Sentence intonation is often described as having a rhythmic, even melodic character. It is the ups and downs in pitch we use as a sentence rolls merrily along. But lists stand apart and have a special nature all their own. Let's look into how Rime gets list intonation right...

Lists as a glitch in intonation

Unlike standard sentence intonation, lists function like a glitch in the intonation system, sort of like a skipping record.

At Rime, we've created a model that does list intonation better than any other TTS offering in the world. We do this because lists are immensely important in the world of business. We use them when reading out numbers, spelling, and in ordering. Take a list to the following example samples:

Lists can drone on indefinitely, with each chunk in the list repeating the same sort of rising intonation until the final item in the list, which generally gets a falling intonation. Rime voices get this right, but why? First we need to explore what intonation is to begin with.

What is Intonation?

In spoken language, we raise and lower our pitch as we produce a sentence to convey many different things. This is called intonation and it's extrememly important for modern TTS to get right.

Intonation is made up of pitch. And since pitch is a continuous variable, it's often representented impressionistically with lines indicating the rise and fall of pitch over the sentence. We can see this below in a few examples:

Even the title of Rime hero Dwight Bolinger's book on the subject gets into the fun of visual intonation representation!

In list intonation, each successive item in the list receives a rising pitch, until the ultimate item, which receives a lower pitch, indicating that the list is finished.

But this is still just a description. In the next section we describe why list intonation works this way.

Why care?

Intonation is a vehicle to convey countless things beyond the compositional meaning of the string of words in sentence. It can be used to express emotion or sarcasm, focus attention, and indicate whether a sentence is a question or not. The last case is by far the most common and fundamental use of intonation in English. Below, the string of words "you like the cake" is instantiated with differing intonation, clearly distinguishing a declarative sentence from a interrogative.

Question intonation in English and many other languages, ends with a rising pitch. This is generally linked with a notion of 'openness' or a sense that things are not yet concluded (see Alan Cruttenden's classic book on this for more). Basically: a question is usual not the end of a conversation, it awaits an answer. Whereas falling intonation generally indicates finality.

In fact, the fact that rising intonation is used to indicate that things are not yet concluded not only explains its use in questions, but also its more generalized use in what is often called up-talk (or it's fancier name high rising terminal intonation). That is, up-talk is not the speaker being unsure of themselves necessarily, but rather an indication that they are seeking feedback or confirmation or affirmation of some sort, in the same way that a question seeks an answer.

So we raise our pitch when we want to indicate that something isn't finished, and lower it to signal completion. This makes lists an interesting scenario: you need to repeatedly indicate that you're not done; after each item in a list, there is another one coming down the pike, until you reach the final item.

The result is a series of rising tones followed finally by a falling one: List intonation simply falls out from the general use of rising and falling pitch as a repetitive glitch in our intonation system!

Wrapping Up

Getting realistic human conversation right, means getting lists right. Lists are incredibly common, especially is business scenarios. And at Rime we understand their importantance and offer the best solutions to this on the market. Contact us to hear more about our offerings or just to chat, talk, converse, gab, prattle, jabber....