A Review and Design of Semantic-Level Feature Spatial Representation and Resource Spatiotemporal Mapping for Socialized Service Resources in Rural Characteristic Industries

Abstract

Socialized services for rural characteristic industries are becoming a key support for promoting rural industries’ transformation and upgrading. They are permeating the development process of modern agricultural service technologies, achieving significant progress in specialized fields such as mechanized operations and plant-protection services. However, challenges remain, including low efficiency in matching service resources and limited spatiotemporal coordination capabilities. With the deep integration of spatiotemporal information technology and knowledge graph technology, the enormous potential of semantic-level feature spatial representation in intelligent scheduling of service resources has been fully demonstrated, providing a new technical pathway to solve the above problem. This paper systematically analyzes the technological evolution trends of socialized services for rural characteristic industries and proposes a collaborative scheduling framework based on semantic feature space and spatiotemporal maps for characteristic industry service resources. At the technical architecture level, the paper aims to construct a spatiotemporal graph model integrating geographic knowledge graphs and temporal tree technology to achieve semantic-level feature matching between service demand and supply. Regarding implementation pathways, the model significantly improves the spatiotemporal allocation efficiency of service resources through cloud service platforms that integrate spatial semantic matching algorithms and dynamic optimization technologies. This paper conducts in-depth discussions and analyses on technical details such as agricultural semantic feature extraction, dynamic updates of rural service resources, and the collaboration of semantic matching and spatio-temporal matching of supply and demand relationships. It also presents relevant implementation methods to enhance technical integrity and logic, which is conducive to the engineering implementation of the proposed methods. The effectiveness of the proposed collaborative scheduling framework for service resources is proved by the synthesis of principal analysis, logical deduction and case comparison. We have proposed a practical “three-step” implementation path conducive to realizing the proposed method. Regarding application paradigms, this technical system will promote the transformation of rural industry services from traditional mechanical operations to an intelligent service model of “demand perception–intelligent matching–precise scheduling”. In the field of socialized services for rural characteristic industries, it is suggested that relevant institutions promote this technical framework and pay attention to the development trends of new technologies such as knowledge services, spatio-temporal services, the Internet of Things, and unmanned farms so as to promote the sustainable development of rural characteristic industries.

1. Introduction

1.1. Background and Significance

Deeply exploring and utilizing regional resource endowments to develop rural characteristic industries is an essential pathway for developing regional economies and promoting rural revitalization [1]. Socialized services constitute a service system that supports agricultural production through specialized, standardized, and market-oriented means, specifically mechanical operations, intelligent seedling cultivation, plant protection services, agricultural material distribution, production trusteeship, technical training, and brand marketing. As an intensive production service model that integrates labor-intensive and service-intensive segments, it represents an inevitable trend in promoting rural industrial development against the backdrop of accelerating urbanization and rural labor shortages. It is an effective model for activating the rural economy and achieving value co-creation. In terms of labor factors, it wholly or partially replaces the labor of aging farmers; in terms of land factors, it promotes farmland quality improvement through pesticide reduction and soil fertility enhancement; in terms of technological aspects, it promotes scientific and technological progress through improved varieties and modern technologies; in terms of capital factors, it promotes increased capital investment through the increase in social capital and financial capital, ultimately comprehensively promoting the improvement in food production capacity [2]. The reform of rural land “separation of three rights” has laid the policy foundation for moderate-scale operations and socialized services [3,4]. Currently, socialized services are relatively mature in urban–rural suburban areas in the form of modern agricultural parks and agricultural cooperatives. However, in rural areas, due to remote geographical locations, weak information infrastructure, and scattered service resources, the coverage and efficiency of socialized services remain low. Many socialized services rely on traditional telephone calls and personal networks for information transmission and service docking. This approach has problems such as information asymmetry, slow service response speed, and difficulty ensuring service efficiency. For example, the Qiannan region (China) is in mountainous areas with scattered farmland. During production peaks, local agricultural machinery service entities cannot meet the needs of all regional users. The supply and demand of socialized services cannot be effectively matched, requiring cross-regional operations from entities in Guangxi, Yunnan, and other regions to ensure that overall production is not affected. Under such circumstances, traditional contact methods cannot achieve rapid and efficient service resource scheduling, resulting in low supply–demand matching efficiency and constraining the overall development of socialized services in rural areas. Therefore, to promote the high-quality development of rural characteristic industries, it is first necessary to overcome the problem of farmers being unable to access sufficient socialized services due to poor information flow [5]. In this regard, using information technology to match supply and demand resources precisely is a critical approach and development trend [6].

1.2. Objective and Contribution

The overall objective of this study is to propose a collaborative scheduling framework for socialized service resources based on a semantic feature space and a spatio-temporal graph to address the issue of low efficiency in supply and demand matching. In light of the current situation of agricultural resource services, it can be refined into two sub-goals: addressing the matching error caused by the ambiguity of farmers’ demand semantics and breaking through the scheduling lag problem brought about by the spatio-temporal dynamics of rural service resources.

Compared with the existing methods, the main contribution of this paper is that it applies semantic-level feature space expression to the field of agricultural socialized services for the first time, which can significantly improve the matching accuracy.

1.3. Scope and Limitation

By defining the scope of data, time-space, and application scenarios, the controversy over the generalization of research can be avoided. It mainly includes:

(A)

Data scope: Only structured (GPS trajectories, service orders) and unstructured data (agricultural question-and-answer texts) are processed, and remote sensing images or multimodal data are not involved for the time being.

(B)

Spatio-temporal scope: The time covers the growing season from 2020 to 2023, and the space is concentrated in the rice-growing areas in southern China.

(C)

Application scenarios: Prioritize solving semantic matching problems in agricultural question answering, and do not expand to complex scenarios such as agricultural machinery scheduling for the time being.

1.4. Organization Structure

In the subsequent sections, Section 2 introduces the semantic expression and spatio-temporal technical framework of characteristic industry service resources, Section 3 presents the spatial expression of semantic-level features, Section 4 covers spatio-temporal semantic matching, Section 5 introduces the construction of spatio-temporal resource map services, Section 6 presents the construction of a spatio-temporal platform for industrial socialized services oriented towards rural characteristics, and Section 7 presents the conclusions and prospects.

2. Semantic Expression and Spatiotemporal Technology Framework for Characteristic Industry Service Resources

In research related to precise matching and scheduling of socialized service resources for rural characteristic industries, key technologies such as semantic analysis, knowledge graph construction, and intelligent service platforms have been applied [7,8], mainly reflected in multi-dimensional expression of resources, multi-path semantic matching, precise recommendation, and efficient scheduling (

Table 1. Related Research on precise matching and scheduling of socialized service resources for rural characteristic industries.

Matching Method	Year	Technical Name	Research Object	Effect	Reference
Multi-semantic Features	2022	TF-IDF, Bidirectional Gated Recurrent Neural Network, Multi-granularity Convolutional Neural Network, BERT	Semantic feature and rich semantic feature extraction from short-text Q&A data in the agricultural technology Q&A community of “China Agricultural Technology Extension Information Platform”	The model integrating bidirectional GRU and multi-granularity CNN achieved 95.9% accuracy in agricultural Q&A classification, 94.15% in semantic matching, and 92.07% in pest and disease entity recognition, demonstrating excellent performance	[7]
Keywords	2024	CiteSpace	Research hotspots and development trends of agricultural socialized services	The quantity of research outcomes on agricultural socialized services has maintained a high level over the past 5 years, with related research focusing on five major areas: modern agriculture, food security, rural revitalization, farmers’ needs, and transaction costs	[8]
Graph Neural Network	2021	Knowledge Graph	Resource Recommendation	Proposed a knowledge graph-driven multi-layer graph neural network (KGLN) model, optimizing node feature fusion and aggregation. AUC values improved by 0.3–5.9% and 1.1–8.2% on MovieLens-1M and BookCrossing datasets, respectively	[9]
Graph Neural Network	2023	Knowledge Graph	Rice Fertilization Recommendation	Proposed a model combining knowledge graph reasoning and case-based reasoning to obtain qualitative and quantitative fertilizer amounts. Prediction accuracy for nitrogen, phosphorus, and potassium fertilizer application rates and nitrogen fertilizer operation ratios reached 92.85%, 82.61%, 79.17%, and 90.92%, respectively	[11]
Graph Neural Network	2022	Spatiotemporal Information Fusion	Crop Yield Prediction	Proposed a GNN-RNN model integrating geospatial and temporal knowledge. In corn and soybean yield prediction, compared to the CNN-RNN model, R2 improved by an average of 10.44% and RMSE decreased by an average of 9.6%	[13]
Graph Database Technology	2022	Spatiotemporal Map	Digital Scheduling of Rural Resources	Constructed a Neo4j-based characteristic resource knowledge graph combined with spatiotemporal maps, achieving spatiotemporal information visualization scheduling of characteristic cultural resources through interactive maps and timeline approaches	[14]
Graph Database Technology	2022	Geographic Knowledge Graph	Rural Revitalization Analysis, Ontology Model Design	Through geographic knowledge graph construction including knowledge extraction and entity alignment, analyzed community division results and explored implementation characteristics and potential connections of poverty alleviation projects	[16]

) [9]. In addition to the abovementioned aspects, related research also focuses on data-driven intelligent decision-making [10,11,12,13,14,15,16,17]. The main trends are toward deepening multi-dimensional resource expression and semantic analysis, intelligent resource scheduling, personalized and precise service models, and the development of multi-technology integrated ecosystems.

The precise matching and efficient scheduling of socialized service resources for rural characteristic industries represent critical technical challenges for rural revitalization and characteristic industry development. It can be seen that traditional methods only model from a single dimension of time or space, and the knowledge they can obtain has limitations, which affects the effectiveness of demand matching and service scheduling. To solve the supply–demand contradictions caused by poor information flow, this paper proposes a technical framework centered on semantic-level feature spatial expression and resource spatiotemporal information, as shown in Figure 1. As can be seen from this figure, this framework consists of two layers: Related Technologies and Application. The former consists of a data resources layer and a semantic feature representation and processing layer. The data resources layer provides various resources required for socialized services, including characteristic industry data, service provider registry, demand-side profiling data, contracted service data, etc. The semantic feature representation and processing layer provides semantic analysis technologies required for socialized services, including semantic feature extraction, geographic knowledge graph, spatiotemporal map construction, spatial database, and interfaces between different modules. The latter is the relevant application program for social services, including industrial knowledge services, spatiotemporal coupling recommendation, intelligent matching, etc. The process of using this framework for social services is as follows: First of all, data is collected from the designated data source. Then, it is executed in the order of “semantic feature extraction- > geographic knowledge graph- > spatio-temporal map construction- > spatio-temporal semantic matching- > service module invocation”. Finally, the result of service execution is returned.

From the above analysis, it can be seen that this framework aims to improve resource scheduling and matching efficiency through precise semantic analysis, feature annotation, and spatial mapping of socialized service resources. Essentially, this dynamic spatio-temporal modeling strategy can fully leverage the complementary advantages of multi-modal spatio-temporal data fusion modeling, facilitating precise matching of farmers’ demands and efficient scheduling of service resources. It is an important approach to achieving the grand goal of rural revitalization.

3. Semantic-Level Feature Spatial Expression

Ontology provides semantic networks and logical structures for semantics, helping semantic processing understand relationships between vocabularies. Xue Xiaojuan et al. applied ontology technology to the organization and sharing of village-level data resources, solving the problem of multi-source heterogeneous village-level data resource sharing [17], which can quickly and effectively realize cross-data queries, achieve knowledge utilization, and deeply explore application value.

3.1. Theoretical Origins of Semantic-Level Feature Spatial Expression

From existing research, it is not difficult to find that knowledge graphs play an important role in semantic expression. As a semantic network that reveals relationships between entities, it integrates multi-source heterogeneous data, organizes fragmented knowledge into structured graphs, and intuitively presents concepts, entities, and their interrelationships in the form of graphs. In recent years, it has been widely applied in many fields, including natural sciences and social sciences, promoting the development of intelligent information services and decision-making [18]. From the “expert systems” that emerged in the 1970s, to the use of ontology from the philosophical field to create computer models in the mid-to-late 1970s, to the “semantic web” and “linked data” proposed by Tim Berners-Lee, the father of the World Wide Web, all were predecessors of “knowledge graphs.” The semantic web aims to enable information on the internet to be understood and processed by machines, thereby achieving more intelligent information retrieval and exchange, which is also the essential connotation of semantics [19].

The core technologies of the semantic web include Resource Description Framework (RDF), Web Ontology Language (OWL), and Simple Knowledge Organization System (SKOS). These technologies provide the foundation for implementing semantic-level feature spatial expression. RDF describes resources and their relationships through standardized triple structures, enabling machines to understand and process this information [20].

3.2. Semantic Feature Extraction

Semantic feature extraction involves identifying and extracting meaningful information from text or data to support deeper semantic analysis and understanding. In modern information processing and natural language processing fields, semantic feature extraction aims to extract key semantic information from large amounts of data, which can be used to construct knowledge graphs or improve information retrieval effectiveness. Extraction methods are typically based on word vector models such as Word2vec to construct semantic representation vectors [21,22,23].

Deep learning technology has driven the rapid development of natural language processing technology [24]. By using neural network models, especially models based on the Transformer architecture, it is possible to better capture long-distance dependencies and complex semantic relationships. These models can learn from large-scale corpora while extracting implicit features, significantly improving the accuracy of semantic extraction [25,26].

To enhance the accuracy of semantic feature extraction, this paper proposes a “Three-step Method for Agricultural Semantic Feature Extraction”, which consists of three important steps:

(1)

Service type identification: The BiLSTM + CRF model is adopted. The farmers’ demand text (such as “drone spraying”) is input, and the service type label (such as “plant protection service”) is output. The model is trained on a self-built agricultural service corpus (containing 50,000 labeled data entries).

(2)

Keyword vectorization: For the identified keywords (such as “drone” and “pesticide spraying”), Word2vec is used to train 128-dimensional vectors. The pre-trained word vectors are derived from the abstracts of agricultural literature on the China National Knowledge Infrastructure (CNKI) (approximately 2 million).

(3)

Semantic amplification: By integrating agricultural knowledge graphs (such as “rice pests and diseases → Need for plant protection services → Recommended drones/sprayers”), implicit demands are supplemented through rule reasoning (for example, when farmers input “rice planthopper”, the system automatically associates “pyaphil agent + drone spraying”). After amplification, the coverage rate of semantic features is significantly improved.

3.3. Geographic Knowledge Graph

Knowledge graphs provide a new analytical framework in rural spatial analysis by integrating geospatial and temporal attributes, effectively processing multi-source heterogeneous data, and supporting dynamic expression and analysis of rural spaces. This framework can not only replace traditional spatial analysis methods but also combine with geographic grid technology to improve the efficiency of spatial analysis, thereby better understanding the changes and evolution processes of rural spaces [27].

Geographic knowledge graph, driven by the rapid development of artificial intelligence and big data technology, has gradually become an essential component of geographic information systems and intelligent applications. It can be constructed using ontology modeling tools such as Protégé, with administrative divisions, railways, and transportation-related geographic data as primary data sources. It uses OWL as the ontology description language to promote the transformation from geographic information to geographic knowledge services [28].

Geographic knowledge graphs support precise management and decision-making in the agricultural industry by integrating multi-source data [29]. In digital rural construction, IoT technology and data platform architecture have been applied to achieve intelligent perception and integrated management of rural ecological environments and agricultural production operations, enhancing agrarian production’s refinement and intelligence level [30].

3.3.1. Geospatial Data Characteristics

As a medium carrying spatial information, geospatial data has unified mathematical specifications and extensive geographic coverage [31]. It abstracts the spatial characteristics of features through geometric shapes such as points, lines, and surfaces. Its main characteristics include hierarchical and block storage structures, which may reduce the interconnections between entities [31]; through mathematical transformation, geospatial information can present three-dimensional geographic entities on a two-dimensional plane while maintaining their topological connections [31]. Although simplifying actual planar geographic features into geometric shapes can depict spatial layouts, their geographic meaning expression is relatively limited [31]. In addition, the cognition of geographic space has hierarchical characteristics, and cross-level relationships can be extracted through spatial relationship operators. Meanwhile, knowledge acquisition and fusion improve storage and query efficiency through graph and spatial databases [31].

3.3.2. Attribute Data Characteristics

The attributes of geospatial data are often singular and need to be combined with information from other dimensions. Liu Junnan et al. selected Baidu Baike, which has rich data sources, to obtain attribute data. This data source has independent entity pages, standardized concept labels, and high-quality text. It is constructed by experts and updated frequently, can effectively complete missing attributes of geographic entities, and provides richer Chinese information, making it suitable as the primary data source for information extraction [31].

3.3.3. Geocoding

Geocoding converts text descriptions (such as place names and addresses) into machine-readable coordinate systems [32], whose processing flow is shown in Figure 2. It can be seen that Geocoding mainly includes multiple steps such as the construction of the address space semantic model, address standardization, and address matching. Geocoding results can be directly used to construct geographic knowledge graphs, enabling geographic entities in the graph to be precisely located through their coordinates.

3.3.4. Geographic Knowledge Graph Construction Methods

Geographic knowledge graph is a knowledge representation framework that organizes geographic entities and their relationships using a graph structure (node-edge), integrates geographic spatial data, attribute data, and semantic relationships, and supports reasoning and decision-making. It is the integration of geospatial and attribute data, elevating spatial data and attribute data into a knowledge network through semantic relations. Its overall architecture is shown in Figure 3. In research on geographic knowledge graph construction methods, multiple aspects are typically considered, including conceptual system organization, data layer processing, relationship construction, and storage method selection. Conceptual system organization aims to construct a clear and reasonable conceptual framework for geographic entities; data layer processing focuses on how to extract and refine geospatial knowledge from raw data effectively; relationship construction is a key step in clarifying various connections between geographic entities; storage method selection must balance data storage and retrieval efficiency.

The open-source Geos library, written in C++, can implement geometric operations such as the OGC (Open Geospatial Consortium) Simple Features Specification for SQL, and is widely applied in the geospatial data processing field. Its core functions include geometric object operations, topological relationship checking, reverse geocoding, and distance and area calculations. The topological relationship checking and geometric object operation functions can be utilized when processing entities with the same name but different locations. By converting the geometric information of same-named entities in other places (such as points formed by latitude and longitude coordinates, surfaces formed by regional boundaries, etc.) into geometric objects that the Geos library can process, the topological relationship checking function can be used to determine spatial relationships between these geometric objects, such as whether there is overlap, intersection, containment, etc. Suppose the geometric objects of two entities with the same name are completely unrelated spatially (i.e., no topological relationship). In that case, they can be determined to be different entities, thereby achieving disambiguation of same-named entities in other locations. Meanwhile, with the Geos library’s powerful geometric object operation capabilities, such as merging, splitting, and intersection operations, complex geospatial data can be processed and analyzed, providing support for more accurate determination of entity identity.

Regarding geographic knowledge graph construction, Liu Junnan et al. constructed geographic knowledge graphs using four parts: schema layer, data layer, data storage, and computational applications. The schema layer adopts a “top-down and bottom-up” approach to organize the conceptual system of geospatial data. The expansion of spatial objects establishes geographic entity concepts (GeoEntity) and their attribute relationships, deriving spatial data geographic entities (GeoDatasetEntity) and Baidu Baike geographic entities (GeoBaikeEntity). Entities are identified through features, and constructs have feature relationships with spatial data geographic entities. Geometry is used to represent the geometric appearance of entities [31], using WKT and EPSG character strings to identify coordinate information.

Additionally, spatial relationships are constructed through SpatialRelation, deriving three types of sub-relationships: topological, directional, and distance, to clarify the proximity and associations between geographic entities. The data layer automatically extracts geospatial knowledge from raw data. It refines semantic information, including knowledge acquisition (extracting concepts, entities, relationships, and supplementing Baidu Baike attributes) and knowledge fusion (identifying same-named geographic entities and achieving attribute alignment). For same-named entities in different locations, the Geos library processes sameAs relationships. To improve retrieval efficiency, relational databases store data and semi-structured semantic information. In contrast, graph databases are used to store semantic relationships, constructing four tables: “GeoEntity”, “GeoField_Baike”, “re_Geo_Geo”, and “GeoRelation” [31].

In response to the frequent changes in rural service resources (such as cross-regional agricultural machinery operations and the consolidation of temporary service points), the system adopts a dynamic update mechanism of “time window + conflict detection”. The core operations are as follows:

(1)

Incremental Synchronization: Position data is synchronized once every 15 min through the agricultural machinery Beidou terminal and the API of the service agency, and only nodes with a difference of more than 500 m from the previous record are transmitted.

(2)

Conflict Detection: If the same service node receives more than three conflict coordinates within 10 min (for example, agricultural machinery appears simultaneously in County A and County B), it will be marked as “Pending Review” and pushed to the mobile end of the township administrator.

(3)

Index Optimization: R-tree spatial index reconstruction is performed on the spatio-temporal graph every 24 h to enhance query efficiency. It is measured that under a scale of 100,000 nodes, the response time of the nearest neighbor query is reduced from 1.2 s to 0.3 s.

The pseudo-code of this dynamic update mechanism is shown as follows (

Table 2. Pseudo-code of dynamic update mechanism of "time window + conflict detection".

Pseudo-Code(Python Language): Dynamic Update Mechanism

def update_spatiotemporal_graph(new_data):         for record in new_data:                 if record[‘timestamp’] > last_update_time: # only process new data                         existing_node = query_graph(record[‘service_id’])                         if existing_node:                                 if conflict_detected(existing_node, record): # conflict detection                                         notify_admin(record[‘service_id’]) # notify administrator                                 else:                                         merge_node(existing_node, record[‘location’], record[‘time’])                         else:                                 create_node(record[‘service_id’], record[‘location’], record[‘time’])         optimize_graph_index() # reconstruct the spatial index

):

In the above code, the “conflict_detected()” function determines the conflict by comparing the Euclidean distance between the historical trajectory and the current coordinate; The “optimize_graph_index()” function calls the ST_IndexBuild method of the GeoSpark library.

The above methods have strong reference significance for the spatiotemporal map services of rural characteristic industries, enabling multi-source data integration and precise spatiotemporal information services. Graph databases like Neo4j demonstrate advantages in processing dynamic data, efficiently storing complex relationships through RDF triples and “node-edge” models. The application layer provides various service interfaces, and further leveraging middleware such as Neo4j spatial can support the query, analysis, and visualization of spatiotemporal data, while interoperating with GIS platforms that comply with Open Geospatial Consortium (OGC) standards, ensuring compatibility and security of data sharing [35,36,37,38].

4. Spatial Semantic Matching

Spatial semantic matching is essential for achieving efficient scheduling and precise matching of socialized service resources in rural characteristic industries. A comprehensive analysis of the spatial distribution characteristics and semantic information of resources can enhance the targeting and timeliness of service supply, thereby meeting the diverse needs of farmers.

4.1. Spatial Distribution Characteristic Analysis

The spatial distribution of rural characteristic industries is influenced by multiple factors, including natural resources, policy support, and market conditions [39]. These factors jointly affect the layout and development of rural characteristic industries, forming industrial agglomeration characteristics in different regions. For example, the spatial distribution of “one village, one product” demonstration villages and towns in Hunan Province shows significant clustering characteristics, mainly concentrated in areas with obvious regional characteristic resources and around cities with convenient transportation and developed economies, which also provides favorable conditions for the effective implementation of socialized services [40,41].

4.2. Research on Spatial Semantic Matching Methods

4.2.1. Traditional Semantic Matching

Traditional semantic matching methods mainly rely on rule-based matching strategies and static databases. These methods typically employ keyword matching and string processing techniques, using similarity calculations (such as cosine similarity and the Jaccard coefficient) to evaluate the similarity between different place names or addresses [25].

Standard methods include rule-based matching, character similarity comparison, and dictionary-based matching (

Table 3. Traditional semantic matching methods.

Objective	Principle	Technology
		Type	Name	Purpose
Text Semantic Matching	Character-based	Edit Distance	Levenshtein	Calculate similarity between two strings, suitable for spelling correction and text matching
			Jaccard	Measure similarity between two sets, used for text clustering and classification
	Rule-based	Character Similarity Comparison	Hamming	Compare character differences between place names or addresses, but unable to handle semantic-level associations
	Dictionary-based	Dictionary Matching	Dictionary-based	Associate place names with features, but limited in matching capability for new words or variants

). Rule-based matching matches input place names with those in the database one by one through preset rules, lacking flexibility and having difficulty handling newly generated place names or changing linguistic expressions. Character similarity comparison uses algorithms such as edit distance and Hamming distance to compare character differences between place names or addresses. Although simple and effective, it cannot consider semantic-level associations. Dictionary-based matching relies on constructed toponymic dictionaries, associating specific place names with their corresponding features. While it can solve some synonym problems, its matching capability for new words or variants is limited [25].

4.2.2. Collaborative Process of Semantic Matching and Spatio-Temporal Matching

The limitations of traditional semantic matching methods indicate that relying solely on text similarity cannot meet the precise demands of agricultural scenarios. Therefore, a collaborative process for semantic and spatio-temporal matching needs to be proposed to achieve a leap in matching accuracy by dynamically fusing the two types of features. This process includes the following three steps:

(1)

Seman-to-space mapping: Convert the feature vectors output by the semantic analysis module (such as “harvester = [0.8, 0.2,..., 0.1]”) into geofencing parameters (center coordinates, radius) through the radial basis function (RBF) kernel, for example:

$$Radius=50\times \sigma \left(w^{T}x+b\right)$$

(2)

Spatio-temporal query execution: After receiving the geofencing parameters, the spatio-temporal module returns the matching node through Neo4j’s Cypher query. An example is shown as follows (

Table 4. Example of spatio-temporal query execution.

Example: Spatio-Temporal Query Execution
>      ‘‘‘cypher     >      MATCH (n:ServiceNode)     >      WHERE n.feature CONTAINS ‘Harvester’     >      AND n.location WITHIN 50 km OF [116.4, 39.9] // Beijing coordinates     >      AND n.available_time BETWEEN ‘2024-06-01’ AND ‘2024-06-10’     >      RETURN n.service_id, n.location     >      ‘‘‘

):

(3)

Result fusion: For the nodes returned by spatio-temporal queries, a secondary screening is conducted based on semantic similarity (such as cosine similarity > 0.85), and they are ultimately recommended to farmers.

4.2.3. Comparison and Analysis of Spatial Semantic Matching Method

Although the collaborative process significantly improves the matching effect, its utilization of spatio-temporal data is still relatively coarse-grained (such as only using GeoHash grids). To this end, this section will explore how to mine deeper semantic information from spatio-temporal data and further optimize matching performance.

Traditional methods have limitations when dealing with polysemy, synonyms, and semantic changes [42,43,44,45,46,47,48,49,50]. Therefore, spatial semantic matching needs to utilize various techniques such as string matching, content feature comparison models, and semantic similarity calculations to achieve matching between place names and addresses [51]. The common-used methods of spatial semantic matching are shown in

Table 5. Spatial semantic matching methods.

Objective	Principle	Technology
		Type	Name	Purpose
Spatial Semantic Matching	Spatial Distance-based	Distance Measurement	Euclidean Distance	Used to evaluate spatial proximity between two geographic entities
			Manhattan Distance	Calculate distance between spatial entities through coordinate axes, suitable for distance calculation in grid layouts
	Spatial Relationship-based	Adjacency Relationship	Intersection	Determine whether two spatial objects intersect, commonly used in geographic information systems
	Numerical-based	Numerical Matching	Attribute Comparison	Compare similarity of attributes such as road width and reservoir area
	Encoding-based	Encoding Matching	Classification Code	Achieve matching by comparing the similarity of feature classification codes
	Spatial Semantic-based	Semantic Matching	Knowledge Graph	Use contextual information of geographic entities for semantic association to enhance matching accuracy

. When creating semantic databases and service platforms for toponymic addresses, these methods can effectively describe and label basic resource data, forming comprehensive toponymic address data knowledge bases [51].

A spatial data semantic expression similarity calculation is achieved by comparing the semantic similarity of geographic entities from different data sources. This method utilizes numerical, encoded, and textual information in multi-source vector spatial data for semantic matching, thereby improving the accuracy and efficiency of spatial data fusion. For example, numerical-based similarity calculation can be used to compare the similarity of attributes such as road width or reservoir area; code matching is achieved by comparing the similarity of feature classification codes; text similarity calculation matches toponymic information through string similarity algorithms, thereby achieving semantic association of geographic entities across different data sources [50].

Furthermore, modern methods have introduced natural language processing and machine learning technologies. These technologies achieve mapping between geographic information and natural language descriptions through the semantic web and full-text retrieval, significantly enhancing matching accuracy and efficiency [51]. Specifically, these methods combine the comprehensive advantages of text segmentation methods based on natural language processing and semantic similarity calculations of spatial data [51], achieving more precise spatial data semantic association [33,51].

The application of machine learning in spatial semantic matching mainly focuses on feature learning and semantic similarity matching. More precise semantic similarity matching can be achieved by learning features from training sets to build models. Combined with deep learning models such as BERT and Siamese networks, important semantic representation features can be automatically learned from training data, avoiding the limitations of traditional manual feature engineering [52].

4.3. Application of Spatial Semantic Matching in Rural Characteristic Industries

Spatial semantic matching has extensive application scenarios in the rural characteristic industry socialized services (

Table 6. Typical Application scenarios of spatial semantic matching for rural characteristic industry socialized services.

Service Scenario	Application Requirements	Solutions
Intelligent Rice Seedling	Accurately grasp the suitable spatial distribution of seedling sites and transplanting timing, rationally recommend seedling resources, ensure matched resources meet optimal agricultural timing and geographical conditions, improve seedling quality and efficiency	Collect data on rural natural resources, regional environment, seedling environmental requirements, and production activities of subjects and objects. Apply spatial semantic matching and time tree technology to construct spatiotemporal graphs, and intelligently recommend service subjects through cloud service platform big data analysis algorithms
Mechanical Rice Transplanting	Precisely plan mechanical transplanting operation areas, optimize transplanter scheduling, improve precision and efficiency of mechanical transplanting operations, reduce agricultural machinery resource waste	Integrate spatial data including terrain and landform of rural farmland, land ownership, distribution of planted varieties, and crop maturity distribution in different regions, as well as semantic information such as mechanical transplanting operation standards and agricultural machinery subsidy policies. Through spatial semantic matching technology, determine suitable areas for mechanical transplanting, service subjects, and operation timing
Rural Characteristic Fruit Picking and Delivery	Optimize picking and delivery services, improve service efficiency, rationally allocate picking subjects and delivery vehicle resources, ensure fruits are delivered from orchards to markets timely and efficiently	Apply spatial semantic matching technology to construct geographic knowledge graphs of orchards, picking subjects, and delivery vehicles. Combine time tree technology to integrate temporal graphs of maturity periods, working hours, and transportation time. Through spatiotemporal correlation analysis and dynamic optimization, achieve precise matching of location and time data and resource scheduling

). In rural characteristic industries, applying spatial semantic matching technology is critically important, enabling deep analysis of the intrinsic connections between industrial layout and natural resources, policy support, and market conditions. By precisely matching industrial spatial data with the semantic information of these factors, particularly spatiotemporal information, it is possible to effectively identify the core driving factors of industrial development and further explore targeted optimization strategies [53,54,55].

5. Resource Spatiotemporal Map Service Construction

The socialized service links of rural characteristic industries are numerous, with complex personalized demands, thus requiring a classified analysis of multi-source heterogeneous service resources. Spatiotemporal maps of service resources can be formed by exploring and visualizing their relationships based on spatiotemporal characteristics, varieties, and service links. These maps are essential for optimizing the allocation and efficient utilization of rural characteristic industry service resources.

Spatiotemporal maps integrate multi-dimensional spatial data and temporal information to intuitively display the spatial distribution, temporal changes, and interrelationships of characteristic industry resources. It provides an essential basis for decision-makers, service providers, and farmers in their decision-making processes, supported by real-time data. Constructing a spatiotemporal knowledge graph for socialized services in characteristic industries is an effective data organization method. It helps understand the dynamic associations and characteristics between various service links by constructing a triple network structure of entities, their attributes, and relationships. The construction process of the knowledge graph is shown in Figure 4: Firstly, geographic information data from different sources are classified into three categories: structured data, semi-structured data, and unstructured data. D2R and KBP are, respectively, used for knowledge extraction to obtain the geographical entities and various relationships (spatial, semantic, and temporal) contained therein. Meanwhile, spatial relationship extraction is carried out on the three types of data according to geographical rules to obtain spatial relationships such as topology, direction and distance. At the same time, the updated geographic knowledge base is obtained through entity extraction and knowledge merging of the data collected from Baidu-Baike. Finally, by fusing the three types of data obtained from knowledge extraction, spatial relationship extraction and updated knowledge base with entity linking techniques such as relationship extraction and entity disambiguation, the geographic knowledge graph with spatiotemporal feature can be generated.

In the construction process, it is first necessary to integrate multiple data sources, including basic geographic data, rural characteristic industry spatial data, encyclopedic data, and ubiquitous geographic information data. Using advanced technologies such as Geographic Information Systems (GIS), remote sensing technology, and big data analysis, these data undergo structured, semi-structured, and unstructured data processing to form a comprehensive data foundation [56].

Based on these data, multi-dimensional models combine indicators from multiple dimensions, including social, economic, environmental, industrial, and service aspects, to form resource spatiotemporal knowledge graphs. The typical example is shown in Figure 5. Key elements in the graph, such as topological relationships, directional relationships, and distance relationships, help better understand the spatial and temporal dynamics between resources. Additionally, by employing knowledge extraction technologies, geographic entities, spatial relationships, and temporal relationships are integrated, enhancing resource dynamic updating and real-time service capabilities [35].

In recent years, researchers have extensively studied neural network-based automatic construction of knowledge graphs, particularly the application of deep learning technology in natural language processing and image recognition, promoting the efficiency of knowledge graph construction and updating. By utilizing Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), researchers can more effectively extract relationship and entity information from large-scale datasets [57,58], making knowledge graph generation more automated and intelligent. Furthermore, the introduction of technologies such as Graph Neural Networks (GNNs) can further enhance the learning capability of graphs and the representation of structured information [59], which is of great significance for constructing complex resource spatiotemporal maps.

It is worth noting that large language models have demonstrated strong potential in autonomous knowledge graph construction, bringing new ideas and methods to geographic knowledge graph construction. Couto and Ebecken constructed knowledge graphs for domains like psychoanalysis using large language models. They achieved efficient knowledge graph construction by employing models such as Llama 2 7B, Llama 3 8B, ChatGPT 3.5, and ChatGPT 4 to determine nodes (relevant terms) and edges (relationships between terms) through specific prompt engineering, thereby streamlining the traditionally labor-intensive process. It provides a valuable reference for applying large language models in geographic knowledge graph construction, potentially improving the efficiency of geographic knowledge graph construction, discovering latent relationships, optimizing graph structures, and promoting the development of knowledge management and decision support in the geographic field [60].

Resource spatiotemporal map construction also needs to focus on the user interaction experience. Users can customize queries and analyses by developing user-friendly visualization interfaces and intelligent analysis tools, thereby enhancing service targeting and user engagement. With the advancement of information technology, dynamically updating map information and optimizing resource scheduling paths will provide continuous support for the sustainable development of rural characteristic industries.

In summary, constructing resource spatiotemporal maps provides a scientific basis for socialized services in rural characteristic industries. It facilitates the efficient utilization and management of resources, promoting the sustainable development of rural industries.

6. Construction of Spatiotemporal Platform for Socialized Services in Rural Characteristic Industries

Developing rural characteristic industries is essential for promoting rural revitalization, and constructing socialized service platforms is key to enhancing industrial efficiency and competitiveness. At the technical level, relevant spatiotemporal platforms involve technologies such as Spatial Data Infrastructure (SDI), OGC-compliant ArcGIS/GeoServer map services and spatial analysis platforms, and Neo4j spatiotemporal data relationship mining and analysis. Through integrating and applying these technologies, efficient data management and sharing can be achieved, narrowing the digital divide and thus supporting the sustainable development of rural characteristic industries [61,62,63,64].

Existing socialized services mainly include the China Agricultural Technology Extension Information Platform and the China Agricultural Socialized Service Platform, which serve the entire nation, and regional service platforms that serve specific areas [65]. Based on existing research findings, industrial needs, and various technological advantages, the spatiotemporal platform architecture for socialized services in rural industries can be designed as a layered distributed architecture, as illustrated in Figure 6. The main components of the spatiotemporal platform include relational databases, spatiotemporal middleware, graph databases, Java/Python Web Server, and GeoServer. The relational database stores directory databases, services, demands, orders, credit, and other rural characteristic socialized service data resources; the spatiotemporal middleware selection can integrate with Neo4j’s Timetree/Spatial Middleware, which can construct spatiotemporal graphs according to resource matching and scheduling needs; Java/Python Web Server provides rural characteristic industry socialized service portals and AI algorithm services; GeoServer provides OGC-compliant visualization map services for geographic knowledge graphs and spatiotemporal graphs, offering efficient spatiotemporal scheduling services for matching supply and demand resources between subjects and objects.

6.1. Intelligent Matching of Socialized Service Resources

In recent years, the problem of supply–demand mismatch in the agricultural socialized service system remains prominent, presenting a state where large-scale and fragmented services coexist [66]. The intelligent matching system for socialized services is one of the core objectives in current socialized service platform construction, aiming to achieve efficient matching between service supply and demand through intelligent means, thereby improving service quality and user satisfaction [67]. Gao Mengtian et al. applied the socialized service system platform to match fragmented and discrete farmland service demands intelligently, achieving minimum-cost plant protection drone scheduling based on the agricultural socialized service platform [68].

Related intelligent matching technologies are mainly reflected in the following three categories: (1) Application of big data and artificial intelligence analysis technologies. Intelligent matching technologies that construct spatiotemporal knowledge graphs from actual operational data of rural characteristic industry socialized services based on business logic and use graph convolution and attention mechanisms to mine users’ potential intentions have begun to be researched and applied. For example, Wang Xinyue et al. proposed a knowledge recommendation method integrating rural residents’ intentions, achieving precise recommendations through relational path modeling and improving service resource matching efficiency [69]. In rural characteristic industry socialized services, such technologies can be used for intelligent matching, such as mining potential correlations between multi-source information of subject–object users and industrial service resources. (2) Spatiotemporal analysis. Matching spatiotemporal graph entities according to the shortest time and shortest path rules. For example, Zhang Tianming et al. used CTG-tree (Compressed Transformed Graph tree) indexing, first compressing and transforming temporal graphs, then constructing indexes, and quickly calculating the shortest temporal paths based on this during online query phases to achieve efficient resource matching [70]. In rural characteristic industry socialized services, the most suitable service resources can be quickly found based on the spatiotemporal relationships between service providers and demanders. (3) Spatiotemporal scheduling. In addition to the above model algorithms for intelligent resource matching, spatiotemporal resource scheduling can be directly performed through functions provided by the Neo4j graph database and its spatiotemporal middleware in the spatiotemporal map system integration environment. For example, Neo4j’s shortestPath is used to schedule entities with the shortest path meeting conditions, determining optimal connections between service supply and demand; GraphAware Neo4j TimeTree’s ga.timetree.events.range can filter service events within specific time ranges for precise supply–demand time matching; Neo4j spatial’s spatial.bbox achieves spatial positioning and filtering of service resources through precise geographic coordinate range queries, assisting in completing intelligent resource matching and scheduling.

6.2. Spatiotemporal Graph System

Using Neo4j combined with the Neo4j Spatial library or self-developed rules and tools for geographic knowledge graph construction has become a new trend at the system platform construction level. Hui Li addressed problems faced by relational databases in spatiotemporal data modeling and storage, such as difficult structural conversion, complex queries, low efficiency, and poor scalability [71], establishing a spatiotemporal data model based on Neo4j and the Neo4j Spatial library. This model integrates and stores the three basic elements of temporal geographic information systems—time, space, and attributes—and explicitly expresses the spatiotemporal characteristics of temporal geographic information systems semantically [71].

Furthermore, temporal graphs based on GraphAware Neo4j TimeTree have unique advantages in managing and expressing the temporal dimension, which can further refine the association and display of time nodes. Taking mechanized seedling cultivation service tasks as an example, TimeTree can present the expected start and completion times of tasks and the time dependencies between subtasks, as shown in Figure 7.

Using a three-dimensional spatial representation, this figure illustrates a spatiotemporal graph of mechanized seedling cultivation service tasks constructed based on Neo4j. In the graph, temporal entities center around task nodes connected to start and end time nodes through “SHOULD_BE_STARTED” and “SHOULD_BE_COMPLETED” relationships. In contrast, each time a node is connected to other subtask nodes through relationships such as “FIRST,” “LAST,” and “CHILD,” it clearly presents the temporal context of task execution and hierarchical task relationships. This spatiotemporal graph not only reflects expectations in the temporal dimension but can also manage spatial location information, such as seedling cultivation sites, in combination with the Neo4j Spatial library, achieving integrated display and analysis of spatiotemporal data, providing strong support for task planning and resource scheduling in socialized services for rural characteristic industries.

6.3. Spatiotemporal Map Service System

In constructing spatiotemporal map service systems, ArcGIS 10 and other platforms [29,72] have been widely used as basic platforms, which can be deployed in the cloud and provide spatial services through WMS, WFS, WCS, etc. In research on characteristic agricultural industries in Shaanxi Province, researchers used ArcGIS 10.2 software to analyze the spatial distribution of characteristic agricultural industry points. By obtaining agricultural industry point-of-interest data crawled from Amap and combining it with DEM elevation data provided by the geospatial data cloud platform, researchers successfully visualized the spatial distribution characteristics of characteristic agricultural industries in Shaanxi Province [29,73].

Based on existing technical research, this paper designs a prototype system for spatiotemporal map services for socialized services in rural characteristic industries, as shown in Figure 8. This system is based on the technical architecture in Figure 6, developed using open-source tools such as GeoServer, and addresses the needs of socialized services for rural characteristic industries by implementing management services for resources such as orders, service organizations, service demands, and demonstration bases through spatiotemporal graphs. The system integrates the Neo4j graph database through Timetree/Spatial middleware, uses Java language to implement interactions with the Neo4j graph database and GeoServer, provides spatiotemporal map management services, and provides technical support for socialized services in rural characteristic industries.

Although the spatio-temporal map service system has achieved breakthroughs at the technical level, its implementation in rural areas still faces multiple challenges, such as infrastructure and user capabilities.

6.4. Challenges and Responses to Technology Implementation in Rural Areas

(1)

Lightweight deployment

In response to the unstable problem of rural networks, Raspberry Pi 4B edge nodes (configuration: 4 GB RAM, 32 GB SD card) are deployed in towns and townships to cache high-frequency data (such as the location of agricultural machinery and weather warnings). The caching strategy is:

Heat ranking: Prioritize caching data that has been queried more than 10 times within 24 h.

Expiration cleanup: Delete data with a cache time exceeding 48 h every 6 h.

In actual tests, under a 2G network, edge nodes can independently support 10 concurrent users, and the map loading time has been reduced from 8 s to 3 s.

(2)

Low-bandwidth optimization

The map tiles are compressed in WebP format (originally in PNG format), with the compression rate increased by 70%. The specific parameters are:

Quality parameter: 80 (0–100, the higher the value, the higher the quality).

Compression algorithm: Lossy (lossy compression, the difference is hard for the human eye to detect).

Under a 2G network (with a bandwidth of approximately 50 kbps), loading a 10 km × 10 km area map has been reduced from 12 s to 3 s.

(3)

Digital Literacy Training

The Joint Agricultural Technology Station is launching a “Mobile Phone Farming” workshop. In 2024, the pilot program will cover 50 administrative villages. The training content includes:

Voice input requirement: Dialect recognition is achieved through iFLYTEK SDK (supporting six dialects, including Henan and Shandong), with a recognition accuracy rate of over 90%.

Agricultural machinery location check: Teach farmers to use the WeChat mini-program to check the real-time location of nearby agricultural machinery. The operation steps are:

1. Open the mini-program → 2. Click “Find Agricultural Machinery” → 3. Select the service type (such as “Harvester”) → 4 View the available agricultural machinery within 50 km.

7. Conclusions and Outlook

7.1. Conclusions

As the demand for socialized services in rural characteristic industries deepens and artificial intelligence technology accelerates its penetration into the agricultural field, semantic-level feature spatial expression and resource spatiotemporal map technology for socialized service resources in rural characteristic industries are developing toward intelligence, precision, and integration. This process enhances the rural industries’ service efficiency and resource optimization allocation capabilities. It provides new opportunities for integrating the latest digital technologies, cloud computing, and big data.

Constructing a comprehensive spatiotemporal resource-mapping service platform can make full use of the complementary advantages of multi-modal spatio-temporal data fusion modeling and promote the efficient scheduling and precise matching of various service resources, help farmers better access corresponding socialized services, facilitate the sustainable development of rural characteristic industries, and achieve rural revitalization goals.

7.2. Trend Analysis

7.2.1. Transformation from Pure Information Services to Deepened Integration of Knowledge Services

The scope of socialized services has undergone a leapfrog transformation. Like traditional agricultural machinery services for cultivation, planting, management, and harvesting that solve problems of low production efficiency and labor shortages, agricultural socialized services, especially socialized services for rural characteristic industries, also play an essential role in comprehensively improving food production capacity and solving the question of who will farm. Moreover, this process requires urgent agricultural technology and agricultural knowledge, and traditional information services in socialized services will inevitably evolve toward deeper penetration of knowledge services.

7.2.2. Enhanced Application of Spatiotemporal Services Represented by Geographic Knowledge Graphs

Regarding information service forms, without excluding traditional graphical interface services, as the spatial characteristics of rural characteristic socialized services deepen, geographic knowledge graphs integrating geospatial data and temporal attributes have unique advantages in enhancing semantic reasoning and matching. Particularly, the integrated application of generative large models and graphs will inevitably be widely applied in human–computer interactions, knowledge services, and resource scheduling.

7.2.3. Deepened Application of Standardized Production Services Connected to the Internet of Things

With the development of big data and artificial intelligence technology, IoT data mining and analysis oriented toward efficient production management will be deeply applied, playing an essential role in the standardized production of socialized services [74].

7.2.4. Rising Demand for Smart Unmanned Farm Socialized Services

As urbanization and the aging of agricultural practitioners deepen, unmanned farm technology has already been prototyped in many domestic regions [75]. Before a large-scale application is achieved, the demand for unmanned farm socialized services based on entities with specific operational capabilities is becoming increasingly strong. It can be used to leverage market-oriented applications in advance and accelerate the development process of smart agriculture.

7.3. Implementation Path

In response to the actual application demands in rural areas, we have proposed a “three-step” implementation path as follows:

(1)

2024–2025: Pilot projects will be carried out in five provinces to verify the technical feasibility.

(2)

2026–2027: Formulate industry standards and promote them to major grain-producing areas.

(3)

After 2028: Integrate AI prediction models to achieve automatic scheduling.