(site under construction)


Essential component of any Quality Center:
Produce more efficient and coherent technical documents

Based on research supported by the ANR (national research agency) and the French CNRS

LELIE is your personal intelligent assistant that helps you improve your technical documents. The main characteristics of LELIE are:

- PRESCRIPTIVE: dynamic modelling of the main principles and norms of technical authoring, over several industrial domains,

- ADAPTABLE and RELIABLE: auto-adaptative approach based on an accurate analysis of technical writing,

- SIMPLE and INTUITIVE: available as a Plug-In for Word and Excel,

- RELEVANT and EFFICIENT: LELIO automatically customises LELIE to your industrial domain and authoring habits,

LELIE is FREE: open source, languages: French and English. We celebrate 100 free copies distributed!

The facets of LELIE and LELIO:

LELIE in 10 major fonctions

LELIO: the art of customizing LELIE

Perspectives: our road-map

Illustrations: LELIE outputs

The download corner

The juridical corner

Main industrial uses

The main events around LELIE and LELIO, LELIE Consulting Network


Address of main Contact:

Patrick SAINT-DIZIER, senior researcher at CNRS
ADRIA / ILPL, Institut de Recherches en Informatique de Toulouse
118, route de Narbonne
31062, Toulouse Cedex, France

Email: stdizier@irit.fr
Phone: +33 5 61 55 62 44

LELIE: an intelligent technical text clinic

In 10 points, the art of making your documents more coherent and efficient:

The functions offered by LELIE are based on existing norms and the everyday practice of technical authors. LELIE also offers specific controls dedicated to precise types of documents: specifications, requirements, procedures, equipment notices, etc. and to specific theoretical approaches (e.g. pattern-based authoring, minimalist technical writing).

The 10 major functions are the following:

(1) Fuzzy and vague terms detection: since these affect the understandability and the verifiability of instructions or requirements.

(2) Terms to avoid detection: such as very generic terms, buzz terms, ambiguous terms, or common terms not in the technical domain ontology.

(3) Sentence complexity analysis: LELIE detects via a metrics those sentences which are difficult to read because of their complex structure (with e.g. long coordinations, embedded relatives), their length, or the presence of too many technical terms.

(4) Ambiguous pronoun reference detection: LELIE detects these pronouns in different configurations since they frequentlty introduce ambiguities.

(5) Various types of negation detection: a number of negative forms should be avoided: they introduce completeness and verifiability problems.

(6) Tenses to avoid detection: in particular agentless passives, future tenses and conditional and subjunctives must be avoided as much as possible since they affect the completeness of the document.

(7) Controls on the syntactic patterns imposed by the document style: the objective is to ensure that sentences follow the schemas or patterns required by the domain or company.

(8) Enumeration regularity controls: enumerations and cells in charts are easier to understand if they follow a regular syntactic format.

(9) Cohesion controls: LELIE controls the terminological regularity of documents, given a domain ontology. This avoids term proliferations.

(10) Similarities analysis: LELIE identifies in a text, even a large text, those elements which are very similar: this may indicate redundancies or possible incoherences.

Via Plug-Ins in Word and Excel, annotations are produced as Word comments. These comments include several features including the severity level of the error. Comments can be customized to the user habits.
A global metrics gives an overall estimate of the text authoring quality with respect to norms.


LELIO or the art of customisation

LELIE is a generic kernel: it works rather well on any type of technical text.

For better results, in error diagnosis and corrections to make, LELIO is a research action that allows you to create your own LELIE, customized to your domain and authoring practices. These elements are specified via a collaborative environment.

To realize this customization, LELIO requires:
- samples of your texts,
- your authoring guidelines or habits or preferences, specified in a given format,
- your preferences in terms of disgnosis display.

Contact us to use LELIO, developed by an independent group (Lelie Consulting Network).


Perspectives - Road-Map

Our perspectives are organized around major topics in technical writing. A close collaboration with technical writers allows us to identifiy the following topics:

- writing documents in English when English is not your mother tongue: typical errors and remediation strategies, IREB project,

- the characterization of inocuity: the difficulty to establish error diagnosis which are relevant: indeed many so called errors are perfectly understandable, in spite of norms which are too strict and general,

- the analysis of forms of incoherence among requirements or in technical documents,

- the developpement of an auto-adaptative system to correct errors,

- social and cognitive ergonomics in LELIE.


Some examples of LELIE outputs

Errors expressed in Word mode: (scroll on the image to see the whole example):

Errors expressed in sentence mode:


Download section

TextCoop and LELIE are good quality research prototypes, which are used in real industrial contexts. They are distributed FREE of charge (see section juridique).

The TextCoop and LELIE archives contain a user manual to start the system, running examples are provided.

discourse semantics analysis platform, SOFTWARE freely available (French and English), OPEN SOURCE.

TEXTCOOP V5.1 (September 2015, archive with the TextCoop engine, and large samples of updates discourse analysis rules and related lexical data). TEXTCOOP V5.1,
TextCoop has been distributed to more than 35 research centers or companies.

LELIE V1: old version (2014), runs well but based on a simple pattern technology, with some linguistic and technical limitations.
CODE: LELIE V1 French and LELIE V1 English.

LELIE V2: version based on constraints and rules, much more efficient and accurate. Associated with LELIO for customization, compatible with RAT-RQA.
Last version: LELIE V2.2 (May 2016), about 100 copies sent at the moment.
Please send email to stdizier@irit.fr or to lelieconsultingnetwork@gmail.com to get LELIE V2.2 in the Word/Excel Plug-In, CC BY NC, see the juridical corner section.

A version LELIE V3 is under preparation: it places the user at the very center of the process of the technical text improvement.

Short and funny introductory video to LELIE (in French)

LELIO V1 is not distributed. For any customisation of LELIE by LELIO, contact stdizier@irit.fr


The juridical corner

TextCoop, LELIE and LELIO prototypes are results of research projects developed by the CNRS, the ANR and the University in collaboration with the CNAM and EDF. They belong to the CNRS and the university.

Authors are happy to share these results and associated prototypes and data provided that some return of experience, positive or negative is sent back to us, the goal being to help us further improve these prototypes.

- TextCoop V4 and V5 are under CC BY license (open-source, commercial applications allowed) since August 2015.

- LELIE V1 and V2.2 are under license CC BY NC (free usage for research, commercial or technical opportunity analysis, internal tests).

The plugs-Ins Word and Excel are Open Source.

- To make a commercial use of LELIE V2.2, there are several simple solutions, contact: stdizier@irit.fr


LELIE users

LELIE has been tested or is being tested in several companies in various industrial sectors: EDF, SNCF, IBM, Reuters (financier), Orange, Siemens, Accenture, LGM, Pôle Emploi, Daimler-Benz, IFDS, RequirementOne, INCOSE, the ReuseCompany, etc.

LELIE is used for teaching purposes for technical text authoring, in particular requirements, in cooperation with the IREB.

LELIE is being integrated in various environments (Quality Center).

Products based on LELIE with our collaboration: Semiosis.zeen (Lelie V2.1 redone, plug-In low cost), LM (Lelie V2.1 for multilingual documents), Inuite.de. More ambitious systems in preparation.


events, readings, notes

under construction

- P. Saint-Dizier, Challenges of Discourse processing: the case of technical documents, April 2014, Cambridge Scholars Publising.

- P. Saint-Dizier: An Introduction to Prolog, Springer-Verlag, 1989.



under construction