Economic policies assessment and judgement during the pandemic with semantic and social network joint analysis

Abstract

This paper delves into the economic policies of China during the pandemic and investigates the relationships between policy-issuing institutions. Firstly, we conduct keyword extraction and statistical analysis based on policy texts to understand policy contents and distribution. Then, we calculate the co-occurrence matrix of policy keywords and publishers from social networks and visualise the relationships with UCINET6 and Gephi software. Through a combination of semantic analysis and social network analysis, we examine the content of relevant economic policies, laws, regulations, and the relationships between their publishers. Our findings reveal three stages of China’s economic policies during the crisis, i.e. the shock, stable, and boost periods, which align with the crisis’s impact on China’s economy. Initial policies supported small-sized enterprises (SMEs), followed by a focus on industries like tourism. During the boost phase, policies underscored various support measures, including tax and fee reductions. We also identified a “local agglomeration” characteristic among policy-issuing entities, suggesting potential improvements in cooperation, especially at the provincial level. Our findings provide valuable insights for future policy design in response to public security events.

KEYWORDS:

• Semantic analysis
• social network analysis
• economic policy content
• economic security

1. Introduction

The outbreak of the COVID-19 pandemic has unleashed profound and far-reaching consequences on global economies. Governments and enterprises have responded with regulations and initiatives, leading to a stabilisation of the pandemic’s effects and an economic recovery (Shen & Xu, 2020; Wang, 2021). During the pandemic, it is essential to analyse policy changes and entities involved uncovers efficient macro-control strategies through the computer technologies (Dai, 2021; Wang, 2020), thus improving emergency response capabilities. In this paper, we investigate economic policies and employ semantic analysis and social network analysis to explore the paths of policy changes, priorities, and strategic deployment in China’s economic development during the pandemic.

There are a number of studies that utilise social media and internet analytics to investigate the impact of COVID-19 on different aspects of society and economy, e.g. tourism (Abbas et al., 2023), healthcare (Azadi et al., 2021; Iorember et al., 2022), business (Yu et al., 2022), education (Abbas et al., 2019; Meng et al., 2023), and environment. For policy analysis, existing studies mainly follow two major research encompasses two main research directions: (i) directly summarising the characteristics and key points of the policies from classical and important policies contents and texts (Hu, 2020; Liu & Pan, 2021; Qie & Zhang, 2020; Wang, Wang, et al., 2023). (ii) using natural language analysis tools, such as semantic analysis and text mining, to conduct knowledge learning from a large amount of text (Wang, 2021; Zhang et al., 2018; Zhang et al., 2019). For instance, thematic analysis studies variability, evolutionary patterns, and mismatches in science and technology policies (Wang, 2021; Zhang et al., 2018; Zhang et al., 2019), while machine learning examines the implementation effects of financial policies during different periods of the pandemic (Wang, Abbas et al., 2023; Zou, 2021).

Regarding policy-issuing entities, scholars have focused on two aspects. Firstly, the hierarchical structure of policy-issuing entities and the governance capacities at each level have been explored (Liu et al., 2017; Ma et al., 2020; Pei et al., 2016). Secondly, the synergistic operations of policy subjects have been investigated using social network analysis, identifying core nodes and network structures (Ding et al., 2021; Liu et al., 2021; Liu & Xu, 2012).

The outbreak of the pandemic has disrupted various sectors and necessitated the implementation of new policies and initiatives by governments and enterprises to stabilise the effects and facilitate economic recovery. In this context, it is crucial to analyse the policy changes and entities involved in order to uncover efficient macro-control strategies and improve emergency response capabilities. However, existing research on policy analysis during the pandemic has mainly focused on summarising policy characteristics or conducting knowledge learning from a large amount of text. Moreover, studies on policy-issuing entities have either explored hierarchical structures or investigated synergistic operations using social network analysis, but have not effectively integrated semantic analysis and social network mining to thoroughly investigate the content of economic policies and entity relationships.

In this paper, we utilise semantic analysis and social network modelling simultaneously to analyse policy content and the relationships among policy-issuing entities. We present an in-depth data-driven analysis of Chinese economic data during the pandemic and summarise and analyse China’s economic policies, laws, and regulations. In addition, we employ semantic analysis and social network techniques to explore policy content and agent relationships. These methods enable us to analyse policy characteristics and identify key policy agents and their evolution trends. In summary, the contributions of this work are as follows.

• Firstly, we analyse the economic policies of China during the pandemic, with a specific focus on the computer field. By integrating the methodologies of semantic analysis and social network analysis, we offer a novel approach to understanding policy content and the relationships among policy-issuing entities. The proposed conceptual framework provides valuable insights into crisis management and policy design.

• Secondly, we leverage advanced natural language processing and social network modelling techniques to analyse a large volume of policy texts and construct relationship networks. Semantic analysis empowers us to extract crucial information and unveil concealed patterns embedded within policy content. Social network analysis enables us to visualise and quantify the connections between various policy actors.

• Lastly, we conduct a comprehensive examination of economic policies implemented during the pandemic, focusing on the three stages of shock, stability, and boost. Through quantitative methods and data visualisation, we identify the priorities and strategic deployments in these stages and provide a nuanced understanding of policy responses. These empirical findings contribute to the existing body of knowledge on crisis management and provide valuable insights for policymakers in future global crises.

2. Related work

Many scholars have explored and analysed the field of policy analysis in the past. Wang, Hu et al. (2023) analysed the policy of the Belt and Road Initiative. From the perspective of policy data sources, the articles generally use public policy documents as data sources, and some articles divide policy research groups according to important time nodes. For example, the policy documents can be divided into three stages in the study of the characteristics and evolutionary logic of new technology policies. From the perspective of policy analysis, scholars mainly use two research paths. One is to start from classic and important policies and summarise the characteristics and key points of the policies; for example, Hu Yanni and Zhang Lin (Hu, 2020) summarised the important policies during the COVID-19 epidemic and obtained the key points of China’s anti-epidemic work at each stage. The other is to use natural language analysis tools, such as semantic analysis, text mining, etc., to perform knowledge learning on a large amount of text. For example, Zhang Baojian and Li Pengli (Zhang et al., 2019) used topic analysis in text mining to study the differences, evolution laws, and mismatches in science and technology policies. From the perspective of the final result of policy analysis, the articles mainly focus on the policy characteristics, evolution paths, policy priorities, and themes of each stage. Focusing on the economic policy analysis during the epidemic, scholars use policy qualitative combing, natural language analysis, and machine learning methods to explore the policy characteristics and effects of China’s fiscal, financial, import and export, industrial chain, and supply chain under the impact of the epidemic. For example, Liu and Pan (2021) used the combination of economic data analysis and policy research to refine China’s policy path and put forward suggestions for future policy design; Zou (2021) used machine learning methods to study the final implementation effect of financial policies in the three major industries.

From the perspective of policy subject analysis, scholars’ research mainly focuses on three aspects. One is to sort out the hierarchical structure of policy-issuing subjects, such as Liu et al. (2017) based on China’s network public opinion governance issues, sorted out the hierarchical structure of 53 national-level policy subjects, and discussed the governance capabilities of each level of subjects; the second is to use historical analysis to study the changes of policy-making subjects and their cooperation mechanisms, and then expound the impact of policy subject coordination on policy implementation effects (Abbas et al., 2023; Micah et al., 2023). For example, NeJhaddadgar et al. (2020) investigated the effectiveness of telephone-based screening in the recommended health system during the COVID-19 outbreak and made decisions; the third is based on big data processing methods such as social networks and human–computer interaction, to analyse the degree of coordination between policy subjects and explore their cooperation (Abbas et al., 2019; Azadi et al., 2021; Meng et al., 2023), this kind of research method has been used by more scholars in recent years, such as Liu et al. (Liu & Xu, 2012) used social network analysis methods to draw the cooperation network map of China’s science and technology policy-making subjects at different development stages, and identified and extracted the core nodes and network structure characteristics. Liu et al. (2021) used social network analysis methods to draw the cooperation network of policy-issuing subjects during the COVID-19 epidemic, and calculated the structural characteristics of the cooperation network using social network analysis methods. Yu et al. (2022) proposed social media application as a new paradigm for business communication. Al-Sulaiti et al. (2023) analyse the dine-out behaviour of tourists by collecting their online information, thereby influencing relevant decisions.

In addition, some studies have evaluated and discussed the corresponding policies from an environmental perspective, combining social and natural factors (Al-Sulaiti et al., 2023; Schmidt et al., 2022; Shah et al., 2023). Iorember et al. (2022) discussed the criticality of human capital development and energy use from an ecological perspective. In terms of sustainable development policies, Balsalobre-Lorente et al. (2023) pointed out the interdependence between tourism, urbanisation, natural resource rents, and environmental sustainability in their paper, thereby setting sustainable development goals. Li et al. (2022) also discussed the role of policies related to social value creation in environmental sustainability from the perspective of tourism and business.

3. Motivation

This section introduces the motivation of our work, i.e. the changes in the economy under the impact of the epidemic. By examining Chinese economic data from December 2019 to June 2021, we identify that the impact of the epidemic on China’s economy is categorised into three distinct stages.

3.1. Macro trade volume changes

As shown in Figure 1, witnessed significant fluctuations in merchandise trade over the years. Notably, there was a sharp decline of 20.1% in 2008 due to the financial crisis, followed by a further decrease of 26.9% in 2016. In 2020, the world experienced yet another challenging period, with a substantial decline of 6.3% in merchandise trade. It is important to note that these declines were influenced by various factors, with the financial crisis being the primary cause in 2008.

Figure 1. Total World Trade and Growth Rate (2007–2020).

From 2011 to 2015, the world trade scenario remained relatively stable. However, starting in 2017, the global economy faced mounting downward pressure, resulting in a modest growth rate of 3.4% throughout the year. The situation worsened in 2020, as numerous economies implemented isolation measures and blockades to combat the spread of the pandemic. Consequently, global trade suffered severe shocks, leading to a significant drop in the volume of commodity trade, which reached its lowest point in a decade, amounting to $4.34 trillion.

While countries around the world grappled with the economic repercussions of the pandemic, China displayed resilience and sought opportunities amidst the crisis. In 2020, China’s total goods trade import and export value reached an impressive 32.15 trillion yuan. Furthermore, in the first half of 2021, China continued to demonstrate its momentum by achieving a trade value of 18.07 trillion yuan, representing a remarkable increase of 27.1% compared to the same period last year.

As shown in Figure 2, a distinct pattern of stages is evident when examining the total import and export trade. The outbreak of the pandemic had a significant impact on the economic trade, including the computer field, during the period spanning from December 2019 to February 2020. This period witnessed substantial disruption to the economic trade between various nations, including China.

Figure 2. Monthly Statistics of China’s Import and Export Trade, 2019–2021.

Following the initial impact, there was a notable recovery observed from February 2020 to March 2020, characterised by stabilised total trade and positive growth in import and export values. Subsequently, a period of stabilization ensued, with trade gradually and steadily recovering from March 2020 to December 2020. Regrettably, in January and February of 2021, the pandemic underwent a mutation, resulting in a resurgence of international cases. This had a severe impact on trade. However, by March, there was a rapid recovery in total trade volume. Regarding the trade balance between imports and exports, China generally maintained a trade surplus from December 2019 to June 2021. However, during the peak of the pandemic in February 2020, the scale of China’s exports was severely affected, leading to a temporary trade deficit. By June 2021, the scale of China’s imports and exports had largely returned to pre-pandemic levels, indicating a significant rebound in trade activity.

As evident from Figure 3, amidst the impact of the pandemic, China’s trade experienced a significant decline in year-on-year growth, entering a state of negative growth. However, as the situation stabilised, there was a gradual recovery in trade with an accelerated increase in export growth. By October 2020, the year-on-year growth rate of exports had reached 10.9%. Conversely, the growth rate of imports displayed a slower rise. In 2021, both exports and imports witnessed a remarkable year-on-year growth rate surpassing 18%. Particularly noteworthy is the fact that in March, the year-on-year growth rate of exports exceeded 100%, indicating a substantial rebound compared to the previous year’s sharp decline. Therefore, it concluded that China’s export trade has successfully regained its momentum.

Figure 3. Monthly Year-on-year Growth Rates of China’s Imports, Exports, Imports and Exports, 2019–2020.

The motivation for conducting this research stems from the need to understand and analyse the dynamics of global merchandise trade, particularly in the context of significant fluctuations and disruptions caused by events such as financial crises and the COVID-19 pandemic. By examining China’s trade performance and resilience during these challenging periods, we aim to gain insights into the factors that contribute to successful trade recovery and the role played by China as a major player in the global economy. This research can provide valuable information for policymakers, businesses, and economists, helping them make informed decisions regarding trade strategies, risk management, and economic stability in the face of potential future disruptions. Additionally, understanding the patterns and trends in global merchandise trade can contribute to a deeper understanding of the interconnectedness of economies and the potential impacts of trade policies and external shocks on international trade flows.

3.2. Phased change of export structure and international competitiveness of products

The United Nations International Trade Standards classify products into 10 categories. To provide a clearer representation of the prevailing trends in China’s major export product categories and facilitate the advancement and enhancement of China’s foreign trade product structure, this paper further consolidates them into the following 5 categories, as shown in Table 1.

Table 1. Classification of products according to factor tendencies.

Product Type	SITC Category
Primary products Labor-intensive products Capital, technology-intensive products Low-tech manufactured products Medium and high technology products	SITC0-SITC4 class products SITC6, SITC8 products SITC5, SITC7 products SITC5 products SITC7 products

In this paper, we have analysed data spanning from January 2016 to June 2021, and based on this analysis. In Figure 4, the horizontal axis represents the year and month, while the vertical axis is divided into two sections, representing the product structure and the absolute value of exports. Different types of products are depicted by various coloured lines, with the corresponding legend providing specific information about each type.

Figure 4. Folding Line Chart of China’s Export Product Structure and the Absolute Value of Product Exports.

As observed in Figure 4, the trends of China’s export product structure and the absolute value of exports exhibit a consistent pattern, forming two major echelons. The higher echelon encompasses “capital and technology-intensive products”, “medium and high technology manufactured products” and “labor-intensive products”. The lower echelon comprises “primary products and low-technology manufactured products.” Notably, the product structure and absolute value of exports for capital and technology-intensive products rank the highest. This indicates that China has experienced a significant proportion of exports in product categories with substantial technological content in recent years. Moreover, the overall trend of the export product structure appears to be relatively healthy.

Figure 4 illustrates that the period from December 2019 to March 2020 corresponds to a period of shock. During this time, there is a noticeable and rapid decline in the absolute values of exports for capital and technology-intensive products, medium-tech, high-tech manufactured goods, as well as labour-intensive products. However, it is important to note that the product structure curves for capital-intensive and technology-intensive products, as well as medium-tech and high-tech manufactured goods, exhibit an upward trend rather than a decline. This indicates that although the total export value of these two categories of products decreases, the rate of decline is relatively lower compared to China’s overall exports. It suggests that these product categories maintain their resilience and display strong dynamism even during a period of significant disruption caused by the epidemic.

The period from March 2020 to December 2020, as depicted in Figure 4, is characterised as a stable period. During this time, China’s exports of products that were heavily impacted during the shock period experienced a significant rebound. It signifies that China has successfully emerged from the shadow of the epidemic and entered a phase of recovery. Notably, the export value of capital-intensive, technology-intensive products, medium-tech, high-tech manufactured goods, and labour-intensive products not only recovered rapidly to pre-epidemic levels but also surpassed them, reaching a clear peak.

From December 2020 to August 2021, China’s economy entered a period of uplift, as illustrated by the line graph in Figure 4. During this period, there was a consistent and steady increase in the absolute value of exports across product categories, with a minor decline in February 2021. The curve trends indicate a positive short-term growth trajectory. Similar to the shock period, the curves for capital-intensive and technology-intensive products, as well as medium-tech and high-tech manufactured goods, continue to show a decline in the absolute value of exports. However, their relative product mix within the export structure increases, suggesting that these two categories not only display resilience in the face of the epidemic’s impact but also sustain their advantage in terms of vitality. On the other hand, the product structure advantage of labour-intensive products is no longer prominent, which is closely correlated with the continuous loss of China’s traditional labour advantage.

Therefore, understanding the phased changes in China’s export structure and the international competitiveness of its products is essential for policymakers to formulate effective trade policies, businesses to make informed investment decisions, and economists to assess the overall health and resilience of China’s economy. By examining these dynamics, we can gain a deeper understanding of China’s role in the global market and formulate strategies to enhance its position and drive sustainable economic growth.

4. Main design

This section employs text mining methodology to construct a model for semantic analysis and visually analyse policies. Semantic analysis is a natural language processing technology extensively utilised in the computer field, including Sentiment Analysis (Xu et al., 2020) and Relation Extraction (Qiu et al., 2020). It involves processing large-scale text data to extract essential keywords and key topics. The application of semantic analysis technology aims to extract the work priorities of various departments during challenging periods, such as the current global situation. These priorities encompass areas of significant concern, pivotal policy entry points, real-time strategic approaches, and key industries of focus. By utilising this approach, it becomes possible to explore the trajectory of policy changes and the overall strategic deployment within the field. Furthermore, knowledge mining unveils implicit, previously unknown, and potentially valuable macro-regulatory patterns (Zhang & Yan, 2016; Zhao et al., 2019; Zhu et al., 2020).

4.1. Data pre-processing

4.1.1. Data source

We utilised the PKULAW platform (PKULAW, 2023) to download relevant economic policy documents, laws, and regulations issued by both the country and Guangdong Province during the epidemic period.

4.1.2. Policy text acquisition

The policy text data are collected from December 2019 to June 2021. To ensure comprehensive coverage, we conducted a targeted search using keywords such as “epidemic prevention and control”, “resumption of work and production”, “industrial chain supply chain”, “foreign trade” and “foreign investment”. This approach allowed us to gather a diverse range of policy texts that were directly related to our research problem.

After retrieving the documents, we carefully reviewed each file to exclude those that were “obviously inconsistent with the research theme” or “more general in content”. This step ensured that we focused on texts that provided valuable insights into the economic policies implemented during the epidemic period. The remaining collection of files became our data source for further analysis, including social network and semantic web analysis.

In total, we collected 260 policy texts, with national documents accounting for 39.2% of the dataset, and provincial documents accounting for 60.8%. This balanced distribution allowed us to capture both the overarching national policies and the specific regional measures implemented in Guangdong Province.

Furthermore, in order to enhance our analysis and gain a deeper understanding of the policy landscape, we incorporated data perception technologies. These technologies enable us to extract and interpret relevant information from the collected texts using advanced natural language processing and machine learning techniques. By leveraging data perception, we uncover patterns, relationships, and insights that might otherwise remain hidden, thus enriching our analysis of the policy documents.

4.1.3. Pre-processing: keyword discovery, extraction, and classification

In this step, we introduce the process of keyword discovery, extraction, and classification. We identify keywords that are frequently mentioned together and calculate their co-occurrence frequency based on a social network algorithm. With the extracted information, we perform word frequency statistics to determine the importance of each keyword. Additionally, we utilise a logic-based approach to discover new words.

For a given keyword “m” with a frequency of “x_m” and another keyword “n” with a frequency of “x_n” we examine the co-occurrence frequency of the words in the vocabulary. If the ratio of the co-occurrence frequency between the words and the frequency of “m” is greater than or equal to 50% (x_mn/x_m ≥ 50%) and the ratio of the co-occurrence frequency between the words and the frequency of “n” is also greater than or equal to 50% (x_mn/x_n ≥ 50%), then the word is considered as a new word and added to the thesaurus. This approach allows us to identify and include new words that exhibit a significant co-occurrence pattern with existing keywords, based on their respective frequencies.

After extracting the keywords, we proceed to filter and screen the text by removing high-frequency phrases of prepositions and adverbs commonly found in policy texts. Examples of these phrases include “according to” and “based on.” We also filter out repetitive or redundant words such as “epidemic” and “policy” which were already included as search terms. This filtering process helps refine the keyword list and improve its relevance.

In the classification stage, we apply ideographic principles to categorise the keywords. Keywords sharing similar meanings are consolidated into distinct categories. For example, terms like “imports” “exports” and “imports and exports” could be grouped together within a category related to international trade. Similarly, keywords such as “traffic” and “transportation” might fall under a category associated with transportation infrastructure. This categorisation effectively organises the keywords and facilitates the identification of prevalent themes.

Furthermore, we leverage data perception technologies and computational techniques to enhance the keyword extraction and classification process. Natural language processing algorithms and machine learning models enable us to analyse the text data at scale, identify patterns, and uncover hidden relationships between keywords. These computational techniques provide a deeper understanding of the underlying themes and trends present in the policy texts, aiding in the identification of key focal points and emerging patterns.

4.2. Construction of semantic analysis model

4.2.1. Build the co-occurrence matrix

To analyse the relationships between keywords, we construct a co-occurrence matrix using the following formula: $O{c}_{k,j}=\frac{{\sum }_i\sqrt{Z_{\mathrm{ik}}{Z}_{\mathrm{ij}}}}{\sqrt{{\sum }_i{Z}_{\mathrm{ik}}}\times \sqrt{{\sum }_i{Z}_{\mathrm{ij}}}}$

• $O{c}_{k,j}$ represents the co-occurrence probability between two keywords, “k” and “j”. It quantifies how often these keywords appear together in a given set of documents.

• $\sum _i\sqrt{Z_{\mathrm{ik}}{Z}_{\mathrm{ij}}}$ is the numerator of the formula. Here, $Z_{\mathrm{ik}}$ represents the number of occurrences of keyword “k” in document i, and $Z_{\mathrm{ij}}$ represents the number of occurrences of keyword “j” in document i. By summing the square roots of the product of these occurrences across all documents, we capture the overall co-occurrence between the two keywords.

• $\sqrt{\sum _i{Z}_{\mathrm{ik}}}$ calculates the square root of the sum of occurrences of keyword “k” across all documents. This term represents the normalisation factor for the co-occurrence probability.

• $\sqrt{\sum _i{Z}_{\mathrm{ij}}}$ calculates the square root of the sum of occurrences of the keyword “j” across all documents. Similarly, this term serves as a normalisation factor.

Dividing the numerator by the product of the two normalisation factors normalises the co-occurrence probability. It ensures that the values fall within a range of 0–1, where 0 indicates no co-occurrence, and 1 represents a strong co-occurrence relationship between the keywords.

By constructing a co-occurrence matrix using this formula, which can analyse the relationships between keywords based on their co-occurrence patterns in the given set of documents. This matrix provides a quantitative measure of the strength of these relationships, allowing the identification of keywords that frequently appear together and potentially uncover underlying associations or themes. Since this study categorises policies into three stages for research purposes, it is essential to generate separate co-occurrence matrices for each stage. By analysing the co-occurrence probabilities within each stage, we understand the relationships and associations between different keywords within that particular time frame.

The resulting co-occurrence matrices are presented in Table 2, which provides a visual representation of the probabilities of co-occurrence between different keywords within each stage. These matrices serve as valuable resources for analysing the interconnectedness and semantic relationships between keywords, enabling us to identify clusters of related terms and gain insights into the overall structure of the policy discourse.

Table 2. Part of the intercepted overall policy text co-occurrence matrix.

Serial number	Enterprise	Service	Policy	Reform	Regulatory	Innovation	Cross-border Ecommerce	Market	Mechanism	Information	Employment	Platform
Enterprise	1880	726	549	382	325	440	282	309	356	361	221	429
Service	726	1218	523	429	347	505	154	308	435	404	210	416
Policy	529	523	770	367	245	329	62	244	307	260	216	284
Reform	382	429	367	732	279	312	20	189	277	206	106	213
Regulatory	325	347	245	279	684	238	71	211	265	292	44	223
Innovation	440	505	329	312	238	659	88	212	334	251	90	319
Cross-border Ecommerce	282	154	62	20	71	88	579	33	48	59	3	117
Market	309	308	244	189	211	212	33	557	206	224	62	223
Mechanism	356	435	307	277	265	334	48	206	551	301	82	250
Information	301	404	260	206	292	251	59	224	301	528	64	268
Employment	221	210	216	106	44	90	3	62	82	64	508	85
Platform	429	416	284	213	223	319	117	223	250	268	85	486

To generate these matrices, we leverage data perception and computational techniques. Advanced algorithms and statistical models are applied to efficiently process the large volume of data and calculate the co-occurrence probabilities. By utilising data perception technologies, we enhance our ability to uncover complex patterns and relationships present in policy texts, facilitating a more comprehensive analysis of the keyword associations.

4.2.2. Form the semantic network

To visually depict the relationships between keywords and enhance our comprehension of the pandemic policy landscape, we employ UCINET6 software to facilitate the connection of the co-occurrence matrix and the construction of a semantic network graph.

By utilising UCINET6, we analyse the co-occurrence matrix, producing a semantic network graph that visually illustrates keyword connections. This graph offers valuable insights into the internal relationships and distribution patterns of pandemic-related policies. It enables the identification of closely related keyword clusters and the exploration of the overall structure of policy discourse.

Furthermore, in the process of forming the semantic network, we incorporate data perception technologies and computational techniques. These technologies enable us to handle the complexity of the co-occurrence matrix, efficiently process the data, and extract meaningful insights. By leveraging advanced algorithms and graph analysis techniques, we uncover hidden patterns, detect influential keywords, and identify the central themes within the policy landscape.

4.3. Model result analysis

In this part, we employ UCINET6 social network software to visually represent the co-word matrix as a semantic network graph. By utilising this approach, we aim to discover and summarise the internal relationships and distribution patterns of policies issued by China throughout three distinct stages. The semantic network graph provides a visual representation that helps uncover important connections and patterns within the policy landscape, shedding light on the evolving nature of policy responses during the specified periods.

1. SMEs are the key targets for deployment; the division of labour and cooperation has become an important success factor in the fight against the epidemic; employment is the focus of continuous attention, and informatization has also been highlighted.

Figure 5 illustrates that micro-enterprises hold a central position in the network diagram, surrounded by keywords such as “organisation”, “system” and “government.” This reflects the emphasis on government-enterprise cooperation during the epidemic and the recognition of the importance of division of labour and cooperation. The government has encouraged policies that support micro-enterprises to overcome challenges and survive during difficult times. On the other hand, private enterprises and medium-sized enterprises appear on the periphery of the network map, indicating that although the policy system may not be comprehensive enough for these enterprises, their development still receives special attention from policymakers.

Figure 5. Overall Semantic Network of Three Stages.

Furthermore, the utilisation of data perception technologies and computational analysis techniques facilitates a comprehensive understanding of the semantic network. These technologies enable us to identify the central position of micro-enterprises in the network and the surrounding keywords that signify the importance of division of labour and cooperation. The analysis reveals the nuanced relationships and patterns within the policy landscape, shedding light on the specific focus areas of policy responses during the epidemic.

As shown in Figure 6, keywords linked to import, export, and trade hold prominent positions in both the TF-IDF graph and in proximity to the term “customs” in the semantic network, which underscores the central role of customs in the resumption of import and export activities. Additionally, the analysis emphasises the importance of addressing employment concerns through continuous policy measures, particularly crucial amid widespread layoffs and financial strain experienced by individuals during the crisis.

Figure 6. TF-IDF Distribution Diagram of Full-Time Policy.

In addition, the analysis uncovers the repeated mention of informatization tools, policies, and platforms. These keywords are located at the centre of the network, reflecting the recognition of the importance of leveraging technology and digital solutions for effective policy implementation. By employing data perception techniques, we gain insights into the centrality of these keywords and their interconnectedness with other policy-related terms. This knowledge enhances our understanding of the emphasis placed on informatization and its role in ensuring stability and operational continuity during the epidemic.

Through the integration of data perception technologies and computational analysis, we obtain a comprehensive view of the policy landscape. This approach enables us to uncover important connections, patterns, and emphases within the policies, providing valuable insights for decision-making and further research. By utilising innovative computational techniques, we gain a deeper understanding of the policy responses and their underlying dynamics, ultimately contributing to effective policy formulation and implementation.

1. Transportation, medical care, integrated circuits, and other industries are the key fields of concern. Technology and innovation industries are still key directions.

The transportation, medical care, and integrated circuits industries, among others, are positioned on the periphery of the semantic network graph, indicating their significance as key policy focuses during the development of the epidemic. This aligns with the original intent of policies aimed at ensuring the transportation of essential materials (Xia et al., 2023) and the supply of medical resources during the crisis. It reflects the comprehensive and forward-looking nature of policy planning.

Furthermore, the global trade disruptions experienced during the pandemic have heightened the risk of instability or even disruption in China’s industrial chain and supply chain. This has exposed the vulnerabilities associated with China’s heavy reliance on imported intermediate products. Consequently, keywords such as “technology” and “innovation” frequently appear in policy documents, highlighting the attention given to fostering technological advancements. The integrated circuit industry, in particular, has garnered significant attention, as innovation remains a crucial factor for China’s industrial upgrading and resilience in the face of challenges.

As technology continues to play a pivotal role in various industries, there is a growing emphasis on leveraging data-driven computer technologies to drive innovation and address emerging challenges. The advent of advanced data analytics, artificial intelligence, and machine learning has provided new opportunities for industries like transportation, medical care, and integrated circuits. These technologies enable the collection, analysis, and interpretation of vast amounts of data, leading to improved efficiency, decision-making, and resource allocation.

In conclusion, the transportation, medical care, and integrated circuits industries are key fields of concern, and technology and innovation industries remain crucial directions for policy planning. The integration of data-driven computer technologies, such as advanced analytics, artificial intelligence, and machine learning, plays a significant role in driving innovation and addressing challenges in these industries. By leveraging these technologies, industries enhance efficiency, decision-making, and resource allocation, ultimately contributing to their growth and resilience in the face of evolving global challenges.

Policy analysis during the shock period (1) Most of the attributes of nodes are those affected by the epidemic; the key support policies have been released successively.

The impact of the pandemic has been widespread, affecting various industries, production chains, and subsequently enterprises and employees. The role of policies during the shock period is to guide the epidemic prevention and control mechanism and process promptly and identify and support key frustrated industries and key frustrated groups.

Figure 7 illustrates the policy focus during this period, which encompassed areas such as prevention and control measures, the functioning of the market economy, employment, resumption of work and production, trade, logistics, and technological advancements. These issues had significant implications for economic development and social stability amidst the challenges posed by the pandemic. The release of key support policies during the shock period was crucial in mitigating the impact of the epidemic on various sectors. These policies aimed to provide financial support, promote innovation, and ensure the smooth operation of the market economy. They encompassed measures such as tax breaks, financial assistance programmes, subsidies for affected industries, and incentives for technological advancements.

Figure 7. Semantic Network of the Shock Period.

The focus on prevention and control measures was of utmost importance during this period. Policies were designed to strengthen the healthcare system, enhance epidemic prevention and control capabilities, and ensure the safety and well-being of the population. This included measures such as funding for medical research and development, the procurement of medical supplies, and the establishment of quarantine facilities. The functioning of the market economy was another key area of policy focus. Measures were implemented to stabilise the financial market, stimulate consumer spending, and support small and medium-sized enterprises (SMEs). This included interest rate cuts, liquidity injections, and the implementation of targeted support programmes for affected businesses. Employment was a critical concern during the shock period, given the widespread job losses and economic uncertainty. Policies were introduced to protect jobs, provide unemployment benefits, and support job training and placement programmes. These measures aimed to alleviate the financial burden on individuals and ensure a smooth transition back into the workforce.

The resumption of work and production was a priority to restore economic activity. Policies were implemented to facilitate the safe reopening of businesses, ensure the availability of necessary resources and supplies, and promote the efficient operation of production chains. This included measures such as the co-ordination of logistics and transportation, the facilitation of cross-border trade, and the implementation of health and safety protocols in workplaces.

In summary, the release of key support policies during the shock period was essential in addressing the challenges posed by the pandemic. These policies focused on areas such as prevention and control, the functioning of the market economy, employment, resumption of work and production, trade, logistics, and technological advancements. By providing targeted support and guidance, these policies played a crucial role in minimising the impact of the epidemic and facilitating a smoother recovery process.

1. The policy is comprehensive without losing details.

The peripheral area of the semantic network diagram represents the policy’s main areas of focus. These include measures aimed at preventing and controlling the epidemic, ensuring smooth material transportation, addressing unemployment concerns, and promoting technological advancements in industries such as integrated circuits, chips, and semiconductors. The central part of the graph highlights the policy’s emphasis on key entities such as markets, companies, projects, and institutions. Analysing the keywords surrounding small and medium-sized enterprises (SMEs) in the upper right corner reveals that policies implemented during the crisis period primarily concentrate on securing the cash flow of SMEs and supporting the overall industrial chain. Measures such as financing, funds, subsidies, and other resources are implemented to alleviate operational pressures faced by enterprises. Keywords associated with customs include places, products, safety, employees, risks, and enterprises. Examining the policy’s role in import and export protection, it becomes evident that it extends to various critical entities, with a focus on managing risks and ensuring business continuity during the epidemic.

Policy analysis in a stable period (1) The stable period expands the scope of attention and refines the stabilisation and control tools during the shock period.

During the stable period, as depicted in Figure 8 of the semantic network diagram, new terms and connections emerge, such as the industrial chain, supply chain, medical care, finance, tourism, and more. This indicates that the policies implemented during this phase begin to prioritise the development of the industrial and supply chains, as well as focus on improving the medical system and ensuring a robust guaranteed mechanism. The expansion of the semantic network’s scope indicates that more subjects are being influenced by these policies, showcasing their increasing strength and effectiveness in achieving “group control”. Given that this is one of the most significant public health emergencies that China has faced in recent years, the semantic network also reflects the country’s innovative response. For instance, keywords associated with

Figure 8. Semantic Network of the Stable Period.

innovation demonstrate new perspectives, discoveries, and advancements in models, technologies, and trading methods. The term “platform” also appears in the network, highlighting how Guangdong Province has been able to swiftly respond and exercise precise control in epidemic prevention and control through the assistance of national government platforms such as “JianKangBao” and the government cloud.

1. Words such as production, mechanism, regulation, and risk became the core keywords instead of market and service, and the policy force was shifted from internal to external.

During a challenging period, the market experienced significant damage. Through the relentless efforts of various stakeholders, including employees and enterprises, the system was eventually restored. The government played a crucial role in macro-control, focusing on key prevention and control measures, and promptly addressing the impact of the epidemic on our market economy and people’s well-being. As stability returned, the government gradually reduced its policy interventions, allowing internal forces to transition into external forces. This approach ensured the continued development of the market economy and facilitated import and export activities by emphasising the establishment of effective mechanisms, robust supervision, and risk control perspectives.

In the subsequent stable period, the influence of policy measures gradually waned, marking a shift from internal to external forces. The focus turned towards establishing efficient mechanisms, implementing effective supervision, and managing risks to sustain the development of the market economy while facilitating import and export activities. The primary objective was to maintain stability and ensure the ongoing progress of the economy.

Overall, the policies implemented during the stable period reflected a comprehensive and adaptive response to the challenges posed by the pandemic. They focused on multiple aspects such as industrial development, healthcare, innovation, and risk management, aiming to effectively control the epidemic, stabilise the economy, and promote long-term resilience and sustainability.

Overall policy analysis during the boost period (1) Quarantine, health, and other epidemic prevention and control measures become the focus of continued policy attention as the epidemic stabilises.

According to Figure 9, as the epidemic stabilises, people’s daily lives gradually return to normal, leading to an overall boost in the social economy. During this phase, policy attention shifts towards epidemic prevention and control measures, particularly in industries such as tourism. The unique nature of the epidemic and the severity of the global situation make it inevitable to normalise epidemic prevention and control efforts (Gu et al., 2022). As the epidemic stabilises and society returns to a state of normalcy, the challenges associated with prevention and control increase. Therefore, it concluded that quarantine measures, health-care initiatives, and other epidemic prevention and control efforts will remain the focal point of policies during the boost period and will continue to receive attention in the future.

1. Measures such as tax reduction and exemption continue to be promoted, and the direction of high-quality development is firmly established.

Figure 9. Semantic Network of the Boost Period.

The quantitative analysis conducted in this study focuses on import and export policies, revealing that China’s policy approach during the boost period primarily centres around implementing tax reduction and exemption measures. Furthermore, there is an increased emphasis on enhancing tax and fee reduction efforts to alleviate the financial burden on export and foreign trade enterprises, enabling them to overcome challenges effectively. Moreover, keywords that signify the direction of economic development, such as high-quality development and high-level development, appear frequently throughout both the stable and boost periods. While the outbreak of the epidemic had a significant impact on China’s economy, it also presented an opportunity for the country to break free from technological constraints imposed by developed nations and undertake initiatives to reshape the supply chain and industrial chain. Following a smooth transition from the epidemic, policies began to focus on further economic transformation and solidifying the direction of the economy towards achieving high-quality and high-level development (Xu et al., 2020).

In conclusion, the policies implemented during the boost period prioritise epidemic prevention and control measures, tax reduction and exemption, and the pursuit of high-quality development. These policies reflect the government’s commitment to public health, supporting businesses, and driving sustainable and transformative economic growth.

5. Policy subject analysis based on semantic analysis

In this section, we will leverage a social network model to analyse policy-issuing subjects, with a focus on utilising data-driven computing technologies. By applying advanced computational techniques, we extract valuable insights from policy-related data and establish connections among various subjects involved in policy implementation. The social network model serves as a powerful tool for visualising and understanding the relationships among these subjects. It represents policy-issuing subjects as nodes and their connections as edges, allowing us to gain a comprehensive overview of the collaboration patterns and information flow within the policy ecosystem (Qiu et al., 2020; Yan et al., 2022). This approach is particularly valuable in the context of data-driven computing technologies, as it enables us to identify key entities and their interactions, thus facilitating effective decision-making and policy formulation. To enhance our analysis, we integrate statistical analysis with the social network model. By quantitatively measuring the strength and frequency of collaborations among different departments and entities, we identify the most influential and collaborative policy subjects at a micro level. These subjects demonstrate a high level of engagement and contribute to the breadth and depth of policy development and implementation.

Furthermore, the use of data-driven computing technologies allows us to explore macro-level evolutionary trends among policy-issuing subjects. By analysing temporal patterns and changes in collaboration networks over time, we identify emerging trends, evolving relationships, and potential shifts in policy priorities. This information provides valuable insights for policymakers and stake-holders in adapting their strategies and interventions to align with the dynamic policy landscape.

5.1. Analysis of policy subjects relationship based on social network model

• Step1: Construction of policy subject cooperation network

In this part, we extract the policy-issuing subjects from the collected policy documents. The collaborative efforts of multiple departments in issuing epidemic response policies are considered cooperative relationships between the policy-issuing subjects. Using this information, we construct a cooperation matrix that represents the relationships between the policy-issuing subjects. Subsequently, we utilise Gephi software to visualise the cooperation network diagram among the policy-issuing subjects. This diagram provides a visual representation of the collaborative relationships between the various policy-issuing subjects.

• Step2: Structural feature identification of policy subject collaboration network

In this part, we employ social network analysis to identify the overall collaboration concentration, as well as the breadth and depth of cooperation of policy subjects. The specific indicators used for identification are presented in Table 3.

Table 3. Identification index system of network structure characteristics of policy subject collaboration.

Dimensionality	Indicators	Definition
Collaboration Concentration	Network Density	Reflects the average level of interaction of collaboration and communication among policy subjects
	Network Modularity	Reflects the strength of cohesion among policy subjects
	Average Side Distance	Reflects the efficiency of information transfer between policy subjects in the overall network
Breadth of single department collaboration	Degree Center Degree	Number of policy subjects reflecting direct cooperation
	Closeness Centrality	The distance between the reaction and other policy subjects in the network and the efficiency of policy information transmission
Single-sector collaboration intensity	Eigenvector Centrality	Network depth and location advantages of policy
	The Ratio of Jointly Issued Policies to Cooperating Sectors	Reflects the extent of deep collaboration with other policy actors

5.2. Analysis of model results

In this paper, 260 texts were collected, including 38 policy subjects, of which 36 subjects have collaborative relationships, forming 291 cooperative edges. The social network formed is shown in the following Figure 10.

1. There is a close centripetal force between policy release subjects, but the breadth and depth of collaboration need to be optimised.

Figure 10. Policy Publisher Social Network.

The calculated indicators for overall collaboration concentration reveal the following results: the network density (Newman, 2018) of the social network is 0.414 (range from 0 to 1, bigger means more dense). The network modularity (Newman, 2006) is 0.126. The range of values for modularity is typically between −1 and 1, with larger values indicating a stronger community structure in the network. The average edge distance of the social network is 1.730, and the standard deviation of the number of joint publications of each policy issuer is 23.

From the perspective of overall cooperation concentration, 94.74% of the policy release subjects have collaborative relationships with each other. This indicates that after the outbreak of the epidemic, various organs and institutions in China have demonstrated a unifying force to work together, leveraging a “1 + 1 > 2” effect to control the epidemic, stabilise the economy, and ensure the well-being of the people.

However, the calculation results also indicate that the relationships between policy-issuing entities are relatively loose, with a low level of interaction. On average, each pair of policy release subjects has 1.73 collaborative publications. Nevertheless, the large standard deviation of the number of joint documents issued by each policy release subject suggests that some subjects within the network exhibit higher collaboration density than the overall network density. This implies that there are significant variations in the breadth and depth of collaboration among different policy sectors.

1. National departments are at the centre of policy coordination, and single-sector collaboration has an obvious partial aggregation phenomenon.

The single-sector collaboration breadth and intensity indicators were computed for each policy publisher. The top 20 policy publishers with the highest number of joint publications were selected, and the corresponding calculation results for their related indicators are presented in Table 4.

Table 4. Performance of single-sector collaboration breadth and intensity indicators (top 20 joint publications).

Serial number	Organization name	Center degree	Closeness Centrality	Eigenvector centrality	Ratio of Joint to cooperative
1	Ministry of Finance	29	0.795	0.997	3.111
2	Ministry of Commerce	29	0.795	1	2.741
3	State Administration for Market Regulation	28	0.761	0.98	2.769
4	Ministry of Industry and Information Technology	26	0.729	0.959	2.833
5	People's Bank of China	27	0.745	0.97	2.32
6	Ministry of Human Resources and Social Security	24	0.714	0.904	2.591
7	Ministry of Transport	25	0.729	0.861	1.913
8	National Development and Reform Commission	24	0.714	0.878	1.955
9	China Insurance Regulatory Commission	18	0.625	0.67	2.375
10	Ministry of Public Security	23	0.686	0.818	1.81
11	Ministry of Urban-Rural Development	22	0.673	0.842	1.9
12	Ministry of Tourism	23	0.686	0.859	1.762
13	General Administration of Customs	19	0.614	0.689	1.941
14	State Taxation Administration of China	13	0.574	0.521	2.727
15	Ministry of Education	18	0.636	0.696	1.75
16	Ministry of Agriculture and Rural Affairs	19	0.636	0.724	1.412
17	Central Cyberspace Administration	15	0.593	0.602	1.538
18	Assets Supervision and Administration Commission	16	0.603	0.62	1.214
19	Ministry of Justice	15	0.574	0.58	1.308
20	Ministry of Science and Technology of China	15	0.593	0.635	1.154

The single-sector collaboration breadth and intensity indicators were computed for each policy publisher. The top 20 policy publishers with the highest number of joint publications were selected, and the corresponding calculation results for their related indicators are presented in Table 4.

From the perspective of the breadth and intensity of single-sector cooperation, significant differences exist among different departments, indicating the presence of “local aggregation” phenomena. The social network diagram provides a more visual representation, clearly highlighting the close collaborative relationships among policy release subjects such as the Ministry of Finance, the Ministry of Commerce, the State Administration for Market Regulation, the Ministry of Industry and Information Technology, the People’s Bank of China, and the Ministry of Human Resources and Social Security. These subjects occupy a central position in the diagram, indicating their crucial roles in foreign trade mediation and overall economic recovery. They have engaged in comprehensive, multi-level, and extensive cooperation with other policy release subjects. These departments have worked collectively to stabilise the cash flow, employment, information transmission, and product supervision for import and export enterprises.

Furthermore, their collaboration breadth and intensity index scores are high. Among them, the Ministry of Finance and the Ministry of Commerce exhibit prominent collaboration breadth, engaging with the largest number of departments and covering a wide range of collaborative publications. Additionally, they have issued a significant number of independent policies, indicating their substantial influence on stabilising foreign trade. On the other hand, the Ministry of Finance and the Ministry of Industry and Information Technology demonstrate notable collaboration intensity, as they have issued the highest number of collaborative documents within their respective units. They exhibit a dense and frequent collaboration pattern within their departments, showcasing strong sustainability. Consequently, it concluded that the aforementioned policy release subjects, particularly the Ministry of Finance, are core subjects within the co-operation network and play a pivotal role in stabilising foreign trade during the epidemic period.

Considering the social network diagram, it is evident that the main departments with lower scores in terms of breadth and depth of single-sector collaboration have limited collaboration with other major departments. However, each department has its own reasons for this pattern. For instance, although the Office of the National Health Commission and the Department of Emergency Management play crucial roles in overall epidemic prevention and control, their collaboration frequency in issuing economic policies might be comparatively low, placing them at the periphery of the diagram. Similarly, the State Council and the General Office of the State Council exhibit less collaboration with other departments. Upon analysing the policy texts they issue, it becomes apparent that their focus is primarily on macro-level regulation and control, providing guidance and direction for other departments to formulate policies. Consequently, their collaboration frequency is relatively lower. The policies issued by provincial-level units in Guangdong are often forwarded by higher-level government departments and tend to be targeted direct control policies. Therefore, their collaboration frequency is also not as high. Despite this, these policy subjects still play a significant role in stabilising foreign trade.

Considering the breadth and depth of cooperation with other policy release subjects, there is an opportunity to further develop policy cooperation space and improve policy cooperation mechanisms. This could enhance collaboration between different departments and promote more effective and coordinated policy responses.

6. Discussion

6.1. Policy recommendation

In practical terms, we recommend two vital optimisation strategies for economic policymaking during the pandemic. First, improving the connection between economic data and decision-makers will offer a more holistic understanding of the situation, facilitating effective policy formulation. Second, enhancing coordination among provincial and municipal units is essential for seamless implementation and a unified response.

6.2. Implications

Simultaneously pursuing these strategies will enable the government to proactively address public emergencies and support economic and industrial trade recovery. Through our analysis, we’ve observed a concentrated policy focus during the pandemic recovery phase, particularly on implementing tax reduction and relief measures. The aim is to ease the cost burden on export-oriented enterprises and provide support for their resilience. Notably, in the context of international trade-related economic policies, 94.74% of the entities responsible for policy formulation demonstrate collaborative relationships. This suggests a coordinated effort among different departments to stimulate international trade.

In conclusion, by embracing technology and following these optimisation strategies, the government can enhance its responsiveness to public emergencies and create a favourable environment for economic recovery.

6.3. Limitations and future work

In this paper, we analyse the economic policies of China during the pandemic by integrating the methodologies of semantic analysis and social network analysis. One limitation of our work is that there is some inadequacy in the modelling and learning capabilities for the complex and dynamic multi-hop relationships between different policy contents and entities in long-term scenarios. In future endeavours, we plan to leverage the data management and representation capabilities of knowledge graphs to enhance our ability to model dynamic and intricate relationships. Additionally, we will implement corresponding policy recommendation mechanisms to offer more assistance in future economic policy formulation.

7. Conclusion

This paper employs a comprehensive analysis of policies implemented during the pandemic to identify policy priorities and facilitate economic recovery. We find that the policies implemented during various phases of the pandemic effectively targeted critical aspects of prevention and control. We also identify several challenges of economic policies, including the presence of imperfect emergency plans, inadequate linkage mechanisms, and shortcomings in information and data-sharing systems. Particularly, this paper introduces a groundbreaking text analysis framework that merges social network analysis and semantic analysis algorithms, enabling a deeper examination of the characteristics of policy issuance during the epidemic. By incorporating computer technology, the proposed analysis framework helps policymakers gain more valuable insights and make data-driven decisions.

Ethical approval

As this research did not involve any human or animal subjects, no ethical approval was required for the study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Consent to participate

This research did not involve human participants or animal studies; therefore, no consent to participate was required.

Consent to publish

We have obtained consent to publish from all individuals (including participants, co-authors, and other relevant parties) whose personally identifiable information, images, or any other potentially identifying information is included in this manuscript. We confirm that any such information has been appropriately anonymized or de-identified to protect the privacy and confidentiality of the individuals involved.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Miao Yu, Xing Wan, Tianyou Zhu, Yuyue Wang, Mengdi Xu, Zhenzhen Wu and Xinyu Li. The first draft of the manuscript was written by Miao Yu and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Additional information

Funding

This work was supported in part by the Social Science Project (20ZFG63001) from China University of the Political Science and Law, in part by the National Natural Science Foundation of China, grant number 72273152 and the Qian Duan-sheng Distinguished Scholars Program from China University of the Political Science and Law.