initial release

README.md

@@ -0,0 +1,75 @@
+---
+language:
+- "ja"
+tags:
+- "japanese"
+- "pos"
+- "dependency-parsing"
+datasets:
+- "universal_dependencies"
+license: "apache-2.0"
+pipeline_tag: "token-classification"
+widget:
+- text: "全学年にわたって小学校の国語の教科書に挿し絵が用いられている"
+---
+
+# deberta-v3-base-japanese-ud-goeswith
+
+## Model Description
+
+This is a DeBERTa(V3) model pretrained on LLM-jp corpus v1.0 for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [deberta-v3-base-japanese](https://huggingface.co/ku-nlp/deberta-v3-base-japanese) and [UD_Japanese-GSDLUW](https://github.com/UniversalDependencies/UD_Japanese-GSDLUW).
+
+## How to Use
+
+```py
+class UDgoeswith(object):
+  def __init__(self,bert):
+    from transformers import AutoTokenizer,AutoModelForTokenClassification
+    self.tokenizer=AutoTokenizer.from_pretrained(bert)
+    self.model=AutoModelForTokenClassification.from_pretrained(bert)
+  def __call__(self,text):
+    import numpy,torch,ufal.chu_liu_edmonds
+    w=self.tokenizer(text,return_offsets_mapping=True)
+    v=w["input_ids"]
+    x=[v[0:i]+[self.tokenizer.mask_token_id]+v[i+1:]+[j] for i,j in enumerate(v[1:-1],1)]
+    with torch.no_grad():
+      e=self.model(input_ids=torch.tensor(x)).logits.numpy()[:,1:-2,:]
+    r=[1 if i==0 else -1 if j.endswith("|root") else 0 for i,j in sorted(self.model.config.id2label.items())]
+    e+=numpy.where(numpy.add.outer(numpy.identity(e.shape[0]),r)==0,0,numpy.nan)
+    g=self.model.config.label2id["X|_|goeswith"]
+    r=numpy.tri(e.shape[0])
+    for i in range(e.shape[0]):
+      for j in range(i+2,e.shape[1]):
+        r[i,j]=r[i,j-1] if numpy.nanargmax(e[i,j-1])==g else 1
+    e[:,:,g]+=numpy.where(r==0,0,numpy.nan)
+    m=numpy.full((e.shape[0]+1,e.shape[1]+1),numpy.nan)
+    m[1:,1:]=numpy.nanmax(e,axis=2).transpose()
+    p=numpy.zeros(m.shape)
+    p[1:,1:]=numpy.nanargmax(e,axis=2).transpose()
+    for i in range(1,m.shape[0]):
+      m[i,0],m[i,i],p[i,0]=m[i,i],numpy.nan,p[i,i]
+    h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    if [0 for i in h if i==0]!=[0]:
+      m[:,0]+=numpy.where(m[:,0]==numpy.nanmax(m[[i for i,j in enumerate(h) if j==0],0]),0,numpy.nan)
+      m[[i for i,j in enumerate(h) if j==0]]+=[0 if i==0 or j==0 else numpy.nan for i,j in enumerate(h)]
+      h=ufal.chu_liu_edmonds.chu_liu_edmonds(m)[0]
+    u="# text = "+text+"\n"
+    v=[(s,e) for s,e in w["offset_mapping"] if s<e]
+    for i,(s,e) in enumerate(v,1):
+      q=self.model.config.id2label[p[i,h[i]]].split("|")
+      u+="\t".join([str(i),text[s:e],"_",q[0],"_","|".join(q[1:-1]),str(h[i]),q[-1],"_","_" if i<len(v) and e<v[i][0] else "SpaceAfter=No"])+"\n"
+    return u+"\n"
+
+nlp=UDgoeswith("KoichiYasuoka/deberta-v3-base-japanese-ud-goeswith")
+print(nlp("全学年にわたって小学校の国語の教科書に挿し絵が用いられている"))
+```
+
+with [ufal.chu-liu-edmonds](https://pypi.org/project/ufal.chu-liu-edmonds/).
+Or without ufal.chu-liu-edmonds:
+
+```
+from transformers import pipeline
+nlp=pipeline("universal-dependencies","KoichiYasuoka/deberta-v3-base-japanese-ud-goeswith",trust_remote_code=True,aggregation_strategy="simple")
+print(nlp("全学年にわたって小学校の国語の教科書に挿し絵が用いられている"))
+```
+

added_tokens.json

@@ -0,0 +1,7 @@
+{
+  "[CLS]": 96871,
+  "[MASK]": 96867,
+  "[PAD]": 96869,
+  "[SEP]": 96868,
+  "[UNK]": 96870
+}

config.json

@@ -0,0 +1,228 @@
+{
+  "architectures": [
+    "DebertaV2ForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "custom_pipelines": {
+    "universal-dependencies": {
+      "impl": "ud.UniversalDependenciesPipeline"
+    }
+  },
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "-|_|dep",
+    "1": "ADJ|_|acl",
+    "2": "ADJ|_|advcl",
+    "3": "ADJ|_|amod",
+    "4": "ADJ|_|ccomp",
+    "5": "ADJ|_|csubj",
+    "6": "ADJ|_|csubj:outer",
+    "7": "ADJ|_|dep",
+    "8": "ADJ|_|nmod",
+    "9": "ADJ|_|nsubj",
+    "10": "ADJ|_|obj",
+    "11": "ADJ|_|obl",
+    "12": "ADJ|_|root",
+    "13": "ADP|_|case",
+    "14": "ADP|_|fixed",
+    "15": "ADV|_|advcl",
+    "16": "ADV|_|advmod",
+    "17": "ADV|_|dep",
+    "18": "ADV|_|obj",
+    "19": "ADV|_|root",
+    "20": "AUX|Polarity=Neg|aux",
+    "21": "AUX|Polarity=Neg|fixed",
+    "22": "AUX|_|aux",
+    "23": "AUX|_|cop",
+    "24": "AUX|_|fixed",
+    "25": "AUX|_|root",
+    "26": "CCONJ|_|cc",
+    "27": "DET|_|det",
+    "28": "INTJ|_|discourse",
+    "29": "INTJ|_|root",
+    "30": "NOUN|Polarity=Neg|obl",
+    "31": "NOUN|Polarity=Neg|root",
+    "32": "NOUN|_|acl",
+    "33": "NOUN|_|advcl",
+    "34": "NOUN|_|ccomp",
+    "35": "NOUN|_|compound",
+    "36": "NOUN|_|csubj",
+    "37": "NOUN|_|csubj:outer",
+    "38": "NOUN|_|nmod",
+    "39": "NOUN|_|nsubj",
+    "40": "NOUN|_|nsubj:outer",
+    "41": "NOUN|_|obj",
+    "42": "NOUN|_|obl",
+    "43": "NOUN|_|root",
+    "44": "NUM|_|advcl",
+    "45": "NUM|_|compound",
+    "46": "NUM|_|nmod",
+    "47": "NUM|_|nsubj",
+    "48": "NUM|_|nsubj:outer",
+    "49": "NUM|_|nummod",
+    "50": "NUM|_|obj",
+    "51": "NUM|_|obl",
+    "52": "NUM|_|root",
+    "53": "PART|_|mark",
+    "54": "PRON|_|acl",
+    "55": "PRON|_|advcl",
+    "56": "PRON|_|nmod",
+    "57": "PRON|_|nsubj",
+    "58": "PRON|_|nsubj:outer",
+    "59": "PRON|_|obj",
+    "60": "PRON|_|obl",
+    "61": "PRON|_|root",
+    "62": "PROPN|_|acl",
+    "63": "PROPN|_|advcl",
+    "64": "PROPN|_|compound",
+    "65": "PROPN|_|nmod",
+    "66": "PROPN|_|nsubj",
+    "67": "PROPN|_|nsubj:outer",
+    "68": "PROPN|_|obj",
+    "69": "PROPN|_|obl",
+    "70": "PROPN|_|root",
+    "71": "PUNCT|_|punct",
+    "72": "SCONJ|_|dep",
+    "73": "SCONJ|_|fixed",
+    "74": "SCONJ|_|mark",
+    "75": "SYM|_|compound",
+    "76": "SYM|_|dep",
+    "77": "SYM|_|nmod",
+    "78": "SYM|_|obl",
+    "79": "VERB|_|acl",
+    "80": "VERB|_|advcl",
+    "81": "VERB|_|ccomp",
+    "82": "VERB|_|compound",
+    "83": "VERB|_|csubj",
+    "84": "VERB|_|csubj:outer",
+    "85": "VERB|_|nmod",
+    "86": "VERB|_|obj",
+    "87": "VERB|_|obl",
+    "88": "VERB|_|root",
+    "89": "X|_|dep",
+    "90": "X|_|goeswith",
+    "91": "X|_|nmod"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "-|_|dep": 0,
+    "ADJ|_|acl": 1,
+    "ADJ|_|advcl": 2,
+    "ADJ|_|amod": 3,
+    "ADJ|_|ccomp": 4,
+    "ADJ|_|csubj": 5,
+    "ADJ|_|csubj:outer": 6,
+    "ADJ|_|dep": 7,
+    "ADJ|_|nmod": 8,
+    "ADJ|_|nsubj": 9,
+    "ADJ|_|obj": 10,
+    "ADJ|_|obl": 11,
+    "ADJ|_|root": 12,
+    "ADP|_|case": 13,
+    "ADP|_|fixed": 14,
+    "ADV|_|advcl": 15,
+    "ADV|_|advmod": 16,
+    "ADV|_|dep": 17,
+    "ADV|_|obj": 18,
+    "ADV|_|root": 19,
+    "AUX|Polarity=Neg|aux": 20,
+    "AUX|Polarity=Neg|fixed": 21,
+    "AUX|_|aux": 22,
+    "AUX|_|cop": 23,
+    "AUX|_|fixed": 24,
+    "AUX|_|root": 25,
+    "CCONJ|_|cc": 26,
+    "DET|_|det": 27,
+    "INTJ|_|discourse": 28,
+    "INTJ|_|root": 29,
+    "NOUN|Polarity=Neg|obl": 30,
+    "NOUN|Polarity=Neg|root": 31,
+    "NOUN|_|acl": 32,
+    "NOUN|_|advcl": 33,
+    "NOUN|_|ccomp": 34,
+    "NOUN|_|compound": 35,
+    "NOUN|_|csubj": 36,
+    "NOUN|_|csubj:outer": 37,
+    "NOUN|_|nmod": 38,
+    "NOUN|_|nsubj": 39,
+    "NOUN|_|nsubj:outer": 40,
+    "NOUN|_|obj": 41,
+    "NOUN|_|obl": 42,
+    "NOUN|_|root": 43,
+    "NUM|_|advcl": 44,
+    "NUM|_|compound": 45,
+    "NUM|_|nmod": 46,
+    "NUM|_|nsubj": 47,
+    "NUM|_|nsubj:outer": 48,
+    "NUM|_|nummod": 49,
+    "NUM|_|obj": 50,
+    "NUM|_|obl": 51,
+    "NUM|_|root": 52,
+    "PART|_|mark": 53,
+    "PRON|_|acl": 54,
+    "PRON|_|advcl": 55,
+    "PRON|_|nmod": 56,
+    "PRON|_|nsubj": 57,
+    "PRON|_|nsubj:outer": 58,
+    "PRON|_|obj": 59,
+    "PRON|_|obl": 60,
+    "PRON|_|root": 61,
+    "PROPN|_|acl": 62,
+    "PROPN|_|advcl": 63,
+    "PROPN|_|compound": 64,
+    "PROPN|_|nmod": 65,
+    "PROPN|_|nsubj": 66,
+    "PROPN|_|nsubj:outer": 67,
+    "PROPN|_|obj": 68,
+    "PROPN|_|obl": 69,
+    "PROPN|_|root": 70,
+    "PUNCT|_|punct": 71,
+    "SCONJ|_|dep": 72,
+    "SCONJ|_|fixed": 73,
+    "SCONJ|_|mark": 74,
+    "SYM|_|compound": 75,
+    "SYM|_|dep": 76,
+    "SYM|_|nmod": 77,
+    "SYM|_|obl": 78,
+    "VERB|_|acl": 79,
+    "VERB|_|advcl": 80,
+    "VERB|_|ccomp": 81,
+    "VERB|_|compound": 82,
+    "VERB|_|csubj": 83,
+    "VERB|_|csubj:outer": 84,
+    "VERB|_|nmod": 85,
+    "VERB|_|obj": 86,
+    "VERB|_|obl": 87,
+    "VERB|_|root": 88,
+    "X|_|dep": 89,
+    "X|_|goeswith": 90,
+    "X|_|nmod": 91
+  },
+  "layer_norm_eps": 1e-07,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 768,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "tokenizer_class": "DebertaV2Tokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.40.1",
+  "type_vocab_size": 0,
+  "vocab_size": 96900
+}

maker.py

@@ -0,0 +1,54 @@
+#! /usr/bin/python3
+src="ku-nlp/deberta-v3-base-japanese"
+tgt="KoichiYasuoka/deberta-v3-base-japanese-ud-goeswith"
+url="https://github.com/UniversalDependencies/UD_Japanese-GSDLUW"
+import os
+d=os.path.basename(url)
+os.system("test -d "+d+" || git clone --depth=1 "+url)
+os.system("for F in train dev test ; do cp "+d+"/*-$F.conllu $F.conllu ; done")
+class UDgoeswithDataset(object):
+  def __init__(self,conllu,tokenizer):
+    self.ids,self.tags,label=[],[],set()
+    with open(conllu,"r",encoding="utf-8") as r:
+      cls,sep,msk=tokenizer.cls_token_id,tokenizer.sep_token_id,tokenizer.mask_token_id
+      dep,c="-|_|dep",[]
+      for s in r:
+        t=s.split("\t")
+        if len(t)==10 and t[0].isdecimal():
+          c.append(t)
+        elif c!=[] and s.strip()=="":
+          v=tokenizer([t[1] for t in c],add_special_tokens=False)["input_ids"]
+          for i in range(len(v)-1,-1,-1):
+            for j in range(1,len(v[i])):
+              c.insert(i+1,[c[i][0],"_","_","X","_","_",c[i][0],"goeswith","_","_"])
+          y=["0"]+[t[0] for t in c]
+          h=[i if t[6]=="0" else y.index(t[6]) for i,t in enumerate(c,1)]
+          p,v=[t[3]+"|"+t[5]+"|"+t[7] for t in c],sum(v,[])
+          self.ids.append([cls]+v+[sep])
+          self.tags.append([dep]+p+[dep])
+          label=set(sum([self.tags[-1],list(label)],[]))
+          for i,k in enumerate(v):
+            self.ids.append([cls]+v[0:i]+[msk]+v[i+1:]+[sep,k])
+            self.tags.append([dep]+[t if h[j]==i+1 else dep for j,t in enumerate(p)]+[dep,dep])
+          c=[]
+    self.label2id={l:i for i,l in enumerate(sorted(label))}
+  def __call__(*args):
+    label=set(sum([list(t.label2id) for t in args],[]))
+    lid={l:i for i,l in enumerate(sorted(label))}
+    for t in args:
+      t.label2id=lid
+    return lid
+  __len__=lambda self:len(self.ids)
+  __getitem__=lambda self,i:{"input_ids":self.ids[i],"labels":[self.label2id[t] for t in self.tags[i]]}
+from transformers import AutoTokenizer,AutoConfig,AutoModelForTokenClassification,DataCollatorForTokenClassification,TrainingArguments,Trainer
+tkz=AutoTokenizer.from_pretrained(src,model_max_length=512)
+trainDS=UDgoeswithDataset("train.conllu",tkz)
+devDS=UDgoeswithDataset("dev.conllu",tkz)
+testDS=UDgoeswithDataset("test.conllu",tkz)
+lid=trainDS(devDS,testDS)
+cfg=AutoConfig.from_pretrained(src,num_labels=len(lid),label2id=lid,id2label={i:l for l,i in lid.items()})
+arg=TrainingArguments(num_train_epochs=3,per_device_train_batch_size=24,output_dir="/tmp",overwrite_output_dir=True,save_total_limit=2,learning_rate=5e-05,warmup_ratio=0.1,save_safetensors=False)
+trn=Trainer(args=arg,data_collator=DataCollatorForTokenClassification(tkz),model=AutoModelForTokenClassification.from_pretrained(src,config=cfg),train_dataset=trainDS,eval_dataset=devDS)
+trn.train()
+trn.save_model(tgt)
+tkz.save_pretrained(tgt)

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c4fb04d6c1af75d1c7f38389a0b972cb35256c9ed553936475635a93d4f7ad5
+size 639830001

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "[PAD]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fefde905766244f5e613a490d6e35236043d6483c4aae0eaac4b4a8fc365a88
+size 1658609

tokenizer_config.json

@@ -0,0 +1,58 @@
+{
+  "added_tokens_decoder": {
+    "96867": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "96868": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "96869": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "96870": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "96871": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "[CLS]",
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "eos_token": "[SEP]",
+  "keep_accents": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "sp_model_kwargs": {},
+  "split_by_punct": false,
+  "tokenizer_class": "DebertaV2Tokenizer",
+  "unk_token": "[UNK]"
+}

ud.py

@@ -0,0 +1,62 @@
+from transformers import TokenClassificationPipeline
+
+class UniversalDependenciesPipeline(TokenClassificationPipeline):
+  def _forward(self,model_inputs):
+    import torch
+    v=model_inputs["input_ids"][0].tolist()
+    with torch.no_grad():
+      e=self.model(input_ids=torch.tensor([v[0:i]+[self.tokenizer.mask_token_id]+v[i+1:]+[j] for i,j in enumerate(v[1:-1],1)],device=self.device))
+    return {"logits":e.logits[:,1:-2,:],**model_inputs}
+  def postprocess(self,model_outputs,**kwargs):
+    import numpy
+    if "logits" not in model_outputs:
+      return "".join(self.postprocess(x,**kwargs) for x in model_outputs)
+    e=model_outputs["logits"].numpy()
+    r=[1 if i==0 else -1 if j.endswith("|root") else 0 for i,j in sorted(self.model.config.id2label.items())]
+    e+=numpy.where(numpy.add.outer(numpy.identity(e.shape[0]),r)==0,0,numpy.nan)
+    g=self.model.config.label2id["X|_|goeswith"]
+    r=numpy.tri(e.shape[0])
+    for i in range(e.shape[0]):
+      for j in range(i+2,e.shape[1]):
+        r[i,j]=r[i,j-1] if numpy.nanargmax(e[i,j-1])==g else 1
+    e[:,:,g]+=numpy.where(r==0,0,numpy.nan)
+    m,p=numpy.nanmax(e,axis=2),numpy.nanargmax(e,axis=2)
+    h=self.chu_liu_edmonds(m)
+    z=[i for i,j in enumerate(h) if i==j]
+    if len(z)>1:
+      k,h=z[numpy.nanargmax(m[z,z])],numpy.nanmin(m)-numpy.nanmax(m)
+      m[:,z]+=[[0 if j in z and (i!=j or i==k) else h for i in z] for j in range(m.shape[0])]
+      h=self.chu_liu_edmonds(m)
+    v=[(s,e) for s,e in model_outputs["offset_mapping"][0].tolist() if s<e]
+    q=[self.model.config.id2label[p[j,i]].split("|") for i,j in enumerate(h)]
+    if "aggregation_strategy" in kwargs and kwargs["aggregation_strategy"]!="none":
+      for i,j in reversed(list(enumerate(q[1:],1))):
+        if j[-1]=="goeswith" and set([t[-1] for t in q[h[i]+1:i+1]])=={"goeswith"}:
+          h=[b if i>b else b-1 for a,b in enumerate(h) if i!=a]
+          v[i-1]=(v[i-1][0],v.pop(i)[1])
+          q.pop(i)
+    t=model_outputs["sentence"].replace("\n"," ")
+    u="# text = "+t+"\n"
+    for i,(s,e) in enumerate(v):
+      u+="\t".join([str(i+1),t[s:e],"_",q[i][0],"_","|".join(q[i][1:-1]),str(0 if h[i]==i else h[i]+1),q[i][-1],"_","_" if i+1<len(v) and e<v[i+1][0] else "SpaceAfter=No"])+"\n"
+    return u+"\n"
+  def chu_liu_edmonds(self,matrix):
+    import numpy
+    h=numpy.nanargmax(matrix,axis=0)
+    x=[-1 if i==j else j for i,j in enumerate(h)]
+    for b in [lambda x,i,j:-1 if i not in x else x[i],lambda x,i,j:-1 if j<0 else x[j]]:
+      y=[]
+      while x!=y:
+        y=list(x)
+        for i,j in enumerate(x):
+          x[i]=b(x,i,j)
+      if max(x)<0:
+        return h
+    y,x=[i for i,j in enumerate(x) if j==max(x)],[i for i,j in enumerate(x) if j<max(x)]
+    z=matrix-numpy.nanmax(matrix,axis=0)
+    m=numpy.block([[z[x,:][:,x],numpy.nanmax(z[x,:][:,y],axis=1).reshape(len(x),1)],[numpy.nanmax(z[y,:][:,x],axis=0),numpy.nanmax(z[y,y])]])
+    k=[j if i==len(x) else x[j] if j<len(x) else y[numpy.nanargmax(z[y,x[i]])] for i,j in enumerate(self.chu_liu_edmonds(m))]
+    h=[j if i in y else k[x.index(i)] for i,j in enumerate(h)]
+    i=y[numpy.nanargmax(z[x[k[-1]],y] if k[-1]<len(x) else z[y,y])]
+    h[i]=x[k[-1]] if k[-1]<len(x) else i
+    return h