Abstract:
In this information age the number of internet users are growing rapidly. Now a
days people rst search internet if they face any health hazard rather than asking
a doctor for health related advice as online medical help or health care advice
is easier to grasp. Sometimes, people give less importance to minor symptoms
which may cause serious health hazards. In this context, online health advice can
be instant bene ciary. Moreover, existing online symptom checkers give possible
sense of disease but these systems are not reliable enough. Also existing systems
are not interactive and time consuming. Herein, we propose an automated
disease identi cation system that takes unstructured user input and provides a
list (topmost diseases that have greater likelihood of occurrence) of probable diseases.
We use Conditional Random Field and Support Vector Machine to detect
the word phrases and to classify the class labels.By not considering demographic
information, it gives 4.603% accuracy improvement whereas, considering demographic
information we get slightly better performance with 5.783% accuracy
improvement.
