Placing vocab words in a memory palace

I am learning Chinese, and have developed a mnemonic system(1) to easily generate images from every word. It works great, and I have already been able to amp up the speed and effectiveness of memorizing new words. However, this current system has no geographic component, so I feel it must be sub-optimal.

I would love to incorporate memory palaces into my system, but am not sure of the best way to do it. Namely, I am uncertain on how I should organize my vocab words in my palaces. I am also wondering if having a “path” or “journey” through the memory palace is helpful for such a large memorization task that has no order to it. While I’m at it, perhaps a “journey through a memory palace” is not even a good idea as suggested here. Is there another way to take advantage of the strength of geographic information?

Has anyone tried this before? These are the options I am considering:

  1. Place words strategically according to their meaning/category/part of speech
    Have physical verbs in one palace, have cooking utensils in the kitchen, etc. A journey would not work in this case since I have no idea how many words I will ultimately need to fit in a place. I am a firm believer in memorizing common words before obscure, or in approximate frequency order.

  2. Organize words alphabetically (proposed here)
    This would entail a dedicated palace for words that start with p and another for words that start with q etc. This would be somewhat redundant for me since that information is already encoded into my image.

  3. Create a journey for each batch of words as I memorize them
    Is having a journey helpful if there is absolutely no association between neighboring words?

(1) In Chinese there are almost exactly 23 sounds a word can begin with and 37 that it can end with. Then there are the four tones. I have a person for each of the 23 beginnings, a category of objects or activities for each of the 37 endings, then an emotion/color/season for each of the tones. For each word, I then use the person, the object, the emotion/color/season, and of course the definition of the word, to create a detailed image or story. For a two character word, I have two people, two items, and two emotions/colors/seasons to work with. I am not the first one to develop a technique like this. It is very similar to the Marilyn Method and I’m sure many others.

This question come very often on this forum and people have indeed different take on it. Personally I had a good experience with option 1 coupled with anki. My approach was to create/accomodate/prolonge loci on the fly in an imaginary city. Sometimes blocks were copied from places I know, sometimes generated solely from my imagination. This is indeniably less powerfull than the classic loci approach (that’s why anki is necessary here). But it comes with these two advantages : 1/almost no investment and no analysis paralysis concerning palace management 2/lot of fun.

In this approach, loci are just a way to better encode and retrieve associations when hinted with the word to translate. but retrieving words by walking in your city (and without being given any hints) is not working well. In a way with this approach you are still “anki-dependent”. From my point of view, for long term retention, we are anyway spaced repetition dependent, here again people have different opinions on this matter, so it depends on where you position yourself on this debate.

Have you seen this?

Can u tell us about ur good/easy system to generate images from words? I can use the help. Thanks.