Taking at look at letter-to-sound rules (for new words)

[Alexandra S]

Is there are way to take a look at the letter to sound rules used by Festival to find the pronunciation of a word that is outside the lexicon?

[Simon]

Yes – see here. Remember that, for English, they are not rules as such, but a decision tree learned from data (i.e., from the dictionary).

[Alexandra S]

As mentioned in the linked post, reading the decision tree isn’t an obvious task. Are we expected to explain where in the rules the letter-to-sound process failed for a novel word?

If so, is there an easier way to go through the rules (I’m looking at the letter ‘e’ which has an extensive tree) ?

[Simon]

No, you don’t need to explain specific failures by referencing the actual tree. That would not be very insightful anyway: although we often say that decision trees are human readable, the error might occur very deep down the tree and be hard to explain.

Instead, focus on the general properties of the model being used. For example, why might a decision tree make errors at all? How could you do better: a bigger tree, a smaller tree, change from a decision tree to another model, train it on more data, … etc ?

For a word that Festival’s letter-to-sound model got wrong, was it because that word might be particularly hard for some reason? What reason?

[Author]

Posts