Revise document

Author: hankcs

hanlp/pretrained/mtl.py

@@ -20,7 +20,7 @@
 "ERNIE (:cite:`xiao-etal-2021-ernie`) base version of joint tok, pos, ner, srl, dep, sdp and con model trained on close-source Chinese corpus."
 KYOTO_EVAHAN_TOK_LEM_POS_UDEP_LZH = HANLP_URL + 'mtl/kyoto_evahan_tok_lem_pos_udep_bert-ancient-chinese_lr_1_aug_dict_20250112_154422.zip'
 '''
-Ancient Chinese tokenizer with bert-ancient-chinese (:cite:`wang2022uncertainty`) encoder trained on Classical Chinese 
+Ancient Chinese tokenizer, lemmatizer, part-of-speech tagger, dependency parser with bert-ancient-chinese (:cite:`wang2022uncertainty`) encoder trained on Classical Chinese 
 Universal Dependencies Treebank (:cite:`yasuoka2019universal`) and EvaHan corpus (:cite:`li-etal-2022-first`).
 Performance: ``{tok/fine P: 98.91% R: 99.11% F1: 99.01%}{tok/coarse P: 94.71% R: 92.51% F1: 93.60%}{lem Accuracy:98.86%}{pos/upos Accuracy:94.91%}{pos/xpos Accuracy:93.79%}{pos/pku Accuracy:91.91%}{dep UAS: 88.70% LAS: 83.89%}``
 '''