what's the best practice to write a parser for speech recognition results ?

i want to create a smiple parser for speech recognition results ,the grammar going to be complex that’s why i want some advices about how to deal with complex grammar . i have this simple example below using speech recognition api and lark parser :

from lark import Lark,Token, Transformer, v_args
from helpers.file_process import FileProcess
import numpy as np

speech_grammar = """
start              : KEY TYPE FNAME PATH?
KEY                : "create" | "delete" | "open" | "read" | "make" 
TYPE               : "file" | "folder" 
FNAME              : (/[a-z]/)+ 
PATH               : (/[a-zA-Z]/)+":"(/[\\/]/)+DIRNAME
DIRNAME            : ((/[a-zA-Z0-9]/)+(/[\\/]/)?)*

%ignore (" ")+
"""

speech_parser = Lark(speech_grammar)

def run(input):
    try:
        speech = speech_parser.parse(input)
        instr=np.array(speech.children)
        return FileProcess(instr).request_process()
    except Exception as err :
        return err

def main(input):
    return run(input)

def test():
    print(run("make file test d:/test"))
    print(run("delete file test"))

i will add more terminals (Vt) in the grammar part , any suggestions or advices would be much appreciated ?

[erezsh]

It sounds like you’re asking about the lingual interface, rather than parsing.

Maybe if you provide some examples of the output you’re getting (or hoping to get) from the voice API, we can tell you how to parse it.

I would suggest considering a “DSL” for the paths that’s more natural for speaking, like “drive d folder myfolder”

[Morojs]

the grammar describe how users interact with speech api , we can’t use “create file test d:/myfolder” like that because the api doesn’t understand the path and it doesn’t have the concatenation between each path item, is there any way to make it easier to understand and use for speech api?