Supplementary materials for the ACL-IJCNLP 2021 submission 3211
"Accelerating Text Communication via Abbreviated Sentence Input"

enron-abbrev.txt          - Tab delimited file containing data from the reported Amazon MTurk free-form abbreviation user study
rnnlm-rescore-drop0.5.txt - Tab delimited file containing recognition results using a vowel drop probability of 0.5 and RNNLM rescoring
rnnlm-rescore-drop1.0.txt - Tab delimited file containing recognition results using a vowel drop probability of 1.0 and RNNLM rescoring
study-sentence.txt        - Tab delimited file containing user performance for each phrase in the Sentence condition
study-word.txt            - Tab delimited file containing user performance for each phrase in the Word condition

The Amazon MTurk file contains the following fields:
  worker_id        - Anonymized number of the worker completing the tasks
  task_index       - Integer indicating the order they complete tasks
  original_id      - Original phrase ID from the Enron mobile dataset
  original_text    - The original text passage they were asked to abbreviate
  abbreviated_text - The abbreviated version written by the worker
  backspace_count  - Number of times they used the backspace key on this phrase
  start_time       - Millisecond time they started input of this phrase
  end_time		   - Millisecond time they completed input on this phrase

The recognition result files contain the following fields:	
  sentence    - Integer index of the sentence
  reference   - The reference text for the sentence
  recognition - The 1-best recognition result of our decoder
  input       - The nearest key to each touchscreen tap observation in the data
  cer         - Character error rate for each phrase
  wer         - Word error rate for each phrase 
  ser         - Sentence error rate	for each phrase
  decode_time - Decode time in seconds to expand the abbreviated input
  

The sentence condition file contains the following fields:
  userid               - Integer index of the user 
  phraseid	           - Integer index of the phrase 
  condition	           - Name of the condition 
  reference	           - The reference text for the phrase
  input	               - The expanded text from the abbreviated input
  abbreviated_input	   - The abbreviated text written by the user
  oracle_abbreviation  - Abbreviated reference by removing space between words and mid-word vowels
  start_time           - Millisecond time users started input of this phrase
  end_time		       - Millisecond time users completed input on this phrase
  cer	               - Character error rate for each phrase
  wpm                  - Entry rate in words-per-minute for each phrase
  backspace_count      - Number of times a user used the backspace key on this phrase
  
The word condition file contains the following fields:
  userid               - Integer index of the user 
  phraseid	           - Integer index of the phrase 
  condition	           - Name of the condition 
  reference	           - The reference text for the phrase
  input	               - The input text from the user using word suggestion
  start_time           - Millisecond time users started input of this phrase
  end_time		       - Millisecond time users completed input on this phrase
  cer	               - Character error rate for each phrase
  wpm                  - Entry rate in words-per-minute for each phrase 
  backspace_count      - Number of times a user used the backspace key on this phrase
