#####################################################################

ML/NLP/NLU Challenge

Skylar Labs

#####################################################################

Hi folks, this is a small ML/NLP/NLU challenge to see if you've got what we need at Skylar Labs as a potential candidates for a position as ML/NLP/NLU Software Engineer.

##Task

The task is simple, build a Retrieval System to pick the right answer to a certain question, from a dataset.

An user ask a question, your model, process the question and pick the most suitable answer from a knowledge base dataset, in which the right answer is.

The user input can be slightly(or even totally different, that can be cover later on) different from the based question

Example:
In dataset:
Q: what's your address
A: I don't have any address
User input:
Q: where are you located
A: I don't have any address

##Data

You will be provided 4 files to work with, 2 files for each steps(training and test):

• train_dataset.txt, train_dataset_2.txt, are the dataset for training, it's formatted in a way to be easy to process:
  1- first line is a question
  2- second line is the answer to the question

  you can take it another way:
  1- Odd line number is a question
  2- Even line number is an answer

• test_dataset.txt, test-data.txt are the dataset for testing, it's formatted in a way to be easy to process:
  1- first line is a question
  2- second line is the right answer to the question
  you can take it another way:
  1- Odd line number is a question
  2- Even line number is a right answer
  You can use the test file to benchmarck your system, by comparing the answer ouput by your system and the answer expected

  That is useful to benchmarck your system, and tailor it

##Expectation

What is expected is to get the highest accuracy possible, in the near future, the system will evolve in memory awared system, where the retrieving will be based on previous answers and questions, so give your all, cause you're about to build one of your best and sweetest piece of work.

Good luck fellas, may the force be with you !