An AI interactive question-answering method for migratory bird knowledge based on personalized customization

Abstract

The present invention discloses an AI interactive question-and-answer method of migratory bird knowledge based on personalized customization, and relates to the technical field of AI interactive question-and-answer. The present invention, based on a regional matching mechanism, extracts geographic information through user query analysis, and adopts K-nearest neighbor search matching knowledge base based on spatial index, introduces a migratory bird migration prediction model, calculates the distribution probability of migratory bird species in a specific time period based on a time series analysis method, and makes corrections in combination with climate change impact factors, so that AI can dynamically adjust answers; integrates global and local data through an adaptive knowledge graph, introduces regionalized entity linking technology, organizes migratory bird knowledge in a hierarchical manner, and calculates the influence of migratory bird species in the graph, so that query results are more accurate and can match local species that users are concerned about, and adopts semantic disambiguation, accent adaptation and image recognition methods to improve the ability of AI to parse different types of queries.

Description

Technical Field

[0001] The present invention relates to the technical field of AI interactive question-answering, and in particular to an AI interactive question-answering method of migratory bird knowledge based on personalized customization. Background Art

[0002] Artificial intelligence can provide scientific researchers and bird-watching enthusiasts with a convenient way to obtain information in interactive Q&A sessions on migratory bird knowledge. Relying on global data sets and natural language processing technology, AI can quickly answer common questions about migratory bird species, migration routes and ecological habits.

[0003] However, when users want to know about migratory birds in a specific area, it is difficult for AI to provide targeted localized information. The core of AI interactive Q&A lies in global public databases. Although these databases can provide a large amount of global distribution information of birds, they are not updated in a timely manner or have insufficient coverage for bird dynamics in specific areas, such as seasonal changes in migratory birds in a city, records of local bird watching organizations, or the impact of climate change on specific habitats in recent years. In addition, traditional AI mainly relies on knowledge graphs and text matching-based retrieval methods. When the data itself lacks regional segmentation, it is difficult for AI to give answers that are highly consistent with user questions.

[0004] Some systems have introduced retrieval strategies based on geographic information, such as combining the user's IP location or manually entered place names to filter the database for bird information that matches the region; however, since these methods still rely on the integrity of the existing database, when local data is missing or not integrated, AI still has difficulty providing accurate answers. Therefore, under the existing solutions, AI still faces obvious shortcomings in processing localized migratory bird knowledge and cannot meet users' needs for regionalized and dynamic bird information. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides an AI interactive question-and-answer method for migratory bird knowledge based on personalized customization to solve the problem that the current AI interactive question-and-answer method for migratory bird knowledge performs poorly in some localized migratory bird data.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: The embodiment of the present invention provides an AI interactive question-and-answer method based on personalized customization of migratory bird knowledge, which includes: Step S1, constructing a localized migratory bird knowledge base, integrating the global bird database and regional bird observation data, and performing structured storage of the data in the knowledge base; Step S2, establishing a regional matching mechanism based on the localized migratory bird knowledge base of step S1; Step S3, based on the regional matching mechanism of step S2, construct an adaptive knowledge graph, hierarchically integrate global and local migratory bird knowledge, and introduce regionalized entity linking technology to match the geographical location in the user query with the local migratory bird knowledge; Step S4, based on the adaptive knowledge graph of step S3, using multimodal analysis to perform target detection and classification, and matching bird species and habitat information in combination with the localized migratory bird knowledge base of step S1; Step S5, based on the multimodal analysis results of step S4, a causal reasoning model is used to dynamically predict and infer the migration pattern of migratory birds; Step S6, based on the causal reasoning model, combined with natural language generation technology, optimize and output the question and answer expression; Step S7, based on the question and answer statements generated in step S6, adopts a user feedback optimization mechanism to allow users to correct or supplement the answers, and dynamically adjusts the knowledge base data structure and question and answer weights in combination with reinforcement learning.

[0008] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, wherein: the global bird database includes the bird online database eBird and the Global Biodiversity Information Network; The regional bird observation data include observation records from bird watching associations, scientific research institutions and private individuals.

[0009] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized migratory bird knowledge described in the present invention, the steps of constructing a localized migratory bird knowledge base, integrating the global bird database and regional bird observation data, and performing structured storage of the data in the knowledge base are as follows: Standardize the features of the collected raw data and define a standardized migratory bird knowledge base : , in, represents the standardized migratory bird knowledge base, Indicates the data index, Represents the total number of records in the knowledge base, Indicates bird species, Indicates The geographical location of the records, Indicates The recorded observation time, Indicates Characteristics of birds, Indicates Information about the habitats of birds. Create data index : , in, represents the index structure of the migratory bird knowledge base, Indicates The classification level information of birds, Indicates Clustering categories of bird behavior characteristics.

[0010] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, the regional matching machine performs geographical identification on the query input by the user, matches the migratory bird data of a specific area, and dynamically updates the regional migratory bird information in combination with the migratory bird migration prediction model and climate change factors.

[0011] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, the step of establishing a regional matching mechanism based on the localized migratory bird knowledge base in step S1 is: When parsing user queries, define the user input query as , including the query content and geographic information : , in, Indicates a user query request. Indicates the query content. Indicates the geographical location information involved in the query. Using geographic matching function Analysis And match the migratory bird knowledge base : , in, is the geographic matching function, Indicates the geographical location information corresponding to the matched migratory bird data. The geographic matching function uses the K-nearest neighbor search KNN method based on spatial index for matching. The matching formula is: , in, Knowledge Base Geographic location information in is the geographic distance function, which uses the Haversing formula to calculate the distance between two longitudes and latitudes; Make bird migration predictions and define the migration time of migratory birds In position The probability distribution of : , in, Indicates the time when migratory birds In position The probability of represents the number of migratory bird populations, Indicates The weight of a migratory bird population, Indicates The migration distribution function of a migratory bird population, Incorporating climate change factors Modified prediction model: , in, is the corrected probability, is the climate change adjustment factor, Represents the influencing factor of climate factors; Based on the matched geographic location and predicted probability , dynamically update the query results, the update formula is: , in, For the final query matching result, For the query matching function, the user query With knowledge base and prediction models Conduct a joint analysis.

[0012] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge described in the present invention, the multimodal analysis includes: Text parsing: semantic disambiguation of text queries entered by users. Voice analysis: Accent adaptation for voice input, Image parsing: Detect and classify objects in user-uploaded bird photos.

[0013] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, the step of constructing an adaptive knowledge graph based on the regional matching mechanism in step S2 is: Based on the region information matched in step S2 and migratory bird data to build a migratory bird knowledge graph : , in, represents the migratory birds knowledge graph, Represents a set of nodes in the migratory bird knowledge graph, including bird species, geographic location, observation time, and habitat information. represents the set of edges between nodes, defining the relationships between migratory bird species, including predation relationships, habitat similarity, and migration paths. Represents a set of relationships in the knowledge graph, including belonging to the same habitat and having the same migration route; Divide the knowledge graph into layers and define hierarchical sets : , in, represents the migratory bird knowledge node at the global level, represents the migratory bird knowledge node at the local level, Calculate each migratory bird species node in the knowledge graph The influence of is calculated as: , in, Indicates migratory bird species Influence in the knowledge graph, Representation Node The number of connections, Indicates migratory bird species With neighboring migratory bird species The weight between is the similarity function between migratory bird species, defined as: , in, represents the habitat similarity distance between migratory bird species, is the normalization coefficient; Based on the geographic location in the user's query Perform regionalized entity matching, the matching formula is: , in, Indicates the matched migratory bird species node, is the matching function between migratory bird species nodes and geographical locations, defined as: , in, is the matching weight, represents the matching score based on geographic distance, represents the matching score based on migration pattern; The steps of using multimodal analysis to perform target detection and classification based on the adaptive knowledge graph in step S3 and matching bird species and habitat information in combination with the localized migratory bird knowledge base in step S1 are as follows: Perform multimodal analysis, including: Querying text entered by the user To perform semantic disambiguation, we use the semantic similarity calculation based on the Bert model: , in, represents the similarity between the query text and migratory bird species, is the Bert semantic vector of the query text, is the semantic vector of migratory bird species, is the cosine similarity function, Perform audio feature matching on user voice input: , in, represents the dynamic time warping (DTW) similarity between the speech input and the migratory bird sound, is the feature sequence of the user's voice input, is the characteristic sequence of sounds of migratory bird species, Perform object detection and classification on bird photos uploaded by users: , in, For the predicted bird species, is the classification probability of the image recognition model ResNet, For user-uploaded bird images, Based on the recognition results of multimodal analysis, the species and habitats in the migratory bird knowledge base are matched: , in, represents the matching migratory bird habitat, Knowledge-based Species matching function.

[0014] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, in which: in step S5, the migratory pattern of migratory birds is dynamically predicted by combining Bayesian network and time series analysis, inferring the factors affecting the distribution of migratory birds in a specific area, and optimizing the question-and-answer reasoning process by combining historical data and real-time data.

[0015] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, the step of dynamically predicting and reasoning the migratory pattern of migratory birds using a causal reasoning model based on the multimodal analysis result of step S4 is as follows: The causal inference model that defines the migratory pattern of migratory birds is : , in, represents a causal inference model, is the set of variables in the causal inference model, including climate factors, food resources and habitat, is the set of causal relationships between variables, is the set of conditional probability distributions in the causal graph, Modeling causal reasoning using Bayesian networks: , in, Representation variables At its parent node The probability distribution under given conditions, express The joint probability distribution with its parent node, represents the marginal probability of the parent node, Defining migratory bird species In time In position The probability distribution of is: , in, Represents a causal inference model migratory bird species Migration probability, is the number of migratory bird populations, is the weight of migratory bird population, Indicates The distribution of migration patterns of migratory bird populations, constrained by a causal inference model, The time series model is used to dynamically predict the migration pattern of migratory birds. The prediction formula is: , in, Indicates time When the bird is in position The predicted probability at is the time impact weight, For time Probability change rate of migratory birds: .

[0016] As a preferred solution of the AI ​​interactive question-and-answer method based on personalized customized migratory bird knowledge described in the present invention, the step of dynamically adjusting the knowledge base data structure and question-and-answer weights in combination with reinforcement learning is: Define the user feedback collection as: , in, represents the user feedback dataset, is the number of user interactions, For the User queries, The answer returned by the system. Ratings or correction suggestions given by users, Based on user feedback, reinforcement learning is used to optimize the knowledge base weights. The optimization formula is: , in, are the parameters of the knowledge base data structure, is the reinforcement learning reward function, is the learning rate, is the gradient of the reward function, The optimization goal is: , in, is the optimized knowledge base weight parameter, Based on parameters Computed question-answer generation probability.

[0017] The beneficial effects of the present invention are as follows: the present invention, based on a regional matching mechanism, extracts geographic information through user query analysis, and adopts K-nearest neighbor search based on spatial index to match migratory bird data in a knowledge base, introduces a migratory bird migration prediction model, calculates the distribution probability of migratory bird species in a specific time period based on a time series analysis method, and makes corrections based on climate change influencing factors, so that AI can dynamically adjust the answer.

[0018] The present invention integrates global and local data through an adaptive knowledge graph, introduces regionalized entity linking technology, organizes migratory bird knowledge in a hierarchical manner, and calculates the influence of migratory bird species in the graph, so that the query results are more accurate and can match the local species that users are concerned about; in terms of multimodal analysis, it integrates text, voice and image inputs, adopts semantic disambiguation, accent adaptation and image recognition methods, improves the ability of AI to parse different types of queries, and uses the knowledge base to match migratory bird species and habitat information, making AI more flexible in actual interactions.

[0019] The present invention adopts a causal reasoning model to construct a migratory bird migration pattern, uses a Bayesian network to describe the factors affecting migratory bird migration, and combines time series analysis to predict future migration distribution, so that when AI answers users' queries about migratory bird behavior trends, it can provide explanatory answers based on reasoning rather than simple static data matching; in addition, the present invention can dynamically adjust the knowledge base data structure and question and answer weights through user feedback optimization mechanism and reinforcement learning, continuously optimize the knowledge base according to users' correction opinions, and improve the accuracy of query matching.

[0020] In summary, the present invention enhances the ability to understand and update localized bird information, and effectively makes up for the deficiencies of existing AI interactive question-and-answer systems in terms of regional matching and dynamic prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 This is a flow chart of an AI interactive question-and-answer method based on personalized customization of migratory bird knowledge in Example 1. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1, reference Figure 1 This embodiment provides an AI interactive question-and-answer method based on personalized customized migratory bird knowledge, comprising the following steps: Step S1, constructing a localized migratory bird knowledge base, integrating the global bird database and regional bird observation data, and performing structured storage of the data in the knowledge base; The global bird database includes the online bird database eBird and the Global Biodiversity Information Network; Regional bird observation data include bird watching associations, scientific research institutions and private observation records; The steps to build a localized migratory bird knowledge base, integrate the global bird database and regional bird observation data, and store the data in the knowledge base in a structured manner are: Standardize the features of the collected raw data and define a standardized migratory bird knowledge base : , in, represents the standardized migratory bird knowledge base, Indicates the data index, Represents the total number of records in the knowledge base, Indicates bird species, Indicates The geographical location of the records, Indicates The recorded observation time, Indicates Characteristics of birds, Indicates Information about the habitats of birds. Create data index : , in, represents the index structure of the migratory bird knowledge base, Indicates The classification level information of birds, Indicates Cluster categories of bird behavior characteristics; Specifically, step S1 constructs a localized migratory bird knowledge base, integrates global bird database data and regional observation data, and stores migratory bird information in a standardized manner; Data indexes are used to improve query efficiency; Step S2, establishing a regional matching mechanism based on the localized migratory bird knowledge base of step S1; The regional matching machine performs geographic identification on the query input by the user, matches the migratory bird data of a specific area, and dynamically updates the regional migratory bird information by combining the migratory bird prediction model and climate change factors; Based on the localized migratory bird knowledge base in step S1, the steps for establishing a regional matching mechanism are: When parsing user queries, define the user input query as , including the query content and geographic information : , in, Indicates a user query request. Indicates the query content. Indicates the geographical location information involved in the query. Using geographic matching function Analysis And match the migratory bird knowledge base : , in, is the geographic matching function, Indicates the geographical location information corresponding to the matched migratory bird data. The geographic matching function uses the K-nearest neighbor search KNN method based on spatial index for matching. The matching formula is: , in, Knowledge Base Geographic location information in is the geographic distance function, which uses the Haversing formula to calculate the distance between two longitudes and latitudes; Make bird migration predictions and define the migration time of migratory birds In position The probability distribution of : , in, Indicates the time when migratory birds In position The probability of represents the number of migratory bird populations, Indicates The weight of a migratory bird population, Indicates The migration distribution function of a migratory bird population, Incorporating climate change factors Modified prediction model: , in, is the corrected probability, is the climate change adjustment factor, Represents the influencing factor of climate factors; Based on the matched geographic location and predicted probability , dynamically update the query results, the update formula is: , in, For the final query matching result, For the query matching function, the user query With knowledge base and prediction models Conduct joint analysis, Specifically, a regional matching mechanism is established here so that user queries can accurately match migratory bird data in the corresponding area: Parse user queries to extract geographic location information , and then through the geographic matching function Use K nearest neighbor search and Haversine distance calculation to match the closest migratory bird database location , further, using the bird migration prediction model Combined with climate change impact factors Correct the matching results and finally pass the matching function Dynamically update query results ,This mechanism can adapt to the migratory bird data in different regions and improve the accuracy and dynamism of the query; Step S3, based on the regional matching mechanism of step S2, construct an adaptive knowledge graph, hierarchically integrate global and local migratory bird knowledge, and introduce regionalized entity linking technology to match the geographical location in the user query with the local migratory bird knowledge; Step S4, based on the adaptive knowledge graph of step S3, using multimodal analysis to perform target detection and classification, and matching bird species and habitat information in combination with the localized migratory bird knowledge base of step S1; Multimodal analysis includes: Text parsing: semantic disambiguation of text queries entered by users. Voice analysis: Accent adaptation for voice input, Image analysis: Detect and classify bird photos uploaded by users, and optimize species recognition accuracy by combining knowledge base data; Based on the regional matching mechanism of step S2, the steps of constructing an adaptive knowledge graph are: Based on the region information matched in step S2 and migratory bird data to build a migratory bird knowledge graph : , in, represents the migratory birds knowledge graph, Represents a set of nodes in the migratory bird knowledge graph, including bird species, geographic location, observation time, and habitat information. represents the set of edges between nodes, defining the relationships between migratory bird species, including predation relationships, habitat similarity, and migration paths. Represents a set of relationships in the knowledge graph, including belonging to the same habitat and having the same migration route; Divide the knowledge graph into layers and define hierarchical sets : , in, represents the migratory bird knowledge node at the global level, represents the migratory bird knowledge node at the local level, Calculate each migratory bird species node in the knowledge graph The influence of is calculated as: , in, Indicates migratory bird species Influence in the knowledge graph, Representation Node The number of connections, Indicates migratory bird species With neighboring migratory bird species The weight between is the similarity function between migratory bird species, defined as: , in, represents the habitat similarity distance between migratory bird species, is the normalization coefficient; Based on the geographic location in the user's query Perform regionalized entity matching, the matching formula is: , in, Indicates the matched migratory bird species node, is the matching function between migratory bird species nodes and geographical locations, defined as: , in, is the matching weight, represents the matching score based on geographic distance, represents the matching score based on migration pattern; Specifically, in step S3, a knowledge graph of migratory birds is constructed to integrate global and local data. The influence of migratory bird species nodes is calculated using graph structure analysis to determine the hierarchical division of important species. Based on the geographic information in the user query, regionalized entity linking technology is used to Match the corresponding migratory bird species information, so that users can efficiently match the migratory bird knowledge of the corresponding area when searching; Based on the adaptive knowledge graph of step S3, multimodal analysis is used to detect and classify targets, and the steps of matching bird species and habitat information in combination with the localized migratory bird knowledge base of step S1 are as follows: Perform multimodal analysis, including: Querying text entered by the user To perform semantic disambiguation, we use the semantic similarity calculation based on the Bert model: , in, represents the similarity between the query text and migratory bird species, is the Bert semantic vector of the query text, is the semantic vector of migratory bird species, is the cosine similarity function, Perform audio feature matching on user voice input: , in, represents the dynamic time warping (DTW) similarity between the speech input and the migratory bird sound, is the feature sequence of the user's voice input, is the characteristic sequence of sounds of migratory bird species, Perform object detection and classification on bird photos uploaded by users: , in, For the predicted bird species, is the classification probability of the image recognition model ResNet, For user-uploaded bird images, Based on the recognition results of multimodal analysis, the species and habitats in the migratory bird knowledge base are matched: , in, represents the matching migratory bird habitat, Knowledge-based Species matching function; Specifically, step S4 uses a multimodal analysis method to improve the accuracy of migratory bird identification; Step S5, based on the multimodal analysis results of step S4, a causal reasoning model is used to dynamically predict and infer the migration pattern of migratory birds; In step S5, the migratory patterns of migratory birds are dynamically predicted by combining Bayesian networks with time series analysis, the factors affecting the distribution of migratory birds in a specific area are inferred, and the question-answering reasoning process is optimized by combining historical data with real-time data; Based on the multimodal analysis results of step S4, the steps of using the causal reasoning model to dynamically predict and infer the migration pattern of migratory birds are as follows: The causal inference model that defines the migratory pattern of migratory birds is : , in, represents a causal inference model, is the set of variables in the causal inference model, including climate factors, food resources and habitat, is the set of causal relationships between variables, is the set of conditional probability distributions in the causal graph, Modeling causal reasoning using Bayesian networks: , in, Representation variables At its parent node The probability distribution under given conditions, express The joint probability distribution with its parent node, represents the marginal probability of the parent node, Defining migratory bird species In time In position The probability distribution of is: , in, Represents a causal inference model migratory bird species Migration probability, is the number of migratory bird populations, is the weight of migratory bird population, Indicates The distribution of migration patterns of migratory bird populations, constrained by a causal inference model, The time series model is used to dynamically predict the migration pattern of migratory birds. The prediction formula is: , in, Indicates time When the bird is in position The predicted probability at is the time impact weight, For time Probability change rate of migratory birds: ; Specifically, step S5 dynamically predicts the migratory pattern of migratory birds based on the causal reasoning model: Building causal inference structures via Bayesian networks , describe the key factors affecting migratory bird migration, and then, based on Calculate the time of migratory birds The migration probability , and dynamically update it in combination with the time series model so that the prediction results can adapt to real-time changes, thereby improving the prediction accuracy of migratory bird migration patterns; Step S6, based on the causal reasoning model, combined with natural language generation technology, optimize and output the question and answer expression; Step S7, based on the question and answer expressions generated in step S6, a user feedback optimization mechanism is used to allow the user to correct or supplement the answers, and the knowledge base data structure and question and answer weights are dynamically adjusted in combination with reinforcement learning; The steps to dynamically adjust the knowledge base data structure and question-answer weights in combination with reinforcement learning are: Define the user feedback collection as: , in, represents the user feedback dataset, is the number of user interactions, For the User queries, The answer returned by the system. Ratings or correction suggestions given by users, Based on user feedback, reinforcement learning is used to optimize the knowledge base weights. The optimization formula is: , in, are the parameters of the knowledge base data structure, is the reinforcement learning reward function, is the learning rate, is the gradient of the reward function, The optimization goal is: , in, is the optimized knowledge base weight parameter, Based on parameters Calculated question-answer generation probability; Specifically, step S7 introduces user feedback optimization, and uses reinforcement learning to adjust the knowledge base structure and optimize parameters based on user feedback data. , so that the system can dynamically adjust the generation weight of questions and answers, improve the accuracy of answers and user satisfaction. This mechanism enables the question-answering system to have self-learning capabilities and continuously improve the quality of answers as users interact.

[0027] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A personalized AI interactive question-and-answer method for migratory bird knowledge, characterized by: include, Step S1, constructing a localized migratory bird knowledge base, integrating the global bird database and regional bird observation data, and performing structured storage of the data in the knowledge base; Step S2, establishing a regional matching mechanism based on the localized migratory bird knowledge base of step S1; Step S3, based on the regional matching mechanism of step S2, construct an adaptive knowledge graph, hierarchically integrate global and local migratory bird knowledge, and introduce regionalized entity linking technology to match the geographical location in the user query with the local migratory bird knowledge; Step S4, based on the adaptive knowledge graph of step S3, using multimodal analysis to perform target detection and classification, and matching bird species and habitat information in combination with the localized migratory bird knowledge base of step S1; Step S5, based on the multimodal analysis results of step S4, a causal reasoning model is used to dynamically predict and infer the migration pattern of migratory birds; Step S6, based on the causal reasoning model, combined with natural language generation technology, optimize and output the question and answer expression; Step S7, based on the question and answer statements generated in step S6, adopts a user feedback optimization mechanism to allow users to correct or supplement the answers, and dynamically adjusts the knowledge base data structure and question and answer weights in combination with reinforcement learning.

2. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 1, characterized in that: The global bird database includes the online bird database eBird and the Global Biodiversity Information Network; The regional bird observation data include observation records from bird watching associations, scientific research institutions and private individuals.

3. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 2, characterized in that: The steps of constructing a localized migratory bird knowledge base, integrating the global bird database and regional bird observation data, and performing structured storage of the data in the knowledge base are: Standardize the features of the collected raw data and define a standardized migratory bird knowledge base : ， in, represents the standardized migratory bird knowledge base, Indicates the data index, Represents the total number of records in the knowledge base, Indicates bird species, Indicates The geographical location of the records, Indicates The recorded observation time, Indicates Characteristics of birds, Indicates Bird habitat information, Create data index : ， in, represents the index structure of the migratory bird knowledge base, Indicates The classification level information of birds, Indicates Clustering categories of bird behavior characteristics.

4. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 3, characterized in that: The region matching machine performs geographic identification on the query input by the user, matches the migratory bird data of a specific area, and dynamically updates the regional migratory bird information in combination with the migratory bird migration prediction model and climate change factors.

5. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 4, characterized in that: The steps of establishing a regional matching mechanism based on the localized migratory bird knowledge base in step S1 are: When parsing user queries, define the user input query as , including the query content and geographic information : ， in, Indicates a user query request. Indicates the query content. Indicates the geographical location information involved in the query. Using geographic matching function Analysis And match the migratory bird knowledge base : ， in, is the geographic matching function, Indicates the geographical location information corresponding to the matched migratory bird data. The geographic matching function uses a spatial index based The nearest neighbor search KNN method is used for matching, and the matching formula is: ， in, Knowledge Base Geographic location information in is the geographic distance function, which uses the Haversing formula to calculate the distance between two longitudes and latitudes; Make bird migration predictions and define the migration time of migratory birds In position The probability distribution of : ， in, Indicates migratory birds at time In position The probability of represents the number of migratory bird populations, Indicates The weight of a migratory bird population, Indicates The migration distribution function of a migratory bird population, Incorporating climate change factors Modified prediction model: ， in, is the corrected probability, is the climate change adjustment factor, Represents the influencing factor of climate factors; Based on the matched geographic location and predicted probability , dynamically update the query results, the update formula is: ， in, For the final query matching result, For the query matching function, the user query With knowledge base and prediction models Conduct a joint analysis.

6. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 5, characterized in that: The multimodal analysis includes: Text parsing: semantic disambiguation of text queries entered by users. Voice analysis: Accent adaptation for voice input, Image parsing: Detect and classify objects in user-uploaded bird photos.

7. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 6, characterized in that: The steps of constructing the adaptive knowledge graph based on the regional matching mechanism in step S2 are: Based on the region information matched in step S2 and migratory bird data to build a migratory bird knowledge graph : ， in, represents the migratory birds knowledge graph, Represents a set of nodes in the migratory bird knowledge graph, including bird species, geographic location, observation time, and habitat information. represents the set of edges between nodes, defining the relationships between migratory bird species, including predation relationships, habitat similarity, and migration paths. Represents a set of relationships in the knowledge graph, including belonging to the same habitat and having the same migration route; Divide the knowledge graph into layers and define hierarchical sets : ， in, represents the migratory bird knowledge node at the global level, represents the migratory bird knowledge node at the local level, Calculate each migratory bird species node in the knowledge graph The influence of is calculated as: ， in, Indicates migratory bird species Influence in the knowledge graph, Representation Node The number of connections, Indicates migratory bird species With neighboring migratory bird species The weight between is the similarity function between migratory bird species, defined as: ， in, represents the habitat similarity distance between migratory bird species, is the normalization coefficient; Based on the geographic location in the user's query Perform regionalized entity matching, the matching formula is: ， in, Indicates the matched migratory bird species node, is the matching function between migratory bird species nodes and geographical locations, defined as: ， in, is the matching weight, represents the matching score based on geographic distance, represents the matching score based on migration pattern; The steps of using multimodal analysis to perform target detection and classification based on the adaptive knowledge graph in step S3 and matching bird species and habitat information in combination with the localized migratory bird knowledge base in step S1 are as follows: Perform multimodal analysis, including: Querying text entered by the user To perform semantic disambiguation, we use the semantic similarity calculation based on the Bert model: ， in, represents the similarity between the query text and migratory bird species, is the Bert semantic vector of the query text, is the semantic vector of migratory bird species, is the cosine similarity function, Perform audio feature matching on user voice input: ， in, represents the dynamic time warping (DTW) similarity between the speech input and the migratory bird sound, is the feature sequence of the user's voice input, is the characteristic sequence of sounds of migratory bird species, Perform object detection and classification on bird photos uploaded by users: ， in, For the predicted bird species, is the classification probability of the image recognition model ResNet, For user-uploaded bird images, Based on the recognition results of multimodal analysis, the species and habitats in the migratory bird knowledge base are matched: ， in, represents the matching migratory bird habitat, Knowledge-based Species matching function.

8. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 7, characterized in that: In step S5, the Bayesian network and time series analysis are combined to dynamically predict the migratory patterns of migratory birds, infer the factors that affect the distribution of migratory birds in a specific area, and optimize the question-answering reasoning process by combining historical data with real-time data.

9. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 8, characterized in that: The step of dynamically predicting and reasoning the migratory pattern of migratory birds using a causal reasoning model based on the multimodal analysis result of step S4 is as follows: The causal inference model that defines the migratory pattern of migratory birds is : ， in, represents a causal inference model, is the set of variables in the causal inference model, including climate factors, food resources and habitat, is the set of causal relationships between variables, is the set of conditional probability distributions in the causal graph, Modeling causal reasoning using Bayesian networks: ， in, Representation variables At its parent node The probability distribution under given conditions, express The joint probability distribution with its parent node, represents the marginal probability of the parent node, Defining migratory bird species In time In position The probability distribution of is: ， in, Indicates time When the bird is in position The predicted probability at is the number of migratory bird populations, is the weight of migratory bird population, Indicates The distribution of migration patterns of migratory bird populations, constrained by a causal inference model, The time series model is used to dynamically predict the migration pattern of migratory birds. The prediction formula is: ， in, Indicates time When the bird is in position The predicted probability at is the time impact weight, For time Probability change rate of migratory birds: .

10. The AI ​​interactive question-and-answer method based on personalized customization of migratory bird knowledge as claimed in claim 9, characterized in that: The steps of dynamically adjusting the knowledge base data structure and question-answer weights in combination with reinforcement learning are: Define the user feedback collection as: ， in, represents the user feedback dataset, is the number of user interactions, For the User queries, The answer returned by the system. Ratings or correction suggestions given by users, Based on user feedback, reinforcement learning is used to optimize the knowledge base weights. The optimization formula is: ， in, are the parameters of the knowledge base data structure, is the reinforcement learning reward function, is the learning rate, is the gradient of the reward function, The optimization goal is: ， in, is the optimized knowledge base weight parameter, Based on parameters Computed question-answer generation probability.

Images