Difference between revisions of "Open Food Facts Days 2024/rebuilding open food facts from the ground up workshop"
(workshop result in french) |
|||
Line 150: | Line 150: | ||
== English (automatic translation) == | == English (automatic translation) == | ||
+ | Participants: Yassine, Antoine, Gustave, Anthony and Heyman | ||
+ | |||
+ | === Discussion === | ||
+ | |||
+ | ==== On data and their organizations ==== | ||
+ | OFF is not only on food today. | ||
+ | |||
+ | Think about that from the start. | ||
+ | |||
+ | BDD currently Food oriented. | ||
+ | |||
+ | GS1 -GDSN data -> But not really good format | ||
+ | |||
+ | GS1 Voc -> Schema.org extension to map all the fields of the GDSN -400 properties | ||
+ | |||
+ | Openbatra - Let's give the same format to small producers (GSDN + Extension Schema.org) | ||
+ | |||
+ | OpenFoodFacts -> Mongodb, a lower schema, it is a choice that could be kept. | ||
+ | |||
+ | Because we do not know the projects that will be added, therefore be evolving. | ||
+ | |||
+ | At LLM time, structured data (whatever the data), is what is important. | ||
+ | |||
+ | Offer to producers a data restory: PDF, photo sheet, GS1 -> data and deduce the structured data | ||
+ | |||
+ | Today unique identifier = gtin | ||
+ | |||
+ | To come, digital passport of the product: from its beginning of life to sorting waste. With a central register. | ||
+ | |||
+ | It is surely the GS1 format that will impose itself. | ||
+ | |||
+ | Online GS1 Forum this week, note: they are fully in AI (answer user questions). | ||
+ | |||
+ | Note: Return of experience on the standardization of carpooling, | ||
+ | |||
+ | which was difficult to bear for existing actors, | ||
+ | |||
+ | But interoperability at the end wins. | ||
+ | |||
+ | ==== Higher level ==== | ||
+ | What is the objective of Open Food Facts: | ||
+ | |||
+ | <nowiki>*</nowiki> Initially the goal is data transparency | ||
+ | |||
+ | <nowiki>*</nowiki> Today -> Reliability is a little more important | ||
+ | |||
+ | <nowiki>*</nowiki> But the ultimate goal: to respond to a problem, to a decision -making | ||
+ | |||
+ | This product is good or not good - depending on my context. | ||
+ | |||
+ | But it would also be important to have reliability clues (crossing information for example). | ||
+ | |||
+ | Nutriscore: is not intra-category | ||
+ | |||
+ | In England - an intra category score - | ||
+ | |||
+ | The Eren thought that to change the habits, non-intra-category is best. | ||
+ | |||
+ | On the environment it is very important not to do intra-category, because the orders of magnitude are even greater. | ||
+ | |||
+ | Off is at the crossroads of lots of data and use sources. | ||
+ | |||
+ | Off is already on a lot of levels, just wanting to have the data on lots of products from the world, | ||
+ | |||
+ | Making indicators is yet another level, all the level access to the public and ease of contribution is also important. | ||
+ | |||
+ | Look at the Open Street Map side which also has very heterogeneous data: a store opening time, beer prices, elevation. | ||
+ | |||
+ | They let the contributors create their data and applications. For example on climbing we have lots of super precise data. | ||
+ | |||
+ | Can be separated: | ||
+ | |||
+ | <nowiki>*</nowiki> Data collection and quality | ||
+ | |||
+ | <nowiki>*</nowiki> Creation of algos etc. | ||
+ | |||
+ | <nowiki>*</nowiki> Creation of new use cases | ||
+ | |||
+ | Question to ask yourself am I b2c or rather an ecosystem with start-ups, producers, the big public. | ||
+ | |||
+ | Here off was rather large public: to put reliable info to the big public for its decisions (especially purchases) | ||
+ | |||
+ | In areas where the issues are really important. The quality of data is higher here compared to Open Street Map. | ||
+ | |||
+ | The very centralized model of Open Food Facts may find it difficult to go to scale. | ||
+ | |||
+ | The Crowdsourcing + Direct data aspect will remain. | ||
+ | |||
+ | If you are flexible on the data it is the one who re-uses who must filter. | ||
+ | |||
+ | You need a standard all the same for structure the data. | ||
+ | |||
+ | Open Food Facts Technical Stack is today a brake on the contribution. | ||
+ | |||
+ | Reliability: On Open Street Map we have to indicate in which context we had the data. | ||
+ | |||
+ | (on Open Food Facts it is still in boxes to better ask this information) | ||
+ | |||
+ | Indicator also also depends on the context: e.g. not very important for a diabetic. | ||
+ | |||
+ | Contextuality is very important. | ||
+ | |||
+ | Open Food Facts is a brick for decision -making aid, not necessarily its role. | ||
+ | |||
+ | As a re-user it gives me access to a large number of data. | ||
+ | |||
+ | And the concept of quality depends on the context too and what we are looking for. | ||
+ | |||
+ | Are we a nodal point or a final user solution. | ||
+ | |||
+ | Filling the three is too difficult. | ||
+ | |||
+ | At the same time today the mobile app helps to secure the contribution and impact of Open Food Facts. | ||
+ | |||
+ | Central Food Data in the USA (public). Updated every month, with GS1. | ||
+ | |||
+ | Theyda is also quite important. | ||
+ | |||
+ | Can we have public initiatives. | ||
+ | |||
+ | The verticality of the process is an interesting issue. But also allow external actors to plug in at the level of their choice. | ||
+ | |||
+ | Opening: can we have more political things in Open Food Facts? Match for example the declarations correspond (verification). | ||
+ | |||
+ | Should we integrate this kind of information? | ||
+ | |||
+ | Current Off Policy: the most important fact base | ||
+ | |||
+ | Example also, the distribution of value. | ||
+ | |||
+ | === Conclusion === | ||
+ | Open Food Facts should not manage all incoming (acquisition) and outgoing (applications) flows | ||
+ | |||
+ | At least one vertical for project survival (capturing data) and also facilitating the re-user job. | ||
+ | |||
+ | (ex today a problem is that the OpenStreetMap.org site does not reveal the whole project at all) | ||
+ | |||
+ | Important to keep the two cannals Acquisition: Producers, Crowd-Sourcing. | ||
+ | |||
+ | Really base yourself on the GTIN (and its developments) and the GS1 compatibility to prepare the future. | ||
+ | |||
+ | The Open Food Facts brand must be synonymous with information fiablite, | ||
+ | |||
+ | To also bring to more large public contributors. | ||
+ | |||
+ | Also have this side against power / independence, research and confidence helps. | ||
+ | |||
+ | Continue the componentization of the project, to promote re-use at different levels. |
Latest revision as of 11:58, 21 September 2024
French
Participants: Yassine, Antoine, Gustave, Anthony et Heyman
Discussion
Sur les données et leur organisations
OFF n'est pas seulement sur l'alimentaire aujourd'hui.
Penser à ça depuis le début.
BDD actuellement orientée food.
GS1 - données GDSN --> mais pas vraiment bon format
GS1 voc --> extension de schema.org pour mapper tous les champs du GDSN - 400 propriétés
OpenBATRA - donnons le même format aux petits producteurs (GSDN + extension schema.org)
OpenFoodFacts --> mongodb, schema moins fort, c'est un choix qui pourrait être gardé.
Car on ne sait pas les projets qui seront ajoutés, donc être évolutif.
À l'heure des LLM, la donnée structurées (quelque soit la donnée), est ce qui est important.
Proposer aux producteurs un repository de données : fiche PDF, photo, données GS1 --> et en déduire les données structurées
Aujourd'hui identifiant unique = GTIN
À venir passeport digital du produit: de sa début de vie au tri des déchets. Avec un registre central.
C'est surement le format GS1 qui va s'imposer.
Forum du GS1 en ligne cette semaine, note: il sont à fond dans l'IA (répondre à des questions de l'utilisateur).
Note: Retour d'expérience sur la mise en norme du co-voiturage,
qui a été difficile à supporter pour des acteurs existants,
mais l'interopérabilité à la fin gagne.
Plus haut niveau
Quel est l'objectif d'Open Food Facts:
* initialement le but est transparence des données
* aujourd'hui -> la fiabilité est un peu plus importante
* mais le but ultime: répondre à une problématique, à une prise de décision
Ce produit est il bon ou pas bon - selon mon contexte.
Mais il serait important aussi d'avoir des indices de fiabilité (en croisant des informations par exemple).
Nutriscore: n'est pas intra-catégorie
En Angleterre - un score intra catégorie -
L'EREN a pensé que pour changer les habitudes le non intra-catégorie est le mieux.
Sur l'environnement c'est très important de ne pas faire de l'intra-catégorie, car les ordres de grandeurs sont encore plus grand.
OFF est au carrefour de plein de sources de données et d'usages.
OFF est déjà sur plein de niveaux, rien qu'en voulant avoir les données sur plein de produits du monde,
Faire des indicateurs est encore une autre niveau, tout le niveau accès au public et facilité de contribution est aussi important.
Regarder du côté d'Open Street Map qui a aussi des données très hétérogènes: heure d'ouverture d'un magasin, prix de la bière, dénivelé.
Ils ont laissé les contributeurs créer leurs données et leurs applications. Par exemple sur l'escalade on a plein de données super précise.
Peut être séparer:
* collecte donnée et qualité
* création des algos etc.
* création de nouveaux cas d'usage
Question à se poser suis je B2C ou plutôt un écosystème avec des start-up, les producteurs, le grand publique.
Ici OFF était plutôt grand publique: mettre de l'info fiable à destination du grand publique pour ses décisions (surtout d'achats)
dans des domaines ou les enjeux sont vraiment importants.La qualité de donnée est ici supérieur par rapport à Open Street Map.
Le modèle très centralisé de Open Food Facts aura peut être du mal à passer à l'échelle.
L'aspect crowdsourcing + données directes restera.
Si on est flexible sur les données c'est à celui qui ré-utilise qui doit filtrer.
Il faut une norme tout de même pour structure la données.
Stack technique d'Open Food Facts est aujourd'hui un frein à la contribution.
Fiabilité: sur Open Street Map on doit indiquer dans quel context on a eu la donnée.
(sur Open Food Facts c'est encore dans les cartons de mieux poser cette information)
Indicateur dépend vraiment aussi du contexte: par ex nutri-score peu important pour un diabétique.
La contextualité est très importante.
Open Food Facts est une brique de l'aide à la décision, pas forcément son rôle.
Comme ré-utilisateur ça me donnè accès à un grand nombre de donnée.
Et la notion de qualité dépend du context aussi et de ce qu'on cherche.
Est-on un point nodale ou une solution utilisateurs finaux.
Remplir les trois est trop difficile.
En même temps aujourd'hui l'app mobile aide à sécuriser la contribution et l'impact d'Open Food Facts.
Food data centrale aux USA (publique). Mis à jour tous les mois, avec GS1.
USDA est aussi assez important.
Peut on avoir des prises d'initiatives publiques.
La verticalité du processus est un enjeu intéressant. Mais aussi permettre aux acteurs extérieurs de se plugguer au niveau de leur choix.
Ouverture: Peux-on avoir des choses plus politiques dans Open Food Facts ? Matcher par exemple les déclarations correspondent (vérification).
Faut-il intégrer ce genre d'informations ?
Politique actuelle de OFF: le plus important fact base
Exemple aussi, de la répartition de la valeur.
Conclusion
Open Food Facts ne doit pas gérer tous les flux entrants (acquisition) et sortant (applications)
Au moins une verticale pour la survie du projet (capturer les données) et en facilitant aussi le boulot de ré-utilisateur.
(ex Aujourd'hui un problème est que le site d'openstreetmap.org ne révèle pas du tout l'ensemble du projet)
Important de garder les deux cannaux acquisition: producteurs, crowd-sourcing.
Se baser vraiment sur le GTIN (et ses évolutions) et la compatiblité GS1 pour préparer le futur.
La marque Open Food Facts doit être synonyme de fiablité de l'information,
pour porter aussi à plus de contributeurs grand publique.
Avoir aussi ce côté aussi contre pouvoir / indépendance, aide à la recherche et confiance.
Continuer la componentisation du projet, pour favoriser la ré-utilisation à différents niveaux.
English (automatic translation)
Participants: Yassine, Antoine, Gustave, Anthony and Heyman
Discussion
On data and their organizations
OFF is not only on food today.
Think about that from the start.
BDD currently Food oriented.
GS1 -GDSN data -> But not really good format
GS1 Voc -> Schema.org extension to map all the fields of the GDSN -400 properties
Openbatra - Let's give the same format to small producers (GSDN + Extension Schema.org)
OpenFoodFacts -> Mongodb, a lower schema, it is a choice that could be kept.
Because we do not know the projects that will be added, therefore be evolving.
At LLM time, structured data (whatever the data), is what is important.
Offer to producers a data restory: PDF, photo sheet, GS1 -> data and deduce the structured data
Today unique identifier = gtin
To come, digital passport of the product: from its beginning of life to sorting waste. With a central register.
It is surely the GS1 format that will impose itself.
Online GS1 Forum this week, note: they are fully in AI (answer user questions).
Note: Return of experience on the standardization of carpooling,
which was difficult to bear for existing actors,
But interoperability at the end wins.
Higher level
What is the objective of Open Food Facts:
* Initially the goal is data transparency
* Today -> Reliability is a little more important
* But the ultimate goal: to respond to a problem, to a decision -making
This product is good or not good - depending on my context.
But it would also be important to have reliability clues (crossing information for example).
Nutriscore: is not intra-category
In England - an intra category score -
The Eren thought that to change the habits, non-intra-category is best.
On the environment it is very important not to do intra-category, because the orders of magnitude are even greater.
Off is at the crossroads of lots of data and use sources.
Off is already on a lot of levels, just wanting to have the data on lots of products from the world,
Making indicators is yet another level, all the level access to the public and ease of contribution is also important.
Look at the Open Street Map side which also has very heterogeneous data: a store opening time, beer prices, elevation.
They let the contributors create their data and applications. For example on climbing we have lots of super precise data.
Can be separated:
* Data collection and quality
* Creation of algos etc.
* Creation of new use cases
Question to ask yourself am I b2c or rather an ecosystem with start-ups, producers, the big public.
Here off was rather large public: to put reliable info to the big public for its decisions (especially purchases)
In areas where the issues are really important. The quality of data is higher here compared to Open Street Map.
The very centralized model of Open Food Facts may find it difficult to go to scale.
The Crowdsourcing + Direct data aspect will remain.
If you are flexible on the data it is the one who re-uses who must filter.
You need a standard all the same for structure the data.
Open Food Facts Technical Stack is today a brake on the contribution.
Reliability: On Open Street Map we have to indicate in which context we had the data.
(on Open Food Facts it is still in boxes to better ask this information)
Indicator also also depends on the context: e.g. not very important for a diabetic.
Contextuality is very important.
Open Food Facts is a brick for decision -making aid, not necessarily its role.
As a re-user it gives me access to a large number of data.
And the concept of quality depends on the context too and what we are looking for.
Are we a nodal point or a final user solution.
Filling the three is too difficult.
At the same time today the mobile app helps to secure the contribution and impact of Open Food Facts.
Central Food Data in the USA (public). Updated every month, with GS1.
Theyda is also quite important.
Can we have public initiatives.
The verticality of the process is an interesting issue. But also allow external actors to plug in at the level of their choice.
Opening: can we have more political things in Open Food Facts? Match for example the declarations correspond (verification).
Should we integrate this kind of information?
Current Off Policy: the most important fact base
Example also, the distribution of value.
Conclusion
Open Food Facts should not manage all incoming (acquisition) and outgoing (applications) flows
At least one vertical for project survival (capturing data) and also facilitating the re-user job.
(ex today a problem is that the OpenStreetMap.org site does not reveal the whole project at all)
Important to keep the two cannals Acquisition: Producers, Crowd-Sourcing.
Really base yourself on the GTIN (and its developments) and the GS1 compatibility to prepare the future.
The Open Food Facts brand must be synonymous with information fiablite,
To also bring to more large public contributors.
Also have this side against power / independence, research and confidence helps.
Continue the componentization of the project, to promote re-use at different levels.