Skip to main content

Table 4 The left and right tables show the performances of the HRL agent and the proposed agent (A-SIDDA), respectively

From: Symptoms are known by their companies: towards association guided disease diagnosis assistant

Self-report

Self-report

Patient: Hi, I am having difficulty in breathing. What could be its reason ?

Patient: Hi, I am having difficulty in breathing. What could be its reason ?

Conversation

Conversation

Agent: Ok, are you feeling shortness of breadth ?

Agent: Ok, are you feeling shortness of breadth ?

Patient: Yes, I am feeling shortness of breadth.

Patient: Yes, I am feeling so.

Agent: Do you have skin rash ?

Agent: Have you vomited recently ?

Patient: No, skin rash.

Patient: Yes, I vomited a few times.

Agent: Have you vomited recently ?

Agent: Have you felt nausea ?

Patient: Yes, I vomited a few times

Patient: No, I have not felt so.

Agent: Have you observed abnormal appearing skin ?

Agent: Do you have cough ?

Patient: No, I am not suffering from skin issue.

Patient: No, I do not have cough.

Agent: You are most likely to have Fluid overload

Agent: You are most likely to have Fluid overload

Diagnosis success : True, Dialogue length : 4, Match rate : 0.5,

Diagnosis success : True, Dialogue length : 4, Match rate : 0.5,

DC : 1, IReS-1 : 0.3337, IReS-2 : 0.9304

DC : 1, IReS-1 : 0.5271, IReS-2 : 1.121

  1. The significance of the agents’ investigated symptoms differs substantially, yet they perform identically in all current evaluation metrics (success rate, dialogue length, AMR, and DC). The proposed evaluation metric, IReS, successfully captures the relevance aspect and rates the symptom investigations accordingly
  2. The underline signifies the medical entity which is being inspected