{"reviews":[{"answer_id":"a001","verdict":"correct","rationale":"The paper reports the Rei2016 error detection system, trained on the same FCE dataset, for comparison.","source_span_ids":["s00023","s00025"],"issues":[]},{"answer_id":"a002","verdict":"correct","rationale":"The paper also compares systems using artificial data against a model trained only on manual annotation from the FCE dataset. The question permits this baseline as well as the Rei2016 comparison.","source_span_ids":["s00021","s00023","s00025"],"issues":[]},{"answer_id":"a003","verdict":"correct","rationale":"A model trained only on manual annotation is a baseline used to assess the improvement from artificial data; the training dataset was FCE.","source_span_ids":["s00021","s00025"],"issues":[]},{"answer_id":"a004","verdict":"correct","rationale":"A model trained only on manual annotation is a baseline used to assess the improvement from artificial data; the training dataset was FCE.","source_span_ids":["s00021","s00025"],"issues":[]},{"answer_id":"a005","verdict":"correct","rationale":"A model trained only on manual annotation is a baseline used to assess the improvement from artificial data; the training dataset was FCE.","source_span_ids":["s00021","s00025"],"issues":[]},{"answer_id":"a006","verdict":"correct","rationale":"The paper reports the Rei2016 error detection system, trained on the same FCE dataset, for comparison.","source_span_ids":["s00023","s00025"],"issues":[]}]}