Using Data Visualization to Explore International Trade Agreements

poster / demo / art installation
  1. 1. Oliver Ford

    University of Edinburgh

  2. 2. Esteban Serrano

    University of Edinburgh

  3. 3. Xinyu Du

    University of Edinburgh

  4. 4. Anouk Lang

    University of Edinburgh

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

This poster explores what can be learnt by applying different data visualization methods to a corpus of 450 preferential trade agreements, gathered and structured into XML format by the
ToTA: Text of Trade Agreements project (Alschner et al. 2017) and available at As lengthy legal documents, these agreements present an interesting challenge to data mining and visualization: they contain a large amount of boilerplate and have a high degree of similarity to one another, with the “devil in the detail”, so their interpretation by human readers requires significant trade and legal expertise. There is, moreover, currently relatively little crossover between computational analysis and the domain of international trade, so the ToTA corpus opens up an opportunity for discerning which computational methods have the most potential to drive forward research at the interdisciplinary intersection of legal research and digital humanities.

Our research questions centered on two main areas. First, what kinds of relationships between countries can be discerned by examining the text of legal documents that regulate economic interactions between those countries? And second, what is the relationship between the documents themselves, especially concerning how earlier trade agreements may have influenced those that followed them? In addressing these questions, we employed a number of different data analysis and visualization approaches:

Topic modelling and visual rendering of the similarity of documents based on topics within them demonstrated that trade agreements tended to cluster along regional lines, with some chronological patterning. Figure 1 demonstrates how the trade agreements can be broadly separated along continental lines into Europe, CIS and East Asia, East Asia and Central America, and North America and Oceania, with the nations in the Commonwealth of Independent States engaging predominantly only with each other.

Network visualizations were used to show the proximity of different countries based on the trade agreements in which they were involved. Figure 2 gives an example of this: a network representation of the countries who signed agreements with each other in the period 1989-1998. This graph produced some expected clustering along the lines of geographical regions, for instance the former Communist Bloc nations such as Ukraine, Georgia, Kazakhstan and Tajikistan, whose position between the nodes representing the EU, the EC and EFTA on the one hand, and the node representing the Russian Federation visually embodies the way these nations were caught between two power blocs in the period of geopolitical realignment following the Cold War. It also represents some less obvious similarities, for instance the clustering of smaller political entities such as Andorra, San Marino, the Palestinian Authority and Finland. Creating a number of network visualizations using chronological slices of the data enabled us to represent change over time, and using these it is for instance possible to see trends such as the gradual rise in prominence of Asian nations such as Malaysia, Korea and Vietnam in the years from 1999 to 2016.

A word2vec word embedding model was trained on the text of the agreements. Despite the high proportion of boilerplate, the model successfully clustered terms from the same domain, for example import/export (eg.
originating), progressive concerns (
sustainable) and bureaucracy (

Figure 1. Visualisation of trade agreements clustered by topic demonstrating regional similarities and showing change over time.

Figure 2. Network visualization of countries which signed trade agreements between 1989-1998.


Alschner, W., Seiermann, J. and Skougarevskiy, D. (2017).
Text-as-Data Analysis of Preferential Trade Agreements: Mapping the PTA Landscape. UNCTAD Research Paper No. 5, United Nations Conference on Trade and Development, (accessed 10 May 2019).

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.