Reading backwards doesn’t make a lot of sense in general language structure.
“If you have a sentence such as this”
Reading it backwards still needs to be
“If you have a sentence such as this” rather than “this as such sentence a have you if”
So in order for you to read language sensibly backwards you need to arrange it in your mind in such a way that what you process first is actually last. This may require you to keep track of the entire line.
The alternative is to read concepts, which can be read backwards to some extent.
“If you have a sentence such as this” in image concept = floating string of text in your ownership(can be represented by hand) near a wall of sentences. To explain how this can be read backwards, you convert the sentence to images:
this = reference to something
such as = instance of reference to reference to something
sentence = this something that was being referenced
a = one of
you have = yours
if = condition
In reality when you read it visually then you would have your eyes move all the way to ‘sentence a’, from this without visualizing and then have ‘such as’ and ‘this’ turn to images by visualizing at that point, then continue moving along while having representations for something being yours (usually an instance of such concept).
When you recall it backwards in concept you will say something like :
I have a sentence like that one.
When a condition occurs ‘e.g’
“If you have a sentence such as this” to “If you have a sentence such as this then X”
the above changes to
If I hold a sentence like that one then X.
These are naturally not ‘exactly’ the same things because you are using instances for have, so you may forget that it’s have instead of hold or own etc. It entirely depends on how you want to standardize turning the text into images. 100% accuracy needs 100% accuracy in turning an image to the exact word and the reverse. While you can get better at this, at current it isn’t believed to be enough to recall accurately, but if concepts are enough then this can work.
Alternatively you can make a language of images having images for every word you would use in English, which is time consuming but should work. You would be able to leave out all words that have physical representations (such as tree). Again alternatively you could even have a system to help you identify what word it is. You could carve the first two letters into images like ‘have’ with a hand so you know it is have and not hold or own.
However, for the general text to image concepts:
My advise to training this is to turn text into images while looking at the text, so
smile into “”, brain into “”. This is my personal opinion but if you keep the images as small as that it is more manageable. You should practice this like you train regularly until you can do it instantly.
Then you should practice doing multiple statements to images at once accurately and instantly.
Finally you should learn to have the non-image statements guide how these images are represented, after which you should be able to do this. Once you can do all that you could experiment by combining this with existing memory techniques to bolster your memory but for now that should be enough.