Big hat tip to Daniel for sharing with me this topic.
It shows that at least when it comes to the English language, the placement of emphasis on particular words has a lot of importance.
Here’s an example from a Computer Science book chapter on NLP (you can check it out here):
It clearly shows how a change in where you put an emphases changes the meaning of the sentence entirely. Even though the intonation is subtle, the difference in meanings is not.
The issue with many text-to-speech implementations is that they don’t involve the placement and weighing of emphasis when requesting an output. While many services allow for this type of modification, developers might be easily overwhelmed by the daunting task of annotating text with special tags.
One of the new frontiers for AI and natural language would be to automatically detect the tone of content and properly markup text. This might not be far off.