Integrity of the Food Item Database

A complete user generated food database is great alternative to the USDA or FDA’s lack there of. It does, however, introduce the possibility of human error. One solution that might help would be to add a reference field.
This would be something like a UPC bar code, or a URL listing.
Then allow each user to verify the products accuracy or inaccuracy’s.
You would then be able to display (and sort by) this report in a food search.
I don’t know if many users would actually verify a contents accuracy.
I have on a couple of times verified an entries inaccuracy’s.

I agree that something should probably be worked out like this. It’s kind of frustrating to look up an item like coke or coca-cola and have 15 choices, many with different stats. You end up just selecting one “in the middle” hoping it’s right. I’d love to see the junk entries cut out and the accurate ones frozen.

Gyminee took the first step towards a more accurate food database by letting users edit foods wiki-style. There are still a lot of changes and improvements coming down the pipeline, and those improvements should address most (if not all) of the suggestions mentioned above. Great suggestions though, definitely keep it up.

The wiki is a good start, however it is an implementation I’ve been cautious in using. I’ll explain, and perhaps you’ve already addressed this in your upcoming pipeline.

Right now there is a food item for Sour Cream and Onion Baked Lays that has all of the nutritional information and serving size entered correctly, however the calories are doubled. Let’s say there is a user who ate two servings of this food item every day for the last week. If I edit the item’s calories from 220 to 110, that users nutrition log will now be 220 calories lower for every day last week. In many cases this would change their nutritional letter grades over night. This could be a fairly confusing user experience.

There also seems to be quite a variance on standard foods in different countries (as you would expect) Kelloggs All-bran is an example. Also It would be great to be able to remove “recipe” type foods when searching so you only have the base ingredients. ie Searching for Mushroom you would only have mushrooms not soups etc.

When editing foods the % of minerals etc is this of the actual food or is it RDI? Can you also have an option of adding energy by Calories or Kjs?

I don’t know if the thread has gone stale, but these are some good ideas. When I add an item I try to do two things.

1. Include the source, particularly if I got it online, then I put the URL in the item comment box.

2. Comment the serving size—ie “1 cup (two per can)”.

One of the problems I have is that the serving sizes, often reported in grams, don’t mesh in my head. What does 50g of rice look like, or hamburger? How about adding optional fields by the serving size box: container (ie box, can, pouch) and serving per container?
—c

When editing foods the % of minerals etc is this of the actual food or is it RDI? Can you also have an option of adding energy by Calories or Kjs?

It is of the actual food. Well, let me restate that. The main nutrients (cals, fat, protein, etc.) are for the food item. All of the % of minerals are still of the food item, but they are input and displayed as a % of RDI. (I think I said that correctly). We do plan to display in Kj in the future.

Good suggestions @cgk.

If I edit the item’s calories from 220 to 110, that users nutrition log will now be 220 calories lower for every day last week.

not sure that it works this way. i had the same concern, so i did a test. i changed the calories in an item that was already in my log and it didn’t seem to effect my totals until i removed it from my existing day and readded it.

so, it looks to me like when you say “i ate this” it adds the amounts to your totals, but doesn’t keep any other connection to that item in the database other then the note that it was added (with the link to remove it, which probably just subtracts the amounts it ‘remembers’ adding for that item)

The data actually is attached to the food item, so this could modify what other people have eaten. Of course, the idea is that the food would be modified to be more accurate, so it wouldn’t necessarily be a bad thing.

What about getting the official nutritional information from restaurant chains and marking it as such? It would be much easier to pick a restaurant chain and then add menu items to your diet.

I also like adding the bar code information to certain items. I’m sure some of you have seen the Shop Savvy demos on Android. Very cool tech that would be easy to implement I think.

With regard to the UPC/barcode, I think it would be the best way to track individual items. The key with any database is a unique identifier. The UPC system provides that.

Implementation could be as simple as an additional field in the food addition page and then allowing the search algorithm to include that field. And displaying the UPC on the item detail page would be nice, so when searching for a food by name, you can easily see 1) if the number matches (nice when you have the item in front of you) and 2) whether or not it needs to be added.

I’ve recently tried to remember to place the UPC into the description field, but it doesn’t seem as if the search takes that field into account. Of course, not everything in the world has a barcode. But this would be a good start.

+1 for a barcode field. It’s annoying to have 10 pages of results for a given product and to look at it. It will be a way to be sure that we have chosen the right food.

I would love a barcode option, especially since you can still get old CueCat USB barcode readers off the internet for like $20. You have to tweak them to spit out “normal” barcode numbers, but it’s extremely easy and handy once you do. I use mine for tracking books I own/read.

While we’re at it: the sorting of meals into food groups is very much a US only thing – I don’t measure anything in cups, but know that my country’s recommended veggie intake is 6 portions à 100g pr. day (compared to UK’s 5 portions à 80 gs – just goes to show how flexible these recommendations really are)

So I would love to be able to edit the recommended amount of veggies, grain etc. – in metric!

Another thing that is annoying me no end is the inability to move a food item from unassigned to a meal – or just being able to edit its veggie, grain, milk values without adding it to a meal. (it was one of the things that made me go Pro, but on day 2 I’m already wondering whether I’ve wasted my money, if it’s going to be this bothersome).

+1

I would love to move a food item from unassigned to a meal. That would also allow me to copy an unassigned food item or group of items to another day.

One of the reasons I am less consistent with submitting my nutrition is that it’s a pain to put everything in every single day. If I could copy over meals without having to plan them, that would help me immensely.

Thanks for the suggestions everyone. Those are all great ideas. We’ll try to incorporate as many as we can the next time we iterate/improve the nutrition and meal planning portion of the site.

Any date for that ? ;)

As a software developer, I really like the ability to see when something changes, who changed it, what exactly they changed, etc. – time diffs. I understand the increase in cost of data storage, but I think it would be awesome if foods were “versioned”, so to speak, similar to how wikipedia does (ex. http://en.wikipedia.org/w/index.php?title=Nutri…). That way if someone adds a food to their dairy or favorites, they might just add that particular version of the food – subsequent changes couldn’t possibly affect them (ex. the “oldid” param in http://en.wikipedia.org/w/index.php?title=Nutri…). Then there could also be some UI (a sparkle on the food item or something) to indicate that new updates have been made to the item. Then the user can check the changes and decide if they want to move to the new “version” of the food or not. This would also make it really easy to “rollback” food to previous versions if someone changes it maliciously, and the person responsible could get a ding on their “reputation” (another concept that might be useful). Similarly, people could praise a change or a new food addition, which boosts the person’s reputation.

Also, perhaps it would be useful to include a sort of “food rank” metric in the ranking algorithm for the search results. That is, more popular foods (foods more frequently “eaten” or something) are ranked higher than other foods. This should most likely be used with some form of TF-IDF to be most effective.

• Versioned foods
• Diffs between versions for accountability
• Reputation for content submitters/editors
• Popularity of food affects ranking in search results

I think with User Generated Content like this, it would be efficient to let the users also control the ultimate fate of each individual items.

Let users add items, and let other users rank them (thumbs up / thumbs down). If you click on a bag of certain potato chips and the calories are incorrect, thumbs down… if it’s all correct, thumbs up. If you click thumb’s up on everything you find correct, soon enough the ‘good’ items will have pretty positive rankings.

Then when you search for these potato chips sort by highest number of thumbs up. I think this would be nice :)

Also, I read above about using UPC barcodes. This is an interesting idea, but I think it will definitely take off if you implement this into an iphone / android / etc native apps.

Some apps on iphone/android already make use of the camera to read UPC barcodes, so this should be very possible to implement.

If you add this feature I would definitely use it. Imagine: Gyminee, Menu > Add Food By Barcode, Scan, that’s it. That sounds easy :)

Any date for that ? ;)

Sorry, no date… we’re working as fast as we can, but have a lot on our place :-)

@CornellEngr2008 – The versioning concept is interesting, but my concern would be that it could confuse some users.

Also, our search algorithm does take into account the “popularity” of the food. i.e., the more it is recorded, the higher it will show up in search results over time. So, in a way simply recording the best search matches is like giving it a thumbs up/down.

The barcode is possible (and a really cool idea), but it wouldn’t be trivial to implement. If we decide to do it, I think it would first be seen on a native iPhone app.

I think too much is assumed when using database term like versioning to describe a simple history feature. CornellEngr2008 makes reference to a great UI feature in MediaWiki’s(Wikipedia) history tab. Most users do not use the history tab, and have no interest in doing so. Even the most uninformed of users are not burdened by the complexity of the history interface, because they do no use it.

As far as the barcode is concerned. Rather than developing a specific barcode/upc attribute for all foods, I would propose adding more generic metadata. There could be a multitude of extensible attributes: barcode, gluten-free, glycemic index… These could all be searchable with triggers in the search string “Detour Energy Bar -barcode:733913003751 -lowGlycemic:true”

I have another suggestion to improve the user-edit capability. For foods that have been locked so that they cannot be edited, leave the ability to add a picture unlocked at least until a picture is added. For me, the pictures are useful, but the ‘fork-with-a bite-of-salad’ is not.