nltk does not find the packaged perceptron tagger

I needed to add the following (according to error messages emitted during loading) to get automatic local pronunciation generation working:

```
import nltk
nltk.download('averaged_perceptron_tagger_eng')
```