C. Umana Dajud

Software

FtaCoder

Trade agreements have grown from tariff schedules into thousand-page rulebooks. FtaCoder reads them so you don't have to: text analysis and machine learning that codes what each agreement covers — used in the Journal of Economic Geography paper on business travel.

View on GitHub ↗git clone github.com/camiloumanadajud/FtaCoder

What 450 trade agreements talk about

Share of agreements covering each topic; the darker segment is legally binding — enforceable through a dispute-settlement mechanism.

Trade Remedies312 cover · 279 binding
Rules of Origin288 cover · 257 binding
Intellectual Property257 cover · 239 binding
Investment245 cover · 212 binding
Dispute Settlement238 cover · 122 binding
Environment227 cover · 194 binding
Government Procurement225 cover · 199 binding
Technical Barriers218 cover · 204 binding
Sanitary Measures204 cover · 172 binding
Services178 cover · 151 binding
Labour150 cover · 128 binding
Movement of Persons102 cover · 74 binding
Competition76 cover · 57 binding
Mining73 cover · 72 binding
E-commerce64 cover · 54 binding

Look up an agreement

All 450 agreements of the TOTA corpus, coded by FtaCoder.

Live demo

Run it in your browser

The core of FtaCoder is keyword and pattern matching over the full text of agreements (the TOTA corpus), refined with classifiers for dispute-settlement language. This demo runs the same pattern matching, client-side, on four well-known agreements — pick a topic or write your own patterns.

Corpus: Text of Trade Agreements (TOTA), CC BY-NC 4.0. The full pipeline — including the machine-learning classification of dispute-settlement clauses — is in the GitHub repository.