UserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/53.0.2785.116 Safari/537.36
Platform: 8530.81.0 (Official Build) stable-channel winky

Steps to reproduce the problem:
Call API function chrome.tts.speak(...) with custom event listener as described here ("Listening to events" section, https://developer.chrome.com/apps/tts)

What is the expected behavior?
For each word in spoken text we expect to receive two callbacks: one - with type "word" and another - with type "end"
Each callback should contain relevant charIndex of word being spoken (as it was in Chrome OS 52.02743.116 Stable Channel Update) 

What went wrong?
We receive only two callbacks (one - with type "start" and another - with type "end") regardless of number of words in text.
For example:
for "Hello, my awesome world." we receive first callback "start" with charIndex = 0 and second callback "end" with charIndex = 24
We don't receive in-between callbacks at all.

Did this work before? Yes Chrome OS 52.02743.116 Stable Channel Update

Chrome version: 53.0.2785.103 (64-bit)  Channel: stable
OS Version: 53.0.2785.103 (64-bit)
Flash Version: 

Highlighting text when speaking is a key feature for our products as they are targeted for people with special needs who experienced problems with reading. So it's important for us to have it's working.
It worked well in Chrome OS 52.02743.116 Stable Channel Update