| Input a noun phrase (NP) and the model will tag its syntactic head as HEAD. | |
| Non-head tokens will be tagged as NA. | |
| The model is intended to tag multiple heads in compound NPs. It is not | |
| intended to tag all heads in complex NPs. | |
| The training data is derived from a dataset in which NPs are identified | |
| using the Berkeley neural parser which does not identify the heads of | |
| syntactic constituents. The training data includes erroneous NPs in | |
| which each token is tagged as ERROR. | |
| Examples you might try include: | |
| The bottle of ginger beer | |
| John, Peter, and Mary, who live in Sri Lanka | |
| The man who won the race |