The files are:

conversations.tsv - The initial conversations, 5 for each number of targeted information units. Each line has a conversation id, the number of targeted information units in the conversation, and a sentence.

Summary.tsv - Summaries collected across all conditions using crowdsourced targeted summarization techniques. The fields are:

  conversationId	- A unique number for each conversation, referring to the conversation.tsv file

  Number of targeted 
  information units
  per conversation	- The number of targeted information units that should be summarized in the conversation
  
  Condition		- One of the following labels: baseline, highlight, pin-refine, each refers to one 
			  crowdsourced data collection workflow  

  Correctness		- The adjudicated correct score (correct, incorrect, none)

  Recall		- The percentage of correct targeted information units in the summary

  Precision		- The percentage of incorrect targeted information units in the summary

  F1			- The harmonic mean of precision and recall

  Fluency		- One of the following labels: fluent, unnatural conjunction, ambiguous, grammatical 
			  error, broken
  
  timeOnPrimingTask	- The time taken to finish the priming task, in seconds. In the baseline condition, the value is always 0 
			  because there was no priming tasks. In the highlight and pin-refine conditions, a value of 0 is given 
			  for the first conversation a worker wrote (when they also had to read the instructions).

  timeOnSummarization	- The time taken to write the summary, in seconds. In the baseline condition, a value of 0 is given for 
			  the first conversation a worker wrote (when they also had to read the instructions).

  totalTime		- The time taken to finish the priming task and write the summary

  Summary		- The summary