Fun with labels

Mineral Matters #447

Jolyon Ralph, founder and owner of Mindat, recently posted about a proposed Mindat app facility that would allow you to take a photo of a specimen, and a photo of the label, and automatically create an entry in your Mindat mineral catalogue (if you use that for cataloguing). The function uses AI in part. You can find out more here. Jolyon is looking for beta testers.

That got me thinking about how I might be able to use AI to help me (I don’t use Mindat to catalogue my collection).

So I ran a few tests, in this instance using Microsoft Copilot (which uses ChatGPT). But you could probably use any one of a number of AIs.

My first test was a very simple one. This would help me to get the full locality name, as per Mindat.

• Query: If I were to give you a mineral locality (in its short form), would you be able to suggest possible match(es) from Mindat?

• Copilot response: Absolutely! I’d be happy to help you with that. Just provide me with the mineral locality in its short form, and I’ll do my best to suggest possible matches from Mindat. Let’s get started!

• Query: Ok. Lets try this one. "Lake Boga".

• Copilot response: Based on your query, it seems like you’re referring to the Lake Boga granite quarry in Victoria, Australia. This locality is well-known for its mineralogical significance. Location: Lake Boga, Swan Hill Rural City, Victoria, Australia. (The link to the page was also provided)

A couple of other tests also worked out ok. Now for the labels. The idea was to upload a label and get Copilot to read it.

I tried a few different ones (see photos below). These included:

1. A printed label - Grandviewite.

2. A hand-written label (my handwriting) - Burra

3. A European handwritten label - Strontianite

4. A Japanese label - Yugawaralite

The results, as per Copilot:

1. Based on the label, the mineral specimen is Grandviewite from the Grand View Mine, Horseshoe Mesa, Grand Canyon, Arizona, USA. This is indeed the type locality for Grandviewite.

2. Based on the handwritten label, the minerals listed are Azurite, Libethenite, and Chrysocolla from Burra. This suggests the specimen is from the Burra Burra Copper Mine (Burra Mine; Monster Mine) in Burra, North Mt Lofty Ranges, South Australia, Australia.

3. Based on the handwritten label, the mineral is Strontianite from Unternurrbach, Lenggries, Upper Bavaria (Ob.-Bayern), Germany.

4. Based on the label, the mineral is Yugawaralite from Yugawara Hot Springs, Kanagawa Prefecture, Honshu, Japan.

The first three worked well. Number three in particular would be very useful. I quite often struggle with some of the handwritten European ones.

Number 4 wasn’t quite so successful. Copilot used Yugawaralite to provide the response. What it should have done is translate from Japanese first. I picked up on this after I had asked for a translation. The original response gave the Type Locality in the Kanagawa Prefecture, whereas the translation gave the Shizuoka Prefecture.

As it turns out, the Shizuoka Prefecture label locality isn’t in Mindat anyway! But it is in Alfredo Petrov’s Encyclopedia of Japanese Minerals, or at least a pretty close approximation.

Here are the translations and Romaji for the two lines of Kanji/Kana on the label:

湯河原沸石 (Yugawaralite):

• Translation: Yugawaralite

• Romaji: Yugawarafussekki

静岡県田方郡土肥町大洞林道:

• Translation: Ōdō Forest Road, Toi Town, Tagata District, Shizuoka Prefecture

• Romaji: Shizuoka-ken Tagata-gun Toi-chō Ōdō-rindō

The description of the occurrence according to Alfredo is “Shizuoka: Transparent, euhedral, thin tabular yugawaralite crystals to 3cm long, colourless to white, with wairakite (tiny white trapezohedra), laumontite, quartz and epistilbite (or dachiardite?) at Ohbora (Ohora), Toi, in vugs in veins from 2 to 10cm wide cutting propylitized andesite and basalt of the Miocene Yugashima formation.”