What is the key difference between graph-based QA and QA by semantic parsing?

Question

Folowing D. Jurafsky handbook, i struggle with the chapter on QA. It defines 2 paradigms of QA knowledge based which are:

• graph-based QA (1)
• QA by semantic parsing (2)

1. For (1) it takes the exemple: "When was Ada Lovelace born ?". The process is first to find the Entities (here Ada Lovelace) and then it shoud determine the SPARQL query from the question. So the system hast to compute something like:

• "When was Ada Lovelace born ?"-> birth-year (Ada Lovelace, ?x) [I am wondering how we can achieve thjis second step ...]

Then it just compute the query in our Knowledge Base (with help of SPARQL)

2. For QA by Semantic Parsing, the "semantic parser" directly from question to query (query like SQL, SPARQL, etc...) based on a trained corpus.

For me it is exactly the same thing ? The process is the same and the hard part s to translate Question in Query!

I miss something if anyone could help me ...

Thanks