding to any connection to it, including that node, can be found simultaneously in some original data.

[0305] This completes the organization of fragmented knowledge about crop cultivation. Figure 3 The example shown has two keywords for "crop category," four keywords for "crop growth environment," "problem occurrence time," and "crop growth status," three keywords for "crop characteristics," and three keywords for "problem handling measures." The diagram simply illustrates each step of this process, as well as the initial state and the final organizational structure.

[0306] However, this organization is formed only by existing data. In other words, if new data tries to use this structure, it will inevitably cause errors. Therefore, it is necessary to determine a method to utilize this data structure when the data is not the original data.

[0307] Step 7: Decision Calculation

[0308] Based on the organizational structure of fragmented knowledge about crop planting obtained in step 6, there are two basic calculation methods: from "crop growth status" to "crop planting problem"; and from "crop planting problem" to "problem-solving measures." These calculation processes will be explained below.

[0309] 1. From "Crop Growth Status" to "Crop Planting Issues"

[0310] In this calculation process, the new data includes crop categories, corresponding crop growth status, crop growth environment, and the time when the problem occurred. The problem to be solved is the crop planting problem.

[0311] The specific calculation method is as follows:

[0312] The first step is to process the new data in step 2, find the corresponding keywords for "crop category", "crop growth environment", "problem occurrence time" and "crop growth status" in the organizational structure (ignoring default items), and mark the corresponding second-level child nodes. Then, traverse the marked second-level child nodes in the organizational structure to find all third-level child nodes connected to any one of them. After traversing, a series of third-level child nodes are obtained, which are the corresponding keywords for "crop planting problem".

[0313] The second step is to list the obtained keywords for "crop planting problems" and, for each keyword for "crop planting problems", list the keywords corresponding to the second-level sub-nodes that are related to it, and sort them according to the characteristics of "crop growth environment", "time of problem occurrence" and "crop growth status".

[0314] The third step is to calculate the relation number for each keyword "crop planting problem" b under the specific keywords "crop growth environment", "problem occurrence time", and "crop growth status" for that crop category a, using the following formula:

[0315]

[0316] Here, ω is a weight representing the "credibility" of the existing data structure. This value is related to the amount of data used to build and optimize the data structure, the quality of the data, its internal distribution, and its source. Simply put, the better the data, the smaller ω is. n is the number of keywords in the new data, such as "crop growth environment," "problem occurrence time," and "crop growth status," that are the same as the keywords represented by all child nodes in the second level of the organizational structure. ∑d is the sum of the values ​​read from the correlation table between "crop category" and "crop planting problem" after matching all the keywords related to the crop planting problem.

[0317] The fourth step is to take the smallest one or two Hb values ​​among all the obtained values. The corresponding crop planting problem is the crop planting problem calculated under the new data.

[0318] 2. From "Crop Planting Problems" to "Problem-Solving Measures"

[0319] In this calculation process, the new data includes crop category, crop growth status, crop growth environment, time of problem occurrence, and crop planting problem; what needs to be determined is the problem-solving approach. Similarly, the specific calculation method is as follows:

[0320] The first step is to process the new data in step 2, find the corresponding keywords "crop growth status", "crop growth environment", "problem occurrence time" and "crop planting problem" in the organizational structure (ignoring default items), and mark the corresponding second and third level child nodes. Then, traverse the marked third level child nodes in the organizational structure to find all the fourth level child nodes connected to them. After traversing, a series of fourth level child nodes are obtained, which correspond to the keyword "problem handling measures".

[0321] The second step is to list the obtained "problem-handling measures" keywords. For each keyword c, calculate the keywords "crop growth environment," "problem occurrence time," and "crop growth status" in the new data for crop category a, and determine the coefficient under crop planting problem b. The formula is as follows:

[0322]

[0323] The third step is to retrieve all the obtained H values. c The smallest 1 to 2 values ​​correspond to the problem-solving measures calculated under the new data.

[0324] In one embodiment, according to Figure 4 Step 8 provides a detailed explanation of the content.

[0325] Step 8: Crowd Intelligence Service Model

[0326] Steps 1 through 6 provide a processing flow for fragmented knowledge of crop cultivation, and step 6 gives an organizational structure. Step 7 gives two basic calculation processes related to crop cultivation based on this structure. In this step, a service model will be established based on the above information.

[0327] Figure 4 The document details the functions of the service model, which can be summarized into five main modules. The core of this service model is the directed graph structure obtained in step 6. By adding an input module, an output module, a user interaction module, and a correct / incorrect judgment module to this structure, the basic service model can be constructed.

[0328] Directed graph structure: used to carry the core of the model and store knowledge.

[0329] Input module: Used for question input, text input, and image input.

[0330] Output module: Used for text output, image output, and special output.

[0331] User interaction module: Used for interaction between users and interaction between user and service models.

[0332] True / False Module: Used for knowledge assessment.

[0333] This model can fulfill basic functions: externally, for a specific crop, users can query crop planting issues under current conditions (whether growth is normal, whether there is disease, etc.), find solutions after encountering problems (including crop disease, what to do at a certain stage of crop growth, etc.), consult correct methods and information for crop planting, and upload their own text and pictures to seek answers or answer questions, etc.; internally, it has the ability to judge the correctness of multiple knowledge, learn new knowledge on its own within the learning framework, and add new knowledge on the existing structure.

[0334] This model can fulfill the following basic functions:

[0335] 1. Querying crop planting problems under current conditions: The model can determine whether the crop is growing normally or is diseased based on the input crop category and crop growth environment conditions;

[0336] 2. Handling of problems encountered during queries: When users encounter problems with crops, such as crop disease or abnormal growth, the model can provide corresponding problem-solving measures and suggestions;

[0337] 3. Consult on correct methods and information for crop cultivation: Users can inquire about correct methods and information for crop cultivation, such as when cultivation problems occur, growing conditions, etc.

[0338] 4. Upload text and images to seek responses: Users can upload text and images to seek responses or explanations from the model for specific questions, or to seek responses from other users.

[0339] 5. Ability to judge correctness by combining knowledge from multiple sources: The model can combine knowledge from multiple sources to judge the input information and confirm its correctness;

[0340] 6. Adding new knowledge and deleting redundant knowledge on the existing structure: The model can continuously improve its knowledge base based on the existing knowledge structure, improve its performance and accuracy, and adapt to the needs of more different scenarios.

[0341] Therefore, the implementation process of the above functions at the structural and computational levels will be described below. The methods for querying crop planting problems under the current circumstances and querying problem-solving measures after encountering problems have already been provided in step 7.

[0342] 1. Consult the correct methods and information

[0343] The calculation process for this function is actually a combination of "crop growth status" to "crop planting problems" and "crop planting problems" to "problem-solving measures." When agricultural workers encounter such problems, after determining the "crop type," "crop growth environment," "problem occurrence time," and "crop growth status," the organizational structure will give a judgment of "growing well" or other questions, and then provide corresponding problem-solving measures accordingly.

[0344] 2. Upload your own text and images

[0345] After the data is uploaded, it becomes new data. After step 2, it is transformed into a standard format, and then the next operation is carried out according to the problem.

[0346] 3. Knowledge Services

[0347] It is unrelated to structure and computation; it is a communication between two users.

[0348] 4. Judge whether the statements are true or false.

[0349] The calculation process for determining correctness is similar to the two basic processes in step 7, except that "crop planting problem" and "problem handling measures" are treated as unknowns first, and then the calculation is performed. The result obtained through the organizational structure is compared with the given result. If they are the same or the former basically includes the latter, then it is correct.

[0350] It is important to note that since the results given by the organizational structure are obtained in the final step by selecting the 1-2 results with the lowest correlation coefficient from multiple results, the following rule should be added to the judgment criteria for correctness: if the given result appears among all possible results before the organizational structure determines the final result, then it is partially correct; if it does not appear, it is incorrect.

[0351] 5. Learn and add new knowledge

[0352] The data uploaded by the user after step 2 is called the new data, and the data that constitutes the existing structure before uploading is called the original data.

[0353] Clearly, there's a high probability that the new data will differ from any of the original data. Therefore, after the "crop category" is correctly matched, the remaining nodes and connections will not perfectly correspond. If the new data is determined to be correct, it can be added to the structure, which simultaneously completes the model's self-learning process. Adding knowledge is part of the model's own evolutionary process.

[0354] The method for adding is as follows:

[0355] The first step is to determine if the data is correct. If correct, continue; otherwise, terminate. This determination can be made by an expert or by the model itself.

[0356] The second step is to arrange the new keywords of the new data. If the new keywords exist in the original structure, they remain unchanged; if the new keywords do not exist in the original structure, the keywords are arranged in the corresponding layer of the organizational structure according to their respective characteristics, as new nodes.

[0357] The third step is to add connections between the layers based on the logic of the new data. That is, after the first step, traverse all nodes. If the keywords corresponding to nodes N1 and N2 in adjacent layers appear simultaneously in the new data, and the connection N1N2 does not exist, then add a connection.

[0358] This completes the computational process for adding new knowledge.

[0359] 6. Remove outdated or redundant knowledge.

[0360] Clearly, as more knowledge is added to the model, some outdated knowledge will become obsolete, unsuitable for current needs, or redundant. These problems primarily occur in the judgments related to "crop planting problems" and "problem-solving measures." In such cases, it's necessary to remove some connections corresponding to this knowledge from the structure. Deleting knowledge is part of the model's own evolution.

[0361] The deletion method is as follows:

[0362] The first step is to determine if the data entry has the aforementioned problems. If so, continue; otherwise, terminate. This determination should be made by an expert, or if the model's result is found to be inconsistent with the determined correct result.

[0363] The second step is to determine which two features are causing the problem and to list all the keywords corresponding to those features in the data.

[0364] The third step is to mark the corresponding nodes for these keywords in the structure.

[0365] The fourth step is to delete all the connections between the marked nodes.

[0366] This completes the calculation process for removing outdated or redundant knowledge. Note that this step may also delete some correct logic; therefore, it is recommended to add a corresponding correct piece of knowledge after the deletion.

[0367] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for constructing a crop planting service model based on fragmented knowledge processing organization, characterized in that, Includes the following steps: Step 1: Collect agricultural knowledge information for crop cultivation, including text and images; Step 2: Extract keywords based on the agricultural knowledge information and perform semantic segmentation so that each piece of agricultural knowledge information generates a corresponding text data; Step 3: Convert all keywords in the text data into keywords with standardized vocabulary descriptions according to their meanings; Step 4: Classify the processed text data sequentially from two feature dimensions: crop category and crop planting issues; Step 5: Express the crop planting problem and its solution process as reasoning logic, construct the core network and correlation table. The correlation table is used to evaluate the relationship between crop categories and crop planting problems, as well as the relationship between crop planting problems and problem-solving measures. Step 6: Based on the core network, construct a directed graph network from all keywords that have been converted into standard vocabulary descriptions; Step 7: Based on the directed graph network and the correlation table, provide a decision calculation for crop planting based on the directed graph network, which consists of reading the correlation table and formulas. Step 8: Based on the directed graph network and decision computation, a crop planting service model for fragmented agricultural knowledge is given; The generation process of the core network is as follows: Based on the classification results of crop categories, we obtain many sets of data. The crop category characteristics of each set of data are consistent. Therefore, for each set of data, we obtain a core network, as follows: The first step is to construct a tree. Following the inherent logic, a tree is built based on three features: crop category, crop planting problem, and problem-solving measures. The crop category of this data set is used as the root node; the crop planting problem is used as the second-level child node, and these are connected to the root node. Then, for each second-level child node, the problem-solving measures are used as the third-level child node, and these are connected to their corresponding second-level child nodes. All data under the same second-level child node have the same crop category and crop planting problem, thus establishing a tree structure based on these three features. The second step is to add branch information for nodes in two adjacent layers. and Assuming It is the node closer to the root node, and the middle branch. Representative from arrive The process of thinking about a crop cultivation problem begins with a crop category, its growing environment, growth status, and the time when the problem occurred, leading to the conclusion that it is a specific crop cultivation problem. This process then proceeds to determining the appropriate problem-solving measures for a given crop cultivation problem, presented in probabilities, calculated using the following formula: ； The creation process for the two types of correlation tables is as follows: A table showing the correlation between "crop category" and "crop planting issues" was created. There are separate glossaries for "crop growing environment", "time of problem occurrence", and "crop growing status". And the number of keywords contained therein are respectively These sets each contain all words under the corresponding features in the dataset, and are set as follows: For a certain crop, , For crop collection, Regarding the planting of a certain crop, , A collection of problems related to crop cultivation. For a certain problem-solving measure, , A set of problem-solving measures, , , , For crop categories One potential crop cultivation problem it may cause is There is a certain , , So at this time , , For "crop category is And its crop cultivation problem is The magnitudes of their impact are as follows: ， in, , , The weights of the three features—"crop growth status," "crop growth environment," and "problem occurrence time"—are all less than 1. This weighting depends on the magnitude of the influence of these three features on the "crop category" to "crop planting problem." Generally, "crop growth status" has the greatest weight, followed by "crop growth environment," and "problem occurrence time" has the least weight. For "crop category" And the "crop planting problem" is Number of data entries In this case, words appear Number of data entries In this case, words appear Number of data entries In this case, words appear The number of data entries is used to calculate the impact of three characteristics on different crop cultivation issues for a single crop.

2. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, In step 2, the semantic segmentation and extraction of text data are performed using the term frequency-inverse text frequency index method. During the extraction process, six keyword categories are identified: crop category, crop planting problem, crop growth environment, problem occurrence time, crop growth status, and problem handling measures. Keywords are selected by combining the frequency of keyword occurrence in the text and the number of texts containing the keyword.

3. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, In step 5, the reasoning logic gives the relationship between the question and the answer. First, the type of crop is determined. Then, by observing the state of the crop and combining the current environmental conditions and the time when the problem occurred, a judgment is made on the problem of the crop. Based on the problem of the crop, corresponding problem-solving measures are given.

4. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, In step 5, the core network uses crop category as the center of the network. Around the center, keywords related to possible crop planting problems and problem-solving measures related to crop category are logically organized to create a core network for each crop category. The directed graph network is formed based on the core network plus keywords with three characteristics: crop growth status, crop growth environment, and problem occurrence time. In the network structure of the directed graph network, each node represents a specific instance of a certain keyword, and the connection between nodes represents a thinking process or logical association, indicating the reasoning path from one keyword to another.

5. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, The correlation table provides a quantitative way to measure the correlation between different factors and uses the correlation table for reasoning. The reasoning logic includes the logic of judging crop problems by observing the crop's state, from crop growth status to crop planting problems, and the logic of selecting treatment methods based on problems in the crop planting process, from crop planting problems to problem treatment measures. The reasoning logic first infers possible answers from the organizational structure based on existing information, and then filters the most likely results through the correlation table.

6. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, Step 3 includes, The Word2vec model is used to determine whether words from an industry-standard vocabulary set are synonyms with other words. Word vectors are obtained from standard words using Word2vec; For each other word in the non-industry standard knowledge, after obtaining the word vector using Word2vec, the similarity between the standard word and the other words is calculated. If the similarity is greater than a threshold, it is judged as a synonym; otherwise, it is not a synonym. If a word is identified as a synonym, all other words under that feature are deleted; otherwise, no action is taken.

7. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, Create a correlation table between "crop" and "crop planting issues" using the following method. The first step is to determine the "crop category" based on the data generated during the core network generation process. Nodes and their corresponding "crop planting problems" The probability of calculating branches between nodes Determine according to the formula below The distance between "crop planting issues" and "crop categories" , ， ， The second step involves focusing on each crop planting problem. For each crop planting problem, the formula above is used to calculate the value of each keyword. Determine the distance between each keyword and the crop planting question using the formula below. ， Determine the relationship between each keyword and the crop planting problem. The third step is to create a table, with each keyword as a column and each crop planting question as a row. This will help you identify a specific crop category. Fill in all distance values ​​to build a correlation table; Creation of a correlation table between "Crop Planting Issues" and "Issuance Measures": There are separate glossaries for "crop growing environment", "time of problem occurrence", and "crop growing status". , , And the number of keywords contained therein are respectively , , These sets each contain all words under the corresponding features in the dataset, and are set as follows: For a certain crop, , For crop collection, For the problem of planting a certain crop, , A collection of problems related to crop cultivation. For a certain problem-solving measure, , A set of problem-solving measures, , , , For a problem-solving approach And there are , , So at this time , , Regarding "crop planting issues" And its problem-solving measures are as follows The magnitudes of their impact are as follows: ， Meanwhile, crop categories Regarding "crop planting issues" And its problem-solving measures are as follows The magnitude of the impact is: ， in, , , The weights of the three characteristics—"crop growth status," "crop growth environment," and "problem occurrence time"—are all less than 1. This weighting depends on the magnitude of the influence of these three characteristics on the transition from "crop planting problem" to "problem-solving measures." Generally, "crop growth status" has the greatest weight, followed by "crop growth environment," and "problem occurrence time" has the least weight. This refers to the weight of "crop category" when calculating this impact; it is greater than... , , Small, For "crop planting issues" And the "problem-handling measures" are Number of data entries In this case, words appear Number of data entries In this case, words appear Number of data entries In this case, words appear Number of data entries In this case, words appear Given the number of data entries, for a single crop, the impact of four characteristics was calculated to address different crop cultivation issues. Create a correlation table between the "crop planting issues" and "problem-handling measures" using the following method; The first step is to solve the "crop planting problem" calculated during the core network generation process. Nodes and corresponding "problem-handling measures" The probability of calculating branches between nodes Determine according to the formula below The gap between "crop planting problems" and "problem-solving measures" , ， ， The second step involves focusing on each problem-solving measure. For each problem-solving measure, the formula above is used to calculate each keyword. Determine the distance between each keyword and the crop planting question using the formula below. ， Determine the relationship between each keyword and the problem-solving measures, given the identified problem-solving measures. The third step is to create a table, with each keyword as a column and an "Crop Category" column, and each problem-solving measure as a row. This will allow you to define a table for a specific crop planting problem. Fill in all distance values ​​to build a correlation table.

8. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, In the decision-making calculations that shift from "crop growth status" to "crop planting problems", The first step, for a single data entry, is to locate the corresponding keywords for "crop category," "crop growing environment," "problem occurrence time," and "crop growth status" within the organizational structure and mark the corresponding second-level child nodes. Then, iterate through the marked second-level child nodes in the organizational structure, finding all third-level child nodes connected to any of them. After this process, a series of third-level child nodes are obtained, representing the corresponding keywords for "crop planting problem." The second step is to list the obtained keywords for "crop planting problems," and for each keyword, list the keywords corresponding to the second-level child nodes that are related to it, and categorize them according to the characteristics of "crop growth environment," "problem occurrence time," and "crop growth status." The third step is to analyze each keyword related to "crop planting problems". Calculate in this crop category The formula for the number of relations under specific keywords such as "crop growth environment", "time of problem occurrence", and "crop growth status" is as follows: ， in, It is a weight that represents the "credibility" of the existing data structure; It is the distance between "crop planting issues" and "crop categories"; This refers to the total number of keywords in the data related to "crop growth environment," "problem occurrence time," and "crop growth status," which is the same as the total number of keywords represented by all child nodes in the second level of the organizational structure. This is the sum of values ​​read from the correlation table between "crop category" and "crop planting problem" after matching various keywords related to the crop planting problem. Fourth step, retrieve all the obtained results. The smallest 1-2 values ​​correspond to the crop planting problems calculated under that data. In the decision-making calculation from "crop planting problem" to "problem-solving measures", The first step is to find the corresponding keywords "crop growth status", "crop growth environment", "problem occurrence time" and "crop planting problem" in the organizational structure for a single data point, and mark the corresponding second and third level child nodes. Then, traverse the marked third level child nodes in the organizational structure to find all the fourth level child nodes connected to them. After traversing, a series of fourth level child nodes are obtained, which correspond to the keyword "problem handling measures". The second step is to analyze each obtained "problem-solving measure" keyword. Calculated in crop category Under the keywords marked "crop growth environment", "problem occurrence time", and "crop growth status", the crop planting problem is... The relation coefficient for time is given by the following formula: ， in, It is a weight that represents the "credibility" of the existing data structure; It is the distance between "crop planting problems" and "problem-solving measures"; This refers to the total number of keywords in the data related to "crop growth environment," "problem occurrence time," and "crop growth status," which is the same as the total number of keywords represented by all child nodes in the second level of the organizational structure. This is the sum of values ​​read from the correlation table between "crop planting issues" and "issue handling measures" after matching the problem-solving measures with the relevant keywords. The third step is to retrieve all the results. The smallest 1-2 values ​​correspond to the problem-solving measures calculated for that data.

9. The method for constructing a crop planting service model based on fragmented knowledge processing organization according to claim 1, characterized in that, In step 8, based on the directed graph network and decision computation, a crop planting service model for fragmented agricultural knowledge is provided. Based on the directed graph structure, an input module, an output module, a user interaction module, and a right / wrong judgment module are added to construct the service model. On this basis, service functions are provided and the implementation logic of each function is provided. The service model has the following functions: Querying crop planting issues under current conditions: The model can determine whether the crop is growing normally or is diseased based on the input crop category and crop growth environment conditions; How to handle problems encountered during queries: When users encounter problems with crops, such as crop disease or abnormal growth, the model can provide corresponding problem-solving measures and suggestions. Consultation on correct methods and information for crop cultivation: Users can search for correct methods and information on crop cultivation, including when cultivation problems occur and growth conditions; Upload text and images to seek answers: Users can upload text and images to seek answers or explanations from the model for specific questions, or to seek answers from other users; The ability to judge right and wrong by combining knowledge from multiple sources: The model can combine knowledge from multiple sources to judge the input information and confirm its correctness; Adding new knowledge and removing redundant knowledge on the existing structure: The model can continuously improve its knowledge base based on the existing knowledge structure, improve its performance and accuracy, and adapt to the needs of more different scenarios. Logic implementation of some functions: Consult the correct methods and information. Its function is achieved by combining two decisions, namely, from "crop growth status" to "crop planting problem" and from "crop planting problem" to "problem handling measures", or one of them. When agricultural workers have such problems, after determining the "crop type", "crop growth environment", "problem occurrence time" and "crop growth status", the answer to the problem is obtained through a directed graph structure. Upload your own text and images. After the data is uploaded, it becomes new data. After step 2, it is transformed into a standard format, and then the next operation is carried out according to the problem. Judge whether it is true or false. First, treat "crop planting problem" and "problem-solving measures" as unknowns, then perform calculations. Compare the results obtained through the directed graph structure with the given results. If they are the same or the former basically contains the latter, then it is correct. It should be noted that since the results given by the organizational structure are obtained in the last step by selecting the 1-2 results with the lowest relation coefficients from multiple results, the following rule should be added to the judgment criteria for correctness: if the given result appears among all possible results before the organizational structure determines the final result, then it is partially correct; if it does not appear, then it is incorrect. Learn and add new knowledge. The first step is to determine if the data is correct. If it is correct, continue; otherwise, terminate. This determination can be made by an expert or by the model itself. The second step is to arrange the new keywords in the new data. If the new keywords already exist in the original structure, they remain unchanged; otherwise, they are arranged according to their respective characteristics at the corresponding levels of the organizational structure, serving as new nodes. The third step is to add connections between the layers according to the logic of the new data. That is, after the first step, traverse all nodes. If the nodes and corresponding keywords of adjacent layers appear at the same time in the new data and there is no connection, then add a connection. Remove outdated or redundant knowledge. The first step is to determine if the data entry has the aforementioned problems. If so, continue; otherwise, terminate. This determination should be made by an expert, or if the model's result is found to be inconsistent with the determined correct result. The second step is to determine which two features are causing the problem and to list all the keywords corresponding to those features in that data entry. The third step is to mark the corresponding nodes for these keywords in the structure. The fourth step is to delete all the connections between the marked nodes. This step may delete some correct logic at the same time. Therefore, after the deletion operation, it is recommended to add a correct piece of knowledge corresponding to the deleted knowledge.

Images