Performance improvements

[svenseeberg]

We can improve performance by parallelizing long running tasks, for example language classification and initial message checks. Document relevance checks can also be parallelized.

A relevance check with 3 documents costs about 20s.

[dasgoutam]

Here are the results from the tests I performed for 1 query -

Query - "What can I do in Munich?"
Documents retrieved (avg) = 3
Documents retrieved after relevance check (avg) = 2

Single request
Avg. Time = 29s

Two Parallel requests

1. Example 1

Request 1: 50 sec

Request 2: 1.02 min

2. Example 2

Request 1: 52 sec

Request 2: 1.03 min

Three parallel requests*

1. Example 1

Request 1: 1.06 min

Request 2: 1.19 min

Request 3: 1.32 min