Features Engineering for Deep Learning NLP

I want to build deeplearning NLP model with datasets below. The dataset I have contains diseases and their corresponding symptoms like this

Disease Symptom_1 Symptom_2 Symptom_3 Symptom_4 Symptom_5
Fungal infection itching skin_rash nodal_skin_eruptions dischromic_patches None
Fungal infection skin_rash nodal_skin_eruptions dischromic_patches None None
Fungal infection itching nodal_skin_eruptions dischromic_patches None None
Fungal infection itching skin_rash dischromic_patches None None
Fungal infection itching skin_rash nodal_skin_eruptions None None

Before using it for Natural Language Processing (NLP) tasks, I want to preprocess the data to represent symptoms in a suitable format for my deep learning NLP model. I am considering two feature engineering options:

  1. List of Symptoms for Each Disease:
    I could create a new dataset where each row corresponds to a disease, and the symptoms are listed as a string. For example:
Disease Symptoms
Chronic cholestasis itching, yellowish skin, nausea, loss of appetite, abdominal pain, yellowing of eyes
Chronic cholestasis itching, yellowish skin, nausea, loss of appetite, abdominal pain, yellowing of eyes

or,

  1. Transformed Symptom Descriptions:
    Alternatively, I could transform the symptoms into a single string description for each disease. For example:
    "Fungal infection. Itching. Reported signs of dischromic patches. 
    Patient reports no patches in throat. 
    Issues of frequent skin rash. 
    Patient reports no spotting urination. 
    Patient reports no stomach pain. 
    Nodal skin eruptions over the last few days."
    

My question is, which kind of feature engineering should I use that would work better for my deep learning model? I would appreciate input and insights from the community to help me make an informed decision.

Hi,
In my opinion try using the linear Regression model, which will help you create a function. Also try minimizing the cost of the function.
Thanks