There was a thread recently in Russian on using BEM in medium-sized projects written by Nikolay Gromov who complained that writing BEMJSON requires more time than writing HTML with a help of emmet.

I don't feel the same way about time spent on writing code not in percentage relating to time spent on debugging and thinking over not in absolute terms but I thought that spending time for writing analogue for emmet for BEMJSON is good enough task to spend an evening.

It is how this package bemmet emerged that can expand abbreviations like b1>__e1*2>b3_theme_islands+_state_active{hello} into BEMJSON:

{
    block: 'b1',
    content: [
        {
            block: 'b1',
            elem: 'e1',
            content: [
                {
                    block: 'b3',
                    mods: { theme: 'islands' },
                    content: {}
                },
                {
                    block: 'b1',
                    mods: { state: 'active' },
                    content: 'hello'
                }
            ]
        },
        {
            block: 'b1',
            elem: 'e1',
            content: [
                {
                    block: 'b3',
                    mods: { theme: 'islands' },
                    content: {}
                },
                {
                    block: 'b1',
                    mods: { state: 'active' },
                    content: 'hello'
                }
            ]
        }
    ]
}

Using options you can configure custom naming — powered by bem-naming. stringify() method supports indentation options and quotes — powered by stringify-object.

You can try bemmet in action using online demo. You can also install it right away into your editor using plugins for Sublime Text and Atom.

Would be nice if you can commit with plugins for some other editors.

Good luck!