Keyword similarity ranking method and system

Abstract

The invention discloses a keyword similarity ranking method and system, and belongs to the field of information retrieval, and the method comprises the following steps: 1, synchronizing heterogeneous data sources; 2, data preprocessing and conversion, wherein a deep learning model is adopted to conduct automatic preprocessing and conversion on heterogeneous data of different formats; step 3, real-time data synchronization: performing synchronization processing on the real-time data stream by using stream-oriented computation and a real-time data processing framework; step 4, dynamic keyword weight calculation; 5, keyword similarity ranking is carried out, semantic understanding and similarity calculation are carried out on the keywords through a large model, and optimized keyword ranking is generated in combination with dynamically calculated keyword weights; and 6, performing personalized ranking and recommendation. According to the method, the performance is excellent, the expansibility is high, flexible weight adjustment can be carried out when dynamic data are processed, and then the accuracy of a retrieval result is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of information retrieval, and specifically to a keyword similarity ranking method and system. Background Art

[0002] With the rapid development of information technology, enterprises and organizations have accumulated a large amount of heterogeneous data, including structured data (such as table data in relational databases), semi-structured data (such as JSON files or log data), and unstructured data (such as text, pictures, and videos). These data are scattered and stored in different systems, independent of each other, lacking unified management and integration, which seriously affects the value mining and utilization efficiency of the data.

[0003] Existing data synchronization tools often face performance bottlenecks and insufficient scalability when dealing with heterogeneous data sources. At the same time, keyword similarity ranking algorithms lack a flexible weight adjustment mechanism when dealing with dynamic data, resulting in the accuracy and relevance of retrieval results being affected.

[0004] Therefore, those skilled in the art have provided a keyword similarity ranking method and system to solve the problems raised in the above background art. Summary of the Invention

[0005] The purpose of the present invention is to provide a keyword similarity ranking method and system, which has excellent performance and high scalability, and can perform flexible weight adjustment when dealing with dynamic data, thereby effectively improving the accuracy of retrieval results, so as to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] A keyword similarity ranking method includes the following steps:

[0008] Step 1: Synchronization of heterogeneous data sources. Through an intelligent data synchronization framework based on a large model, automatically identify and synchronize heterogeneous data sources from different types, formats, and structures.

[0009] Step 2: Data preprocessing and conversion. Use a deep learning model to automatically preprocess and convert heterogeneous data in different formats, including data cleaning, standardization, and structured conversion, to meet the format requirements of the target data platform.

[0010] Step 3: Real-time data synchronization. Use stream computing and a real-time data processing framework to synchronize real-time data streams, and dynamically adjust the synchronization strategy according to the changes in the data streams to ensure the real-time and consistency of the data.

[0011] Step Four: Dynamic keyword weight calculation. Through a deep learning model, dynamically adjust the weights of keywords based on user behavior data, historical query records, and real-time context, and calculate the weights of each keyword in real time according to actual requirements;

[0012] Step Five: Keyword similarity ranking. Use a large model to perform semantic understanding and similarity calculation on keywords, and combine the dynamically calculated keyword weights to generate an optimized keyword ranking;

[0013] Step Six: Personalized ranking and recommendation. According to real-time information on user behavior, interests, and location, dynamically adjust keyword weights and ranking strategies through a reinforcement learning model to achieve personalized keyword recommendations and search rankings.

[0014] As a further solution of the present invention: In the said Step One, the specific process of heterogeneous data source synchronization is as follows:

[0015] S101: Data source type identification and classification. Automatically identify different data source types through a deep learning model. The data source types include structured data, semi-structured data, and unstructured data, and select appropriate synchronization strategies according to the data types;

[0016] S102: Cross-modal data synchronization. Through multi-modal learning algorithms, achieve the synchronization of data from different modal data sources, and ensure efficient integration and conversion between multiple data types.

[0017] As a further solution of the present invention: In the said Step Three, the specific process of real-time data synchronization is as follows:

[0018] S201: Real-time monitoring and processing of data streams. Monitor the real-time changes of data sources through a streaming computing framework, timely obtain newly added or updated data, and automatically synchronize it to the target system;

[0019] S202: Dynamic synchronization strategy adjustment. Dynamically adjust the synchronization strategy according to the load of the data stream, synchronization latency, and real-time query requirements to ensure the efficiency and real-time performance of the system.

[0020] As a further solution of the present invention: In the said Step Four, the specific process of dynamic keyword weight calculation is as follows:

[0021] S301: Data analysis based on user behavior. Analyze historical data on user search behavior, click-through rate, and dwell time, and use a deep learning model to dynamically calculate the weights of keywords related to user needs;

[0022] S302: Combination of context information. Combine the current user query context, geographical location, and device type information to adjust the weights of keywords in real time, thereby optimizing search results and recommendation effects;

[0023] S303: Response to real-time environmental changes. By combining external environmental changes, dynamically adjust the weights of keywords to make the ranking more in line with real-time user needs and scenarios.

[0024] As a further solution of the present invention: In the fifth step, the specific process of keyword similarity ranking is as follows:

[0025] S401: Similarity calculation based on deep semantic understanding. Use a large model to perform deep semantic understanding on keywords, identify synonyms, near-synonyms, and related words, and perform similarity calculation at the semantic level;

[0026] S402: Multi-dimensional keyword matching. Combine multi-dimensional data to perform multi-level matching on keywords, thereby optimizing the relevance and accuracy of the ranking.

[0027] As a further solution of the present invention: In the sixth step, the specific process of personalized ranking and recommendation is as follows:

[0028] S501: User behavior analysis and model training. By analyzing the historical behavior data of users, train a personalized recommendation model, and use deep learning and reinforcement learning algorithms to optimize the personalized keyword ranking and recommendation strategy;

[0029] S502: Real-time user information feedback. Based on the user's real-time query information, interaction data, and feedback, dynamically adjust the keyword weights and ranking strategy to achieve a more accurate and personalized search result display.

[0030] This application also discloses a keyword similarity ranking system, which is applied to the keyword similarity ranking method and includes:

[0031] A data synchronization module, used to automatically identify and synchronize heterogeneous data sources from different types, formats, and structures through an intelligent data synchronization framework based on a large model. At the same time, use stream computing and real-time data processing frameworks to synchronize real-time data streams, and dynamically adjust the synchronization strategy according to the changes in the data stream to ensure the real-time and consistency of the data;

[0032] A data preprocessing module, which uses a deep learning model to automatically preprocess and convert heterogeneous data in different formats to ensure that it meets the format requirements of the target platform;

[0033] A keyword ranking calculation module, through a deep learning model, dynamically adjusts the weights of keywords based on user behavior data, historical query records, and real-time context, and calculates the weights of each keyword in real time according to actual needs. At the same time, use a large model to perform semantic understanding and similarity calculation on keywords, and combine the dynamically calculated keyword weights to generate an optimized keyword ranking;

[0034] A personalized recommendation module, which is used to dynamically adjust the keyword weights and ranking strategies according to real-time information of user behavior, interests and locations through a reinforcement learning model, so as to achieve personalized keyword recommendation and search ranking.

[0035] This application also discloses a computer-readable storage medium, on which a program is stored, and when the program is executed by a processor, a keyword similarity ranking method is implemented.

[0036] This application also discloses an electronic device, including a memory and a processor, on which a computer instruction capable of running on the processor is stored, and when the processor runs the computer instruction, a keyword similarity ranking method is executed.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. Efficient heterogeneous data synchronization:

[0039] The present invention combines deep learning and streaming computing technologies, and can automatically identify and synchronize heterogeneous data from different data sources (such as structured data, unstructured data, semi-structured data, etc.). Different from the traditional batch-based data synchronization method, the present invention can achieve real-time data synchronization, greatly reducing data synchronization latency and improving data real-time performance and consistency.

[0040] At the same time, through a streaming computing framework (such as Apache Kafka, Apache Flink, etc.) and intelligent data processing algorithms, it can more flexibly handle the dynamic changes and frequent updates of multi-source heterogeneous data.

[0041] 2. Dynamically adjust keyword weights and rankings:

[0042] Traditional keyword ranking methods mostly rely on static weight calculation, usually based on statistical features such as term frequency (TF) and inverse document frequency (IDF). This method fails to effectively capture the semantic information of keywords.

[0043] The present invention introduces dynamic keyword weight calculation based on deep learning and semantic-level similarity calculation. Through dynamic analysis of user behavior, query context and external environment changes, it can adjust keyword weights in real time, thereby improving the accuracy of ranking and user satisfaction.

[0044] Combined with large models (such as pre-trained language models like LLM, GPT, etc.), it can better understand the deep semantics of keywords and overcome the limitations of traditional statistical methods in identifying polysemous words and synonyms.

[0045] 3. Personalized ranking and recommendation:

[0046] Most traditional recommendation systems rely on static rankings based on content or collaborative filtering, failing to fully consider users' real-time behaviors and context changes.

[0047] The present invention adopts a reinforcement learning algorithm. According to users' real-time feedback (such as click-through rate, dwell time, etc.), it dynamically adjusts the weights of keywords and ranking strategies to achieve personalized recommendations. By training a personalized recommendation model, it can accurately predict and display the most relevant keywords and content, greatly enhancing the user experience.

[0048] 4. Optimize query performance:

[0049] In the traditional query processing process, as the data volume grows, query performance often encounters bottlenecks. Especially when facing complex query conditions, data associations, and aggregation statistics, the query response time will increase significantly.

[0050] The present invention significantly improves query efficiency by adopting deep learning to optimize query plans and strategies of multi-level caching and parallel computing. Combining distributed computing and intelligent query optimization, it can still maintain low query latency and high query throughput under a large volume of data.

[0051] 5. Multi-modal data synchronization and processing:

[0052] The present invention can process heterogeneous data including multiple modalities such as images, videos, texts, time series, etc. Through multi-modal learning algorithms (such as the combination of images and texts), during the data synchronization process, it can make full use of the relevance between multiple data modalities to improve the accuracy and effect of data integration.

[0053] Compared with the traditional single data type processing method, the present invention can more comprehensively understand and synchronize data from different modalities and adapt to data requirements in different scenarios.

[0054] 6. Response to real-time environmental changes:

[0055] The present invention dynamically adjusts keyword weights by combining external environmental changes (such as seasons, hot events, news trends, etc.), enabling search engines and recommendation systems to respond to environmental changes in a timely manner and provide more timely rankings and recommendations.

[0056] This is especially important in fields with rapidly changing information (such as social media, news, e-commerce promotions, etc.), and can effectively improve the real-time performance and competitiveness of the system.

[0057] 7. Multi-level optimization and intelligent decision-making:

[0058] Based on a deep learning model, the present invention can optimize data synchronization and query processing at multiple levels and dimensions, including making intelligent decisions in aspects such as user behavior, real-time queries, and environmental changes. Compared with traditional rule-based optimization methods, the present invention can provide more flexible and accurate optimization strategies, improving the intelligence level of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of a keyword similarity ranking method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0061] As mentioned in the background art of the present application, through research, it is found that existing data synchronization tools often face performance bottlenecks and insufficient scalability when dealing with heterogeneous data sources. At the same time, the keyword similarity ranking algorithm lacks a flexible weight adjustment mechanism when dealing with dynamic data, resulting in the accuracy and relevance of retrieval results being affected, and there are certain defects.

[0062] To solve the above defects, the present application discloses a keyword similarity ranking method and system, which has excellent performance and high scalability, and can perform flexible weight adjustment when dealing with dynamic data, thereby effectively improving the accuracy of retrieval results.

[0063] The following will introduce in detail how the solution of the present application solves the above technical problems in conjunction with the accompanying drawings.

[0064] Please refer to Figure 1, in an embodiment of the present invention, a method for keyword similarity ranking includes the following steps: Step 1: Heterogeneous data source synchronization. Automatically identify and synchronize heterogeneous data sources from different types, formats, and structures through an intelligent data synchronization framework based on a large model, including but not limited to structured data, semi-structured data, and unstructured data. Heterogeneous data sources include databases, log files, sensor data, pictures, videos, etc.; Step 2: Data preprocessing and transformation. Use a deep learning model to automatically preprocess and transform heterogeneous data in different formats, including data cleaning, standardization, and structured transformation to meet the format requirements of the target data platform; Step 3: Real-time data synchronization. Use stream computing and a real-time data processing framework to synchronize real-time data streams and dynamically adjust the synchronization strategy according to changes in the data stream to ensure the real-time and consistency of the data; Step 4: Dynamic keyword weight calculation. Dynamically adjust the weights of keywords based on user behavior data, historical query records, and real-time context through a deep learning model, and calculate the weights of each keyword in real time according to actual needs; Step 5: Keyword similarity ranking. Use a large model (such as pre-trained language models like LLM, GPT, T5, etc.) to perform semantic understanding and similarity calculation on keywords, and combine the dynamically calculated keyword weights to generate an optimized keyword ranking; Step 6: Personalized ranking and recommendation. Dynamically adjust keyword weights and ranking strategies through a reinforcement learning model according to real-time information on user behavior, interests, and locations to achieve personalized keyword recommendations and search rankings. The keyword similarity ranking method of this application has excellent performance and high scalability, and can flexibly adjust weights when processing dynamic data, thereby effectively improving the accuracy of retrieval results. Among them, LLM (Large Language Model) is a large-scale language model constructed by a deep neural network containing hundreds of billions of parameters or more. It is trained using a self-supervised learning method with a large amount of unlabeled text, aiming to capture the grammar, semantics, and common sense of natural language. GPT (Generative Pre-trained Transformer) is a series of pre-trained language models based on the Transformer architecture developed by OpenAI. T5 (Text-to-Text Transfer Transformer) is a pre-trained language model released by Google, which regards all NLP tasks as text-to-text conversion problems.

[0065] In this embodiment, in step one, the specific process of heterogeneous data source synchronization is as follows: S101: Data source type identification and classification. Different data source types are automatically identified through a deep learning model. The data source types include structured data, semi-structured data, and unstructured data, and appropriate synchronization strategies are selected according to the data types; S102: Cross-modal data synchronization. Through multi-modal learning algorithms (such as images, text, time series, etc.), the synchronization of data sources from different modalities is achieved, ensuring efficient integration and conversion among various data types. The deep learning model is a branch of machine learning. It is based on the concept of artificial neural networks (ANNs), but has a deeper architecture. These models attempt to learn and understand complex patterns and features in data by mimicking the working mode of neurons in the human brain. The multi-modal learning algorithm refers to an algorithm that can process and understand information from multiple different modalities (such as images, text, audio, time series, etc.). This learning method helps to capture more context information and improve the accuracy and expressiveness of the model.

[0066] In this embodiment, in step three, the specific process of real-time data synchronization is as follows: S201: Real-time monitoring and processing of data streams. The real-time changes of data sources are monitored through a streaming computing framework, newly added or updated data is obtained in a timely manner, and it is automatically synchronized to the target system; S202: Dynamic adjustment of synchronization strategies. According to the load of data streams, synchronization latency, and real-time query requirements, the synchronization strategy is dynamically adjusted to ensure the efficiency and real-time performance of the system. The streaming computing framework refers to a system that can process and analyze continuously generated data streams in real-time or near real-time. These data streams usually come from various data sources, such as sensors, social media, transaction systems, etc. The main characteristics of the streaming computing framework are real-time performance, high throughput, fault tolerance, and the ability to process unbounded data streams.

[0067] In this embodiment, in step four, the specific process of dynamic keyword weight calculation is as follows: S301: Data analysis based on user behavior. By analyzing historical data of user search behavior, click-through rate, and dwell time, the keyword weights related to user needs are dynamically calculated using a deep learning model; S302: Combination of context information. Combining the current user query context, geographical location, and device type information, the keyword weights are adjusted in real-time to optimize search results and recommendation effects; S303: Response to real-time environmental changes. By combining external environmental changes (such as hot events, seasonal changes, etc.), the keyword weights are dynamically adjusted to make the ranking more in line with real-time user needs and situations. Keyword weight reflects the relevance and importance of keywords in the content indexed by the search engine. It takes into account not only the frequency of keyword occurrences in the web page, but also factors such as the position of the keyword, the overall quality of the page, and the relevance to other web pages.

[0068] In this embodiment, in step five, the specific process of keyword similarity ranking is as follows: S401: Similarity calculation based on deep semantic understanding. Use large models (such as LLM, T5, etc.) to perform deep semantic understanding on keywords, identify synonyms, near-synonyms, and related vocabulary, and perform similarity calculation at the semantic level; S402: Multi-dimensional keyword matching. Combine multi-dimensional data (such as user historical queries, external hotspots, social media trends, etc.) to perform multi-level matching on keywords, so as to optimize the relevance and accuracy of ranking.

[0069] In this embodiment, in step six, the specific process of personalized ranking and recommendation is as follows: S501: User behavior analysis and model training. By analyzing the user's historical behavior data, train a personalized recommendation model, and use deep learning and reinforcement learning algorithms to optimize the personalized keyword ranking and recommendation strategy; S502: Real-time user information feedback. Based on the user's real-time query information, interaction data, and feedback, dynamically adjust the keyword weights and ranking strategies to achieve a more accurate and personalized search result display.

[0070] This application also discloses a keyword similarity ranking system, which is applied to the keyword similarity ranking method, including: a data synchronization module, used to automatically identify and synchronize heterogeneous data sources from different types, formats, and structures through an intelligent data synchronization framework based on a large model. At the same time, use stream computing and real-time data processing frameworks to synchronize real-time data streams, and dynamically adjust the synchronization strategy according to the changes in the data stream to ensure the real-time and consistency of the data; a data preprocessing module, which uses a deep learning model to automatically preprocess and transform heterogeneous data in different formats to ensure that it meets the format requirements of the target platform; a keyword ranking calculation module, which dynamically adjusts the weights of keywords based on user behavior data, historical query records, and real-time context through a deep learning model, and calculates the weights of each keyword in real time according to actual needs. At the same time, use a large model to perform semantic understanding and similarity calculation on keywords, and combine the dynamically calculated keyword weights to generate an optimized keyword ranking; a personalized recommendation module, used to dynamically adjust the keyword weights and ranking strategies according to the user's behavior, interests, and real-time location information through a reinforcement learning model to achieve personalized keyword recommendation and search ranking.

[0071] In specific implementation, the specific usage process of the keyword similarity ranking system is as follows:

[0072] (1): Collect user behavior data

[0073] User behavior data is an important basis for dynamically adjusting keyword weights. The collected data usually includes:

[0074] Click-Through Rate (CTR): The frequency at which users click on a certain keyword or link.

[0075] Bounce Rate: The proportion of users who leave immediately after accessing a page, reflecting the attractiveness of the keyword.

[0076] Dwell Time: The time users spend on a certain page, indirectly reflecting the relevance of the page.

[0077] Conversion Rate: Whether users have completed a certain target behavior (such as purchase, registration, etc.).

[0078] Frequency of user input keywords: Keywords frequently queried by users, helping to determine which keywords are of most concern to current users.

[0079] Collect user click behavior data using server-side logs. Store and organize these behavior data for subsequent analysis.

[0080] (2): Collect query log data

[0081] Query logs record the keywords entered by users in the search engine and the relevant search results. These logs can reflect the keywords preferred by users, search patterns, and behaviors.

[0082] Record the keywords queried by users, timestamps, query results, etc.

[0083] Store the log data for subsequent analysis and weight calculation.

[0084] (3): Analyze user behavior data

[0085] Through the analysis of user behavior data, relevant indicators such as click-through rate, bounce rate, and dwell time can be calculated for each keyword. These indicators reflect the relevance of the keyword and user preference for it.

[0086] Calculate the CTR (Click-Through Rate) for each keyword, that is, the number of times a certain keyword is clicked divided by the total number of times the keyword is displayed.

[0087] Calculate the bounce rate and dwell time, and analyze from user behavior which keywords lead to a higher bounce rate or a longer dwell time.

[0088] Calculate the conversion rate, that is, how many users have completed the expected behavior (such as purchase, registration, etc.) after clicking on a certain keyword.

[0089] (4): Calculate the preliminary weight of keywords

[0090] Based on the analysis results of user behavior, assign a preliminary weight to each keyword. These weights can be weighted and calculated according to different metrics. Common calculation methods include:

[0091] CTR weighting: Keywords with a high click-through rate are assigned higher weights.

[0092] Bounce rate weighting: Keywords with a low bounce rate are assigned higher weights.

[0093] Dwell time weighting: Keywords with a long dwell time are assigned higher weights.

[0094] Conversion rate weighting: Keywords with a high conversion rate are assigned higher weights.

[0095] Determine the weights of each metric (such as CTR, bounce rate, dwell time, etc.).

[0096] Calculate the comprehensive weight of each keyword according to the formula:

[0097] Weight = w1 * CTR + w2 * (1 - bounce rate) + w3 * dwell time + w4 * conversion rate

[0098] Among them, w1, w2, w3, and w4 are weight coefficients, which can be adjusted according to actual needs.

[0099] (5): Combine the query log and analyze the keyword frequency

[0100] The keyword frequency in the query log is an important basis for dynamically adjusting keyword weights. Frequent occurrence of a keyword indicates a high current demand for that keyword, and a higher weight can be assigned.

[0101] Calculate the occurrence frequency of each keyword in the query log.

[0102] Keywords with a higher frequency can be appropriately increased in weight.

[0103] For example, assume a keyword has appeared 100 times in the past 24 hours, and the weight can be appropriately increased according to the occurrence frequency.

[0104] (6): Dynamically adjust keyword weights

[0105] Combine the user behavior analysis results with the keywords with a higher frequency in the query log to dynamically adjust the keyword weights. The weights will change continuously according to actual user behavior and query data.

[0106] Adjust the keyword weights according to the user behavior analysis results (for example, increase the weights of keywords with a high click-through rate and a low bounce rate).

[0107] Adjust the weights of keywords according to their frequencies in the query logs (e.g., increase the weights of keywords that have been frequently queried recently).

[0108] Update the adjusted weights to the system to ensure real-time ranking updates.

[0109] (7): Regularly update and optimize the weight model

[0110] To maintain the flexibility and accuracy of the system, regularly update and optimize the weight model of keywords. A certain period (such as daily or weekly) can be set for recalculation and adjustment.

[0111] Regularly collect new user behavior data and query logs.

[0112] Recalculate the weights of each keyword based on the new data.

[0113] Optimize the weight calculation model to ensure that the system can handle changes in user behavior.

[0114] (8): Test and verify

[0115] Test the keyword rankings after dynamic adjustment to ensure that the adjusted weight model meets expectations and can improve the relevance of search results and user experience.

[0116] This application also discloses a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements the keyword similarity ranking method.

[0117] This application also discloses an electronic device, including a memory and a processor. A computer instruction capable of running on the processor is stored on the memory. When the processor runs the computer instruction, it executes the keyword similarity ranking method.

[0118] The present invention realizes the efficient synchronization and integration of heterogeneous data, and at the same time optimizes the keyword similarity ranking algorithm by combining dynamic weight adjustment, significantly improving the accuracy of retrieval results and the real-time response ability of the system. The system has broad application prospects in the fields of multi-source data integration and intelligent retrieval.

[0119] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

[0120] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A keyword similarity ranking method, characterized in that It includes the following steps: Step 1: Heterogeneous data source synchronization. Through an intelligent data synchronization framework based on a large model, automatically identify and synchronize heterogeneous data sources from different types, formats, and structures; Step 2: Data preprocessing and transformation. Use a deep learning model to automatically preprocess and transform heterogeneous data in different formats, including data cleaning, standardization, and structured transformation to meet the format requirements of the target data platform; Step 3: Real-time data synchronization. Utilize stream computing and a real-time data processing framework to synchronize real-time data streams, and dynamically adjust the synchronization strategy according to the changes in the data stream to ensure the real-time and consistency of the data; Step 4: Dynamic keyword weight calculation. Through a deep learning model, dynamically adjust the weights of keywords based on user behavior data, historical query records, and real-time context, and calculate the weights of each keyword in real time according to actual needs; Step 5: Keyword similarity ranking. Use a large model to perform semantic understanding and similarity calculation on keywords, and generate an optimized keyword ranking by combining the dynamically calculated keyword weights; Step 6: Personalized ranking and recommendation. According to real-time information on user behavior, interests, and location, dynamically adjust keyword weights and ranking strategies through a reinforcement learning model to achieve personalized keyword recommendations and search rankings.

2. The keyword similarity ranking method according to claim 1, characterized in that, In Step 1, the specific process of heterogeneous data source synchronization is as follows: S101: Data source type identification and classification. Automatically identify different data source types through a deep learning model. The data source types include structured data, semi-structured data, and unstructured data, and select an appropriate synchronization strategy according to the data type; S102: Cross-modal data synchronization. Through multi-modal learning algorithms, achieve the synchronization of data sources from different modalities, and ensure efficient integration and transformation between various data types.

3. A method for ranking keyword similarity according to claim 2, characterized in that, In Step 3, the specific process of real-time data synchronization is as follows: S201: Real-time monitoring and processing of data streams. Monitor the real-time changes of data sources through a stream computing framework, timely obtain newly added or updated data, and automatically synchronize it to the target system; S202: Dynamic synchronization strategy adjustment. Dynamically adjust the synchronization strategy according to the load of the data stream, synchronization latency, and real-time query requirements to ensure the efficiency and real-time of the system.

4. A method for ranking keyword similarity according to claim 3, characterized in that In Step 4, the specific process of dynamic keyword weight calculation is as follows: S301: Data analysis based on user behavior. Analyze historical data on user search behavior, click-through rate, and stay time, and use a deep learning model to dynamically calculate the keyword weights related to user needs; S302: Incorporation of context information. Combine the current user query context, geographical location, and device type information to adjust the keyword weights in real time, thereby optimizing search results and recommendation effects; S303: Response to real-time environmental changes. Dynamically adjust the keyword weights by combining external environmental changes, making the ranking more in line with real-time user needs and situations.

5. A method for ranking keyword similarity according to claim 4, characterized in that In Step 5, the specific process of keyword similarity ranking is as follows: S401: Similarity calculation based on deep semantic understanding. Use a large model to perform deep semantic understanding on keywords, identify synonyms, near-synonyms, and related words, and perform similarity calculation at the semantic level; S402: Multi-dimensional keyword matching. Combine multi-dimensional data to perform multi-level matching on keywords, thereby optimizing the relevance and accuracy of rankings.

6. A method for ranking keyword similarity according to claim 5, characterized in that In the sixth step, the specific process of personalized ranking and recommendation is as follows: S501: User behavior analysis and model training. By analyzing the historical behavior data of users, train a personalized recommendation model, and use deep learning and reinforcement learning algorithms to optimize personalized keyword rankings and recommendation strategies; S502: Real-time user information feedback. Based on the user's real-time query information, interaction data, and feedback, dynamically adjust the keyword weights and ranking strategies to achieve a more accurate and personalized search result display.

7. A keyword similarity ranking system, characterized in that, Applied to the keyword similarity ranking method described in any one of claims 1-6, including: A data synchronization module, which is used to automatically identify and synchronize heterogeneous data sources from different types, formats, and structures through an intelligent data synchronization framework based on a large model. At the same time, use stream computing and real-time data processing frameworks to synchronize real-time data streams, and dynamically adjust the synchronization strategy according to the changes in the data stream to ensure the real-time and consistency of the data; A data preprocessing module, which uses a deep learning model to automatically preprocess and transform heterogeneous data in different formats to ensure that it meets the format requirements of the target platform; A keyword ranking calculation module, which dynamically adjusts the weights of keywords based on user behavior data, historical query records, and real-time context through a deep learning model, and calculates the weights of each keyword in real time according to actual needs. At the same time, use a large model to perform semantic understanding and similarity calculation on keywords, and combine the dynamically calculated keyword weights to generate an optimized keyword ranking; A personalized recommendation module, which is used to dynamically adjust the keyword weights and ranking strategies through a reinforcement learning model according to the real-time information of user behavior, interests, and locations to achieve personalized keyword recommendation and search ranking.

8. A computer-readable storage medium, characterized in that, A program is stored on the computer-readable storage medium, and when the program is executed by a processor, it implements the keyword similarity ranking method described in any one of claims 1-6.

9. An electronic device, characterized in that, It includes a memory and a processor. A computer instruction capable of running on the processor is stored on the memory, and when the processor runs the computer instruction, it executes the keyword similarity ranking method described in any one of claims 1-6.

Images