Manuel Morales Almora


Sesiones

11/11
12:00
15minutos
From Text to Map: A Methodology for the Automatic Georeferencing of Geographic Entities Extracted from News Articles and Their Relevance in Discourse.
Manuel Morales Almora

Introduction

Today, digital media produce large volumes of information on a daily basis that document social, political, environmental, and economic events, among others. Within these texts, it is common to find numerous references to geographic locations that help contextualize and spatially understand the events described. However, most of these references are merely part of a text, which makes it difficult to integrate them with geographic analysis tools and to study spatial patterns derived from journalistic discourse. The point is that what matters is not only extracting these geographic references but also being able to assess whether they are relevant to the subject of study. Today, digital media produce large volumes of information daily that document social, political, environmental, economic, and other events. Within these texts, it is common to find numerous references to geographic locations that allow us to contextualize and spatially understand the events described. However, most of these references are merely part of a text, which makes it difficult to integrate them with geographic analysis tools and to study spatial patterns derived from journalistic discourse. The key is not only to extract these geographic references but also to be able to assess whether they are relevant to the subject under study.
This paper presents a methodology for automatically converting geographic references found in news articles into geospatial information. Not only that, but it also allows us to measure the relevance of these locations so that we can generate maps more accurately and precisely, filtering and selecting the locations most relevant to our study while discarding those that are not. This proposal integrates various stages of automated processing, ranging from extracting geographic coordinates from textual documents and measuring the relevance of each location to generating cartographic products that can be further analyzed using Geographic Information Systems.

Geocentrality Metric

Geocentrality is a LexRank-based metric that measures the degree of importance of a geographic area in relation to a document’s topic. This metric is not merely equivalent to the presence of place names. A document may mention countries or regions without these having any analytical significance.
To define this metric, we must first define the geocentrality of a place name.
For a place name $t$ and a document $D$, we define the geocentrality metric as a function of $t$ and $D$ according to Equation (1).
Where $s_i$ are the sentences in document $D$, and $c(s_i)$ is a function that scores the importance of sentence $s_i$.

$$\text{geocentrality}(t, D) = \frac{\sum_{i : s_i \ni t} c(s_i)}{\sum_i c(s_i)} \tag{1}$$

Thus, we can define the geocentrality of document $D$ simply as the average geocentrality of all its geographic entities, or as zero if it does not mention any place names (Equation 2).

$$\text{average_geocentrality}(D) = \begin{cases} \frac{1}{n}\sum_{i=1}^{n} \text{geocentrality}(t_i, D), & \text{if } n>0 \ 0, & \text{if } n=0 \end{cases} \tag{2}$$

At the corpus level, we define the geocentrality of $t$ in a $corpus$ according to Equation (3).

$$\text{corpus_geocentrality}(t) = \frac{\sum_{D} \sum_{i: s_i \ni t} c(s_i)}{\sum_{D} \sum_{i} c(s_i)} \tag{3}$$

Metodology

The methodology begins with the collection of news articles using web crawling techniques. The documents obtained are stored in a structured database that preserves both the textual content and supplementary information, such as the publication date, the news outlet, and other relevant data.
Next, the text undergoes preprocessing to prepare the information for analysis. During this stage, the text is segmented into sentences, the content is normalized, and named entities are automatically extracted using Named Entity Recognition (NER) techniques. This process primarily identifies people, organizations, and places, with the latter serving as the main input for the georeferencing process.
Due to the nature of language in general, the entities that are identified often have spelling variations, abbreviations, incomplete names, or synonyms that refer to the same place. Therefore, the methodology includes a standardization process that seeks to unify these variations before performing geocoding. This stage significantly improves the consistency of the results and reduces the number of queries during the geoparsing process.
Once the geographic entities have been standardized, a geoparsing procedure is applied that combines textual information with geocoding services. For this purpose, Nominatim was used, which allows querying the OpenStreetMap open database to obtain the possible standardized address, as well as the geographic coordinates corresponding to each identified entity. It is important to note that this process also incorporated a spatial constraint tailored to the context of the study, reducing ambiguity in cases where multiple locations share the same name.
The result of this stage is a set of geographic entities with spatial information that can be visualized directly on maps or incorporated into geographic analysis tools.
As a case study, this methodology is applied to the analysis of journalistic discourse related to the socio-environmental conflict over water in the state of Jalisco, Mexico, in the year 2024. A corpus consisting of news articles from three media outlets is used: El Mural, El Informador, and La Jornada. The objective is not only to identify geographic entities but also to analyze the spatial distribution of the places mentioned in journalistic discourse and to explore how different media outlets represent this phenomenon territorially.
In addition to spatial analysis, the proposed methodology incorporates the geocentrality metric to estimate the relevance of geographic entities within each media outlet.

Conclusion

Unlike traditional approaches, which typically treat all identified locations equally, geocentrality allows us to classify geographic entities based on their semantic importance within the document, enabling the creation of thematic maps that highlight the locations most relevant to the topic under study. This approach, combined with the extraction of geographic entities and the acquisition of spatial data, can facilitate a spatial interpretation of journalistic discourse, as this methodology highlights the geographic areas that play a central role in the events under analysis.

Sistemas de Información Geográfica, Cartografía y Geoinformática
Sala de Video
12/11
11:45
15minutos
Spatial association between public bicycle trips and tertiary economic structure in Mexico City: an exploratory spatial data analysis approach for last-mile mobility.
Ana Laura Salguero Melgar, Manuel Morales Almora, Raul Mora, Néstor Conrado Vargas Ruiz

Abstract
The growing functional complexity of contemporary metropolises has increased the need to develop methodologies capable of explaining the spatial interaction between urban infrastructure, mobility, and economic activity. In this context, Geographic Information Science (GIScience) provides a set of analytical tools that make it possible to identify spatial patterns, evaluate territorial relationships, and generate useful evidence for data-driven territorial planning. Beyond cartographic representation, spatial analysis is currently a fundamental component for understanding complex urban processes characterized by spatial dependence, territorial heterogeneity, and emerging functional structures.
Mexico City (CDMX) represents an exceptional laboratory for this type of research, as it concentrates one of the largest metropolitan systems in the world, characterized by strong economic tertiarization, high daily mobility flows, and a polycentric organization sustained by specialized employment and service corridors. Within the Integrated Mobility System, Ecobici stands out as a public bike-sharing system implemented as a mechanism to address first- and last-mile trips. However, there is still little evidence regarding the spatial relationship between the use of this infrastructure and the territorial distribution of the economic activities that generate such trips.
Within this context, the present research sets as its general objective determining whether a significant spatial association exists between the intensity of trips made through the Ecobici system and the concentration of tertiary-sector economic units in the central boroughs of CDMX. The central hypothesis holds that the distribution of trips does not follow a random pattern, but rather responds to the location of tertiary corridors that concentrate employment, services, and advanced economic activities.
The conceptual framework integrates contributions from economic geography, urban theory, and geographic information science. From Christaller's central place theory, the notion of functional centrality is drawn; approaches on metropolitan polycentrality and tertiary corridors developed for CDMX are subsequently incorporated, along with the concepts of derived mobility, accessibility, and last mile. Finally, the research adopts the principles of spatial analysis, which allow the study of spatial dependence between geographic variables through global and local statistics.
Methodologically, the study develops a reproducible procedure implemented entirely in R using specialized libraries for spatial processing, spatial statistics, and cartographic visualization. The information base integrates three main sets of georeferenced data: open trip records from the Ecobici System for the year 2019, economic units from the National Statistical Directory of Economic Units (DENUE) classified within the tertiary sector, and neighborhood (colonia) polygons of the city. All data were standardized within a single spatial reference system and subsequently aggregated to comparable territorial units.
As a first stage, an integrated spatial database was built that made it possible to quantify both the total number of trips and the concentration of tertiary establishments within each spatial unit. The existence of spatial autocorrelation was then evaluated using the global Moran statistic with a Queen-type contiguity matrix. This procedure made it possible to determine whether the territorial distribution of trips exhibited random behavior or whether systematic patterns of spatial clustering existed. The results show a Moran's Index of 0.512 with high statistical significance (p < 2.2 × 10⁻¹⁶), confirming the existence of positive spatial dependence and ruling out the hypothesis of randomness.
Once the existence of a global spatial structure was confirmed, Local Indicators of Spatial Association (LISA) and the Getis-Ord statistic were applied in order to identify the precise location of high- and low-intensity territorial clusters. These analyses made it possible to locate High-High clusters concentrated mainly along the urban corridors with the greatest functional centrality, spatially coinciding with Reforma, Polanco, Insurgentes, and the Historic Center, while Low-Low clusters are located in peripheral areas of the system's coverage area.
In order to reduce the bias associated with the Modifiable Areal Unit Problem (MAUP), the study incorporated an additional procedure based on regular tessellation of space. Rather than relying exclusively on neighborhood boundaries, a homogeneous grid of equal-sized quadrants was constructed, onto which both variables were re-aggregated through spatial intersection operations. The cell size was then adjusted iteratively to minimize the presence of non-significant quadrants, thereby strengthening the statistical robustness of the subsequent analyses.
Using this new spatial representation, a bivariate spatial autocorrelation analysis was carried out through the bivariate Moran statistic, making it possible to simultaneously evaluate the relationships between trip intensity and the concentration of tertiary activities in neighboring units. Finally, to ensure the statistical reliability of the identified patterns, validation was implemented through Monte Carlo simulations using 999 random permutations and a significance level of α = 0.05.
The results obtained show a positive spatial association between both variables. High-High clusters are identified where elevated concentrations of tertiary establishments coincide with high levels of Ecobici trips, evidencing that the system operates preferentially in specialized urban corridors where employment and specific services are concentrated. These findings support the hypothesis that Ecobici functions as a last-mile mobility infrastructure closely linked to the metropolitan economic structure, rather than as a system oriented exclusively toward recreational or environmental purposes. Likewise, the identification of Low-Low clusters reveals persistent territorial inequalities in the distribution of cycling infrastructure, suggesting opportunities for future system expansion strategies.
From the perspective of Geographic Information Science, the study's main contribution lies in the simultaneous integration of spatial analysis techniques within a reproducible methodological workflow developed entirely using free software to analyze a complex socio-territorial problem.

Sistemas de Información Geográfica, Cartografía y Geoinformática
Sala de Video