Hi professor, would you please check out these (possible) typos below?

1. Section 4.4.1 How to Measure Progress?
   https://github.com/danqi/thesis/blob/master/chapters/rc_future/questions.tex

How to tell how far these computer systems are sill from genuine human-level reading comprehension?

Would it be still?

2. Section 5.2.4 Distant Supervision
   https://github.com/danqi/thesis/blob/master/chapters/openqa/system.tex

Finally, the size of SQuAD is rather small (80k training examples). It should further improve the system performance if wen can collect more training examples.

Is it we here?

3. Section 6.2 COQA: A Conversational QA Challenge
   https://github.com/danqi/thesis/blob/master/chapters/coqa/dataset.tex

2. It still supports reliable automatic evaluation with a a strong human performance.

It seems there should have only one a.

Finally, thanks for this fantastic work.