Keep the deep_morph structure intact when resolving MWU's

[kosloot]

Keep the deep_morph structure intact when resolving MWU's

[kosloot]

For now, this seems a non-problem to me.
Does a MWU even have a feasible deep morph structure?
What the current implementation delivers seems good enough. eg:

aan de vooravond van Kerstmis
gives the JSON:

[
  {
    "chunking": {
      "confidence": 1.0,
      "tag": "B-PP_I-PP_I-PP_I-PP"
    },
    "index": 1,
    "lemma": "aan_de_vooravond_van",
    "morph": "[aan]preposition_[de]article_[[voor]adverb [avond]noun/singular]noun_[van]preposition/singular",
    "parse": {
      "parse_index": 0,
      "parse_role": "ROOT"
    },
    "pos": {
      "confidence": 0.995519538829995,
      "tag": "VZ(init)_LID(bep,stan,rest)_N(soort,ev,basis,zijd,stan)_VZ(init)"
    },
    "ucto": {
      "new_paragraph": true,
      "token": "WORD"
    },
    "word": "aan_de_vooravond_van"
  },
  {
    "chunking": {
      "confidence": 1.0,
      "tag": "B-NP"
    },
    "index": 2,
    "lemma": "Kerstmis",
    "morph": "[Kerstmis]",
    "ner": {
      "confidence": 0.6666666666666666,
      "tag": "B-misc"
    },
    "parse": {
      "parse_index": 1,
      "parse_role": "obj1"
    },
    "pos": {
      "confidence": 1.0,
      "tag": "SPEC(deeleigen)"
    },
    "ucto": {
      "token": "WORD"
    },
    "word": "Kerstmis"
  }
]

Nothing wrong with that...