• 1
• 2
• 3
• 4
• 5
• 6
• 7
1
Text-mining needs of the 
food microbiology 
research community
Estelle Chaix, Sophie Aubin, 
Louise Deléger and Claire Nédellec
Speaker: Robert Bossy
IN-OVIVE 2017 at EFITA WCCA congress - July 2nd - 6th - Montpellier
2Context
Microorganisms : very abundant and can live in extreme 
conditions
Microbial diversity research: 
● Microbiomes
● Microbial interactions and ecosystems
● Phylogeny
For :
● Human, plant or animal health
● Plant growth, bioremediation or food processing
Use molecular technologies (DNA-sequencing and 
metagenomics) 
3Context
Data and knowledge sources on microbial 
biodiversity:
● Experimental data
● Curated databases
● Textual documents : scientific publications, 
reports, patents or medical records
From Microbial biodiversity community needs 
To Ontology-based Text-Mining applications
4Context
This work is part of the development of the 
European infrastructure for text-mining 
OpenMinTeD (http://openminted.eu/)
This project targets 4 different research 
communities through different use-case 
examples. This work focuses on the Agriculture 
and Biodiversity field.
5Methodology of user needs analysis
1. Identification of end-users and preliminary 
needs → scope of the use-case
2. Identification of stakeholders, their needs and 
theirs interactions
3. Validation of the requirements provided by the 
need analysis
4. Design of the TDM solution
61. End-user identification
Microbiology 
community
Research 
laboratories
Agro-
industrial 
Technology 
Institutes
Agrofood 
Companies
Food Safety 
Agencies
Research 
Networks
72. End-user needs
Target: sub-community of researchers working on 
positive microbial flora 
Method : User-centered method (persona)
1. Who is the persona? 
2. What information are they interested in? 
3. How do they need to access the information? 
4. What solutions would satisfy their needs?
82. Persona example
Lily: 38 years old, former baker 
and expert in bread leavens
● She wants to compare commercial and artisanal 
leavens.
● She needs to know the habitats and phenotypes of 
bread starter cultures.
● She reads scientific papers, browse microbial 
collections, draws expertise from experience and 
baker networks.
● Requires: data and knowledge about microbial 
phenotypes and taxonomic diversity of leaven.
92. End-user interface
One main interface for all personas needs
10
3. Validation of the end-user needs
● Microbial biodiversity: qualify microorganism 
provenance (prevent contamination).
● Microbial phenotype: degradation and 
production of molecules.
● Industrial use: impact of microbial phenotype 
in the context of food processing (flavor, taste, 
biopreservation).
A common vocabulary of microbe habitats and 
phenotypes.
11
4. TDM solutions
Extraction from scientific publications
● relevant entities (microorganisms, habitats, 
molecules, phenotypes and applications) and 
their normalization with ontologies.
● Relations between entities.
State-of-the-art resources
● Bacteria Biotope task of the BioNLP-Shared Task for  
microorganisms and habitats.
● OntoBiotope ontology for habitats and phenotypes.
● BioCreative IV CHEMDNER Task for molecules.
● Methods for relation extraction: TEES or AlvisRE.
12
Conclusion
● Converging needs identified from diverse 
fictional users.
● TDM solutions have to be linked to 
traditional data sources.
● Ontologies as a means to aggregate 
different sources.