PoLaR Labels

On this page, we briefly discuss the four core tiers of PoLaR, and what their labels signify, with an example. (The sound and annotation files [named reindeer_are_running.wav and reindeer_are_running.TextGrid] can be downloaded from the OSF repository.)

Let us begin with the Prosodic Structure ("PrStr") tier. In this tier we find * labels in intonationally prominent syllables, and ] labels at the end of a phrase-final word. Thus this annotation tells us that there are two phrases ("and then I said" and "the reindeer are running"), and four prominent words ("I", "reindeer", "are", and "running").

The next tier is the Points tier, which annotates where the pitch track is changing slope (i.e., where there would be "turning points" in a straightline approximation of the pitch). Each label on the Points tier is time-aligned to one of these turning points. This is graphically represented in the annotated version of this figure, where the turning points labelled on the Points tier correspond to the red dots overlaid on top of the pitch track:

Note that there is a long flat plateau during "reindeer are running". This indicates that the labeller is ignoring the slight movements in pitch during this stretch of time. It also highlights the fact that a PrStr label does not entail any specific number of Points labels. The timing/existence of Points labels is independent of all other tiers. (NOTE: this file has been labelled with "advanced" PoLaR labels, which allow the labeller to add their analysis of the relationship between Points objects and PrStr objects. See the guidelines for details on how to interpret these labels.)

Another important feature of PoLaR labelling is the Ranges tier. The Ranges tier indicates that approximate f0 min and f0 max that correspond to where the speaker is, at this point in the recording, placing their pitch "floor" and "ceiling". In this file, note that there are two Range intervals, corresponding to the analysis that the first prosodic phrase has a narrower pitch range than the latter -- making explicit an analysis that the pitch during "I" is categorically "high" in the same way that the pitch during "reindeer" is (despite the fact that "reindeer" reaches a relatively higher pitch). PoLaR thus allows labellers to capture the dynamic pitch floor and pitch ceiling, over the course of a recording or series of recordings.

One of the intentions of labelling the Ranges tier is to keep track of which parts of the f0 range signal high pitch, and which parts signal low pitch. This aspect of tracking highs and lows is done explicitly through the PoLaR Levels tier. In this tier, the pitch range (as defined by the Range interval) is divided into 5 regions, numbered 1 through 5, with 1 representing the lowest pitch level, and 5 representing the highest. For each label on the Points tier, one is created on the Levels tier, with a value 1-5 — this scaled pitch value is determined by the f0 value at the time of the Points tier label and the range defined by the Range tier interval. This is visualized below:

Further details about these core tiers, their labels, and their relationship to existing research can all be found in the PoLaR Annotation Guidelines.