I want to send multiple line text input to conversation service in that i want response that will check whether input is in positive tone or in negative tone.
Method 1: You can create two entities, one for positive and one for negative. Populate the values for each with positive and negative statements.
Make sure to capture words like not, yes, worse, sure, sad, happy etc. There larger your set, better the answer.
I am not sure if this method captures double negation. IBM claims to having Natural Language Classifier running behind the scenes which will give you positive results for unseen samples.
Method 2: You can integrate watson personlity insights with conversation to get a direct sentiment analysis for every input.