1

I have lots of text documents structured as

{
{
    Item1=[
            {a1=1,
             a2=2,
             a3=3},
            {a1=11,
             a2=22,
             a3=33},
            {a1=41,
             a2=52,
             a3=63},
            {a1=19,
             a2=29,
             a3=39}
    ],
    Item2=[
            {a4=1,
             a5=2,
             a6=3},
            {a4=11,
             a5=22,
             a6=33},
            {a4=41,
             a5=52,
             a6=63},
            {a4=19,
             a5=29,
             a6=39}
    ],
}
}

Now this can be formatted into two csv's as

Item 1

and

Item 2

I can write regex parser for this but is there a way by which a neural network or deep learning model can be trained for this, which can create these csvs?

The above example has been indented for better visuals, the raw text looks something like

{{Item1=[{a1=1,a2=2,a3=3},{a1=11,a2=22,{a1=41,a2=52,a3=63},{a1=19,a2=29,a3=39}],Item2=[{a4=1,a5=2,a6=3},{a4=11,a5=22,a6=33},{a4=41,a5=52,a6=63},{a4=19,a5=29,a6=39}]}}
Sim101011
  • 111
  • 2
  • 1
    Hi and welcome to this community! Why do you wanna use a neural network for this task? If you can solve it exactly (and not approximately) and efficiently, then you don't need neural networks or machine learning. – nbro Jan 10 '20 at 13:29
  • Hi @nbro, thanks for the welcome. The structure I mentioned above is just an example. The variations are infinite. Sometime row separators are {, sometime they are [ and sometime custome delimiters like @#@#@. If a model learn to identify inherent structure, the need to write separate parser will end. That's why I want to train a NN model on this. – Sim101011 Jan 10 '20 at 14:24

0 Answers0