Changes

Jump to navigation Jump to search
1,148 bytes added ,  15:57, 15 November 2022
explanation on ingredients acquisition
Line 12: Line 12:     
=== Filling ingredients data ===
 
=== Filling ingredients data ===
Thanks to automatic Optical Characters Recognition (OCR), this field can be filled by softwares. But OCR is not always good, and you should always verify the result.
+
This is done through an application. Steps as are follow:
 +
 
 +
* text acquisition
 +
** either a user type ingredients
 +
** or
 +
*** he upload an image, or select an existing one
 +
*** we send it to an OCR (image to text)
 +
*** we fill the text but ask user to verify and eventually correct it
 +
* text parsing
 +
** when submitted, text description is parsed by the [https://github.com/openfoodfacts/openfoodfacts-server/ product opener] code
 +
** this code heavily relies on the [[ingredients taxonomy]] to identify as much ingredients as possible
 +
** but also on specific local habits in the way to write ingredients (and you may contribute code for your country, or report specific problems)
 +
** it also tries to distinguish between ingredients, origins, processing, quantities, and sub-ingredients (ingredients composing other ingredients)
    
=== Database ===
 
=== Database ===
In the database, this field is called <code>ingredients_text_[language code]</code>.
+
In the database, this field is called <code>ingredients_text_[language code]</code> which contains the raw data by language.
 +
 
 +
<code>ingredients</code> contains all inferred information thanks to parsing, <code>ingredients_tags</code> contains ingredients taxonomy entries (and their parents) corresponding to identified ingredients.
 +
 
 +
Other fields are derived from this analysis like <code>allergens_from_ingredients</code> or <code>ingredients_analysis</code>
    
=== Tools ===
 
=== Tools ===
Line 24: Line 40:  
==== Ingredients analyzing ====
 
==== Ingredients analyzing ====
 
Analyzing ingredients is not an easy task. A whole page is dedicated to [[Ingredients Extraction and Analysis]].
 
Analyzing ingredients is not an easy task. A whole page is dedicated to [[Ingredients Extraction and Analysis]].
 +
 +
On every products, you can also see how the software identifies ingredients in the text to verify its understanding.
    
As a result we can find metrics related to [[Ingredients Analysis Quality]].
 
As a result we can find metrics related to [[Ingredients Analysis Quality]].
199

edits

Navigation menu