Doesn’t sound too much more likely to me. There are definitely a huge number of typos in the training data.