Neighborhood Characteristics in Sejong City: Text Mining Analysis of Civil Complaints

1. Text mining analysis

Text mining is an approach where new information is automatically extracted from different written materials using computers (Fan et al., 2006). In technical terms, text mining is a technique that extracts specific keywords or topics from character strings, sorts documents into groups of similar types, and identifies the correlation between the extracted keywords and sorted documents. It is difficult for computers to understand natural language. In this regard, text mining has an advantage that it can convert the written and spoken language of humans, which is unstructured data, into core keywords. This allows users to come up with quantitative values, such as the frequency of each keyword. The recent rapid advances in communications technology and platform and interface applications have been promoting the generation and spread of a large amount of text data throughout the internet space. Furthermore, the government and the business sector have started using this approach, thereby expanding its influence beyond individuals and toward the industry across the board. In doing so, text mining is becoming increasingly important, and the relevant analysis techniques are advancing day by day. In the business management community, a crowdsourcing technique has been introduced as a new approach, in which ordinary citizens attempt to collect opinions to solve a given specific problem. Similarly, in the planning field, there has been a need to introduce a new way of promoting resident participation, like crowdsourcing (Brabham, 2009). In practice, in planning and other policy fields, new ways of satisfaction surveys or public opinion polls are being attempted. This is because conventional surveys, polls, forums, committees, councils, round-table conferences, and discussion sessions are likely to be limited in the number of participants and the ability to represent the respective target groups. Also, these approaches are implemented less frequently than needed, and thus it is difficult to read ever-changing public opinions in a timely manner (Hwang, M.H., 2014). According to the study conducted by Hong, S.G. et al. (2017), it is now time to implement more intelligent governance built upon cutting-edge technologies, and it will be possible to formulate policies in a more specific and individualized manner by attempting to identify various microscopic needs associated with regional issues on our own, especially through big data analysis applications. The overall new town planning process can also be implemented based on new approaches, such as web-based resident participation or civil complaint analysis techniques using large-scale text data available on the web. In the past, it was difficult to collect opinions from the public, but new approaches are now available, which allow for the analysis of a large volume of documents or complaints on the Internet. Using them, it will be possible to find out what citizens want with regard to new town planning in a more specific manner. These approaches are similar to what is conducted in ethnography, one of the qualitative research methods, but have the advantage that it is also possible to address not only “what” but also “why” questions (Song, K.Y., 2012).

2. Review of previous studies

In domestic academia, research using text mining has continuously been on the rise since 2011 (Kim, S.G. et al., 2016). Good examples are studies related to product reviews and tourist destination reviews. In studies on residential environment and urban space reviews, text mining is mainly used for analyzing resident satisfaction or demand. Park, J.H. and Kang Y.O. (2014) used tweet data to analyze gu-specific resident satisfaction in Seoul and then compared the results with those obtained through surveys. Hwang, M.H. (2014) proposed an unstructured big data-based spatiotemporal analysis model capable of public opinion monitoring as an approach to explore and formulate policies tailored to the target region and demand. Hong, S.G. et al. (2017) proposed several models for reference as a way to settle conflicts arising from the implementation of local public work projects. In an attempt to understand the emotions and attitudes of the persons concerned and citizens, they collected relevant civil complaint data and posts on different online communities and further performed text mining of them. Park, S.H. and Lee, H.C. (2018) determined that social networking services had a high potential for analyzing the behavior of visitors to markets. In this respect, they collected social networking data related to the two markets of their interest and compared the respective results obtained from the two markets to identify what had caused them to visit those markets. Lee, S.H. and Son, Y.H. (2018) conducted a comprehensive review of blog posts related to the landscapes and actual images of the Taean Coast National Park and different activities available in the place. As a result, perceptions about the park and major relevant issues could be identified. Park, J.S. and Lee, J.S. (2019) analyzed online newspaper articles and demonstrated that the positive sentiment index of those newspaper articles affected the sales price of small apartment houses.

In the meantime, research using text mining to analyze civil complaints is also on the rise. Sung, B.K. and You, Y.Y. (2018) analyzed the civil complaint and policy bulletin boards of the Ministry of Employment and Labor as part of e-People and thereupon identified the major trends of civil complaints and further submitted proposals to improve the relevant schemes and systems. Another study collected and examined environmental complaints and analyzed their spatial patterns (Na, Y.W. et al., 2015) and Kim, H.J. et al. (2018) performed a two-step correlation analysis of posts available on the civil complaint board of Busan City in an attempt to ascertain what the citizens required in a more accurate manner than in previous studies.

The present study applied various analytical techniques used in previous studies to identify the core aspects of the spatial information and civil complaints contained in various posts. However, the present study differed from others in that resident and environmental complaints from new towns, which was rarely the topic of those previous studies, were analyzed, and as a result, their neighborhood-specific characteristics could be identified.

1. Target research sites

The spatial scope of the present study covers 12 administrative neighborhoods (dong) within the Sejong City. Currently, this zone is home to 249,788 of residents, which is half the planned number of the population (0.5 million), and the total number of residents in Sejong City amounts to 344,476 (according to Sejong Statistics, as of late November 2019). The total area of Sejong Metropolitan Autonomous City as an administrative district amounts to 464.84 km², and the area of Multifunctional Administrative City, which is currently under development, is 72.91 km². The present study focused mainly on Neighborhoods 1, 2, and 3, where construction had been completed, and most citizens had already moved in Table 1 and Figure 1.

Table 1.

Site

Figure 1.

Research area

In Sejong City, neighborhood units are uniformly distributed along its annular urban space structure (Ministry of Land, Infrastructure and Transport, 2006). Its center remains empty as an open space, and circular roads for public transportation are designed to connect between its main neighborhoods. The city is a realization of an urban space built on combined concepts of decentralization, uniformity, and democracy (Kwon, Y.S., 2009). Public transportation constitutes a green traffic system that connects pedestrian walkways, bike lanes, and green networks. Also, along the lower part of its residential areas, commercial and community facilities were arranged to build neighborhood streets and thus create a pedestrian-centered environment. At the center of the neighborhoods, complex community centers were built to provide further convenience for the residents. Also, throughout the entire city area, a waste pipeline transportation system was established, and in its proximity, mechanical biological treatment facilities, and combined heat power plants were installed and connected together. This meant that water, waste, and energy were connected and circulated in an organized and mutual manner from an environmental perspective (National Agency for Administrative City Construction, 2006).

2. Analyzed materials

In the present study, civil complaints posted on the official website of the Sejong City were collected and analyzed. A total of 7,506 posts were collected from the civil complaint consultation board, which is currently not in service, and the public complaint board under Citizens’ Window. From the civil complaint consultation board, a total of 1,907 posts uploaded after the residents started to move into the target areas were collected (those posted from January 03, 2012 - March 29, 2017). A total of 5,599 posts were collected among those made available to the public on the board under Citizens’ Window (those posted from May 04, 2015 - December 17, 2019). These two civil complaint boards are places where the citizens can post their civil complaints, suggestions, inquiries, or opinions, and, subsequently, the persons in charge provide answers to them. Citizens must register as members to post content here.

The data collection was conducted on two occasions from May 1 to 3, 2019, and from December 16 to 17, 2019. Posts were automatically collected using a web crawler written in Python. This program is designed to access HTML pages and extract only the necessary parts by using libraries, such as Selenimum and BeautifulSoup.

3. Analysis methods

1. Results of civil complaint keyword analysis

According to the keyword analysis results, the top 50 keywords are as follows in the order of rankings from high to low: “installation,” “bus,” “road,” “apartment complex,” “apartment,” “traffic,” “management,” “safety,” “children,” “Hansol-dong,” “accident,” “commute,” “bus service,” “Daejeon,” “construction,” “Goun-dong,” “route,” “bicycle,” “environment,” “public transport,” “illegal,” “parking,” “regulation,” “signal,” “store,” “bus stop,” “Areum-dong,” “pass,” “garbage,” “facilities,” “school,” “park,” “construction,” “parking lot,” “building,” “Dodam-dong,” “exercise,” “crosswalk,” “cleaning,” “sidewalk,” “walk,” “city bus,” “noise,” “Jongchon-dong,” “smell,” “left turn,” “Saerom-dong,” “taxi,” “lake park,” and “lane” (Table 2). Keywords, such as “installation,” “road,” “apartment complex,” “apartment,” and “safety,” were considered to be related to the infrastructure. The term “installation” was not the direct content of civil complaints about urban planning and had high redundancy with several keywords, such as “safety,” “signal,” “bus stop,” and “crosswalk.” However, the term was included in the keyword analysis results because it was considered to provide important information, such as what facilities were lacking given the nature of new town planning and what facilities the citizens were demanding. It was also confirmed that there were many complaints related to public transport, as shown by such keywords as “bus,” “traffic,” “commute,” “bus service,” and “route.” Furthermore, six dongs were frequently mentioned. They can be listed in the order of frequency of occurrence from high to low, as follows: “Hansol-dong,” “Goun-dong,” “Areum-dong,” “Dodam-dong,” “Jongchon-dong,” and “Saerom-dong.” These were relatively old neighborhoods with many apartment houses.

Table 2.

Keyword analysis results

2. Results of correlation analysis of high-ranking keywords

The top ten keywords were analyzed for their correlation to determine what content each of them would be highly correlated to, as shown in Table 3. The term “installation” was found to have a high correlation coefficient with the following words: in order of correlation from strong to weak, “crosswalk,” “safety,” “camera,” “CCTV,” “walking,” “prevention,” “vehicle,” “speed,” “speed bump,” “apartment complex,” “accident,” “road,” “danger,” “traffic,” “cradle,” “signal,” “speed camera,” “bicycle,” and “street.” These were considered to be related to the civil complaints that demanded the installation of facilities for pedestrian safety.

Table 3.

Analysis of related words of the top 10 keywords

The term “bus” was found to have a high correlation coefficient with the following words: “route,” “bus stop,” “bus service,” “public transport,” “city bus,” “time,” “departure,” “commute,” “the interval between buses,” “transfer,” “reorganization,” “transportation,” “boarding,” “arrival,” “via,” “passenger,” “driver,” “stop,” “get off,” “taxi,” “Daejeon,” “establish,” “inconvenience,” “adjustment,” “interval,” “increase,” and “Banseok Station.” Complaints related to public transport issues, such as routes, bus stop positions, and service intervals, were identified.

The term “road” was found to have a high correlation coefficient with the following words: “vehicle,” “traffic,” “accident,” “construction,” “pavement,” “center,” “bicycle,” “sidewalk,” “approach,” “lane,” “section,” “danger,” “regulation,” “walking,” “signal,” “safety,” “prevention,” “roadway,” “driving,” and “construction.” These complaints regarded the safety of vehicles, bicycles, and pedestrians.

The term “apartment complex” was found to have a high correlation coefficient with the following words: “installation,” “store,” “crosswalk,” “noise,” “vehicle,” “signal,” and “illegal parking.” These were considered to be complaints about the pedestrian environment around stores inside the apartment complexes or near the complexes.

The term “apartment” was found to have a high correlation coefficient with the following words: “move in,” “apartment complex,” “lotting-out,” “store,” “households,” “noise,” “surroundings,” “construction,” “construction company,” “flaw,” and “noise.” Many complaints about defects and construction companies were found. This was considered to be because apartment houses accounted for a large portion of Sejong City’s housing units, and there were many sales of apartment units and residents moving in.

The term “traffic” was found to have a high correlation coefficient with the following words: ”public transportation,” “bus,” “commute,” “vehicle,” “route,” “service,” “signal,” “national highway 1,” “inconvenience, “time,” “intersection,” “congestion,” and “Daejeon.” These were found to be complaints about vehicle driving on National Highway 1 or commuting roads in Daejeon, as well as public transport services.

The term “management” was found to have a high correlation coefficient with the following words: “clean,” “poor,” “facility,” “resident representative,” “weeding,” “park,” “weeds,” and “facility management.” These were complaints related to public space and apartment management.

The term “safety” was found to have a high correlation coefficient with the following words: “accident,” “children,” “crosswalk,” “installation,” “danger,” “walking,” “threats,” “child,” “vehicles,” “pedestrians,” “road,” “driving,” “communing,” “school,” “speed,” and “speed bumps.” These are related to requests to install crosswalks and speed bumps for the safety of children walking along the streets.

The term “children” was found to have a high correlation coefficient with the following words: “school,” “safety,” “kindergarten,” “children,” “mom,” “elementary school,” “daycare center,” “communing,” “parents,” “danger,” “commuting to school,” “safety,” “worry,” and “crosswalk.” These were mainly about the safety of children walking to and from school, similar to the correlation analysis results of the keyword “safety.”

The term “Hansol-dong” was found to have a high correlation coefficient with the following words: “bus,” “loop bus,” “city bus,” “route,” “E-Mart,” “Handuri Bridge,” “terminal,” “Hansol Middle School,” and “cogeneration plant.” It was deemed that there were many complaints about the public transport system and schools in Hansol-dong and nearby co-generation power plants.

3. Topic modeling results

Topic modeling was conducted to identify the topics of various civil complaint documents, as shown in Table 4. The main topics of civil complaints in Sejong City were found to include safety facilities, public transport, educational environment, road traffic, parking, and environmental issues. Topic 1 (Safety facilities) includes the following keywords: “installation,” “vehicle,” “safety,” “road,” “crosswalk,” “signal,” “walking,” and “children.” These were complaints asking for the expansion of facilities related to pedestrian safety. Topic 2 (Public transport) includes the following keywords: “bus,” “route,” “apartment complex,” “Goun-dong,” “public transport,” “bus service,” “Hansol-dong,” and “change.” It was assumed that there were relatively many complaints about the public transport system in Hansol-dong and Goun-dong. Topic 3 (Education) includes the following keywords: “children,” “school,” “day care center,” “kindergarten,” “education,” “mom,” “commuting,” and “safety.” Similar to Topic 1 (Safety facilities), keywords related to the safety of children walking to and from school were included. Topic 4 (Road traffic) includes the following keywords: “vehicle,” “road,” “signal,” “commute,” “traffic,” “turn left,” “lane,” and “traffic jam.” These were complaints about inconveniences related to road traffic. Topic 5 (Parking) pertains to parking-related issues. It includes “parking,” “vehicle,” “regulation,” “illegal,” “parking lot,” “road,” “store,” and “apartment complex.” These were complaints about the lack of parking space for vehicles and requests for the implementation of parking enforcement on the road and near stores and complexes. Topic 6 (Environment) regards environment-related issues, including the following keywords: “smell,” “noise,” “apartment complex,” “apartment,” “Hansol-dong,” “bad smell,” “nighttime,” and “environment.”

Table 4.

Topic modeling analysis results

4. Analysis results of neighborhood-specific civil complaints about the residential environment

Lastly, civil complaints about the residential environment of the six neighborhoods, which were among the top 50 keywords, were compared. The analyzed six neighborhoods were Hansol-dong (ranked 10th), Goun-dong (16th), Areum-dong (27th), Dodam-dong (36th), Jongchon-dong (43rd), and Saerom-dong (47th).

All high-ranking keywords were grouped based on similarity into residential environment keywords, as follows: Public transport (search keywords: “bus” and “public transport”), Road (search keyword: “road”), Installation (search keyword: “installation), Education (search keyword: “school” and “children”), Parking (search keyword: “illegal,” “parking,” and “parking lot”), Accident (search keyword: “accident”), Park (search keyword: “park”), and Bad smell (search keywords: “smell” and “bad smell”). From all civil complaints, residential environment-related documents were separated, from which all documents that contained spatial information were extracted.

The analysis results showed that the content of civil complaints varied from one neighborhood to another (Table 5). The number of posts that contained those keywords in each neighborhood was divided by the number of posts in the corresponding neighborhood to compare the neighborhood-specific characteristics of civil complaints, as shown in Table 6 and Figure 2. Notably, when there was a unique set of keywords mentioned for a specific neighborhood, the content of the corresponding posts was directly reviewed to find out the cause of the dissatisfaction. For example, a post that contains “bus” and “bus stop” as keywords may differ from one that contains “park,” “bus,” and “bus stop” in terms of content. In general, there were many complaints asking for the installation of bus stops, but in some neighborhoods, the installation of bus stops was requested because it was difficult to access a park by public transport. For that reason, the content of individual complaints specific to each neighborhood was thoroughly examined before they were classified.

Table 5.

Frequency analysis of complaints by neighborhood

Table 6.

Comparison of neighborhood by residential environment complaints

Figure 2.

Frequency analysis of residential environment complaints by neighborhood

In the case of Hansol-dong, the number of civil complaints was relatively high. When compared to the other neighborhoods, matters related to public transport, parking, road traffic, and bad smell were more frequently mentioned. When it comes to public transport, there were 98 posts regarding route issues, 80 posts regarding the installation of bus stops, and 53 posts regarding service intervals. With regard to parking, the number of posts asking for the implementation of parking enforcement was the largest at 52. Also, there were 35 complaints about the lack of parking space and facilities. Issues related to illegal parking near the shopping area were frequently raised (31 posts). Among complaints related to road traffic as well, parking problems were most frequently mentioned (41 posts). There were also 15 posts requesting the installation of facilities that help improve the pedestrian environment, such as crosswalks. With regard to complaints related to bad smells, there were 13 posts that said that it smelled even worse during certain seasons and time of day, for example, in summer evenings. There were eight complaints that pointed to burnt smell and illegal incineration from co-generation power plants and another eight complaints about bad smell from nearby restaurants, laundries, and outdoor units of air-conditioners along the roadside. Also, there were eight complaints about garbage smell and another eight complaints about smells coming from barns.

In the case of Goun-dong, the number of complaints about facility installation and parks was relatively high. With regard to facility installation, the number of complaints asking for the installation of facilities inside parks, including Gountteul Park, was the highest at 143. Furthermore, there were 29 posts asking for the installation of bus stops and new routes, 20 posts asking for the installation of crosswalks, and 16 posts asking for the implementation of the bike-sharing system. With regard to parks, there were 39 requests for the maintenance of trails, decks, and lighting facilities in Gountteul Park. Also, there were 19 complaints about the lack of public transport to parks, 15 posts requesting the installation of safety facilities near parks, such as crosswalks, and 14 complaints against the installation of pet dog playgrounds. In the case of Goun-dong, complaints related to Gountteul Park were most frequent.

In Areum-dong, there were many complaints related to education. More specifically, those who failed to enroll in nearby elementary and middle schools because they were overpopulated were forced to commute to and from more distant schools, but the lack of public transport added to their inconvenience. There were 18 posts regarding these problems. There were 14 complaints about the safety of school walkways where illegal parking and speeding were common, and crosswalks were not well prepared. Also, there were six posts asking for the installation of pedestrian decks and tunnels and four requests that sports facilities be made available to the citizens at all times because there was a lack of space for children’s athletic activities, such as swimming pools and sports facilities.

In Dodam-dong, there were many complaints about parking and accidents. It was found that illegal parking was frequently, especially in the shopping district of Dodam-dong (17 posts). There were 27 posts asking for the implementation of parking enforcement and 37 posts complaining about the lack of parking space and poor status of parking lot entrances, as well as asking for making public parking lots available to use.

In Saerom-dong, there were many complaints about facility installation and accidents. There were many posts requesting the installation of bus stops and public transport-related facilities (18 posts). In addition, there were 14 complaints about pedestrian facilities, such as crosswalks, pedestrian overpasses, and lighting facilities; nine posts asking for the installation of bicycle rental facilities; five complaints about CCTV installation; another five complaints about the installation of sports facilities, such as swimming pools and basketball courts; and four requests for the maintenance of speed bumps. With regard to accidents, there were nine complaints about the risk of accidents due to illegal parking, placards, occupied sidewalks, and construction works; five complaints about the risk of pedestrian accidents due to motorcycle messengers’ reckless driving; and five complaints about the failure of the signal system and traffic lights and the danger associated with roundabouts. Among the six neighborhoods, this one was most recently developed, and thus understandably, there were more complaints requesting the adjustment of public facilities, such as bus stops and crosswalks.

As can be seen in the neighborhood-specific analysis results, civil complaints about the residential environment varied from one neighborhood to another. Hansol-dong faced many complaints about public transport, parking, road traffic, and bad smell, while in Goun-dong, there were many complaints about Gountteul Park. Areum-dong received many complaints about students’ commute to and from school, while Areum-dong faced many complaints about parking and the risk of accidents near the shopping district. In Saerom-dong, there were relatively many complaints asking for the installation of safety-related facilities while demanding measures to prevent accidents.