Rewrite probabilities learned for English, averaged over the last 4 epochs on 𝖾𝗇 treebank (blue bars) or 𝖾𝗇_𝖾𝗌𝗅 treebank (orange bars). The header above each figure is the underlying punctuation string (input to NoisyChannel). The two counts in the figure headers are the number of occurrences of the underlying punctuation strings in the 1-best reconstruction of underlying punctuation sequences (by Algorithm 1) respectively in the 𝖾𝗇 and 𝖾𝗇_𝖾𝗌𝗅 treebank. Each bar represents one surface punctuation string (output of NoisyChannel), its height giving the probability.