diff --git a/README.md b/README.md
index 31f6e269e2433ba2de37963ad962d736f6bc953d..cd6834b3e8976cbe5975a6441ff10f6be621bcd2 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,45 @@
-A serialisation of the Universal Networking Language (UNL) using RDF triples.
+# This is the main repository for development and distribution of RDF-UNL.
+
+The main reference to describe RDF-UNL is the article submited to the [SEMANTiCS 2020 EU conference](https://2020-eu.semantics.cc/) :
+
+**Natural language representation and content extraction using RDF, SHACL and the Universal Networking Language (UNL)**
+
+David Rouquet (1), Valérie Bellynck(2), Vincent Berment (2,3,4) and Christian Boitet (2)
+
+1.  Tétras Libre SARL, 38400, St Martin d’Hères, France
+2.  UGA-LIG-GETALP, 38080, Grenoble, France
+3.  CS GROUP, 92350, Le Plessis Robinson, France
+4.  INaLCO, 75214, Paris, France
+
+The submited version of the article is available in the repositiory (in `documents`)
+
+# Presentation
+
+RDF-UNL proposes a complete serialization of the [Universal Networking Language (UNL)](http://www.unlweb.net/wiki/Specs) as a schema on top of RDF. 
+
+UNL is a linguistico-semantic interlingua that represents a sentence in any natural language L as a hypergraph, 
+where arcs bear semantic relations, and nodes bear interlingual lexemes (word senses) taken from an autonomous 
+lexical space, plus semantic and pragmatic features. 
+
+UNL has been acknowledged as a framework suited to machine translation and tasks such as (multilingual) 
+question answering, information extraction, information retrieval, etc.  
+Therefore it is a strong candidate to operate as a linguistic paradigm for machine reading in Semantic Web applications.
+RDF-UNL equips the Semantic Web with a linguistic framework to soundly represent the meaning of a sentence, directly in RDF. 
+
+# Working example instructions
+
+This repository contains a working example that is described in the SEMANTiCS 2020 paper.
+
+To reproduce the results, please follow the instructions bellow :
+
+1.  Download and install [Topbraid Composer (TBC) Free, Standard or Maestro Edition](https://www.topquadrant.com/topbraid-composer-install/)
+2.  Launch TBC and choose your workspace. We will call it myWorkspace.
+3.  `$ cd myWorkspace`
+4.  `$ git clone https://gitlab.tetras-libre.fr/unl/rdf-unl.git`
+5.  In TBC : `File -> New -> Project -> General -> Project -> Next`, and in `Project name` enter "rdf-unl" then click `Finish`
+6.  In TBC, open `rdf-unl/exemple-shacl-rules.shapes.ttl` in the Project explorer panel (bottom left). The file contains the SHACL rules and imports the rdf-unl model and the example.
+7.  In TBC menu : `Inference -> Run Inferences`
+8.  You should now see the infered triples in the Inference tab (bottom center)
+9.  You can now save the infered triples and run an OWL reasonner on them to exhibit the contradiction they contain (for instance in [Protégé 5](https://protege.stanford.edu/) as suggested in the paper)
+
+If you are at ease with Semantic Web programing, you can also execute the SHACL rules with a library like Jena (beware that they are SHACL SPARQL CONSTRUCT Rules and are not supported by all SHACL engines)
\ No newline at end of file