hankcs/HanLP · critical · FileNotFoundError

Chinese Treebank 9.0 is a copyright dataset owned by LDC whi

Error message

Chinese Treebank 9.0 is a copyright dataset owned by LDC which we cannot re-distribute. Please apply for a licence from LDC (https://catalog.ldc.upenn.edu/LDC2016T13) then download it to {path_from_url(_CTB9_HOME)}

What it means

At import time, hanlp.datasets.parsing.ctb9 tries to fetch the CTB9 dataset home and, on HTTPError (the data is not hosted), raises FileNotFoundError explaining that CTB 9.0 is LDC-licensed and cannot be redistributed. The message tells you to obtain an LDC licence and manually place the corpus at the shown local path.

Source

Thrown at hanlp/datasets/parsing/ctb9.py:49

CTB9_BRACKET_LINE_NOEC_TRAIN = _CTB9_HOME + 'tasks/par/train.noempty.txt'
'''Training set for ctb9 constituency parsing without empty categories.'''
CTB9_BRACKET_LINE_NOEC_DEV = _CTB9_HOME + 'tasks/par/dev.noempty.txt'
'''Dev set for ctb9 constituency parsing without empty categories.'''
CTB9_BRACKET_LINE_NOEC_TEST = _CTB9_HOME + 'tasks/par/test.noempty.txt'
'''Test set for ctb9 constituency parsing without empty categories.'''

CTB9_SD330_TRAIN = _CTB9_HOME + 'tasks/dep/train.conllx'
'''Training set for ctb9 in Stanford Dependencies 3.3.0 standard.'''
CTB9_SD330_DEV = _CTB9_HOME + 'tasks/dep/dev.conllx'
'''Dev set for ctb9 in Stanford Dependencies 3.3.0 standard.'''
CTB9_SD330_TEST = _CTB9_HOME + 'tasks/dep/test.conllx'
'''Test set for ctb9 in Stanford Dependencies 3.3.0 standard.'''

try:
    get_resource(_CTB9_HOME)
except HTTPError:
    raise FileNotFoundError(
        'Chinese Treebank 9.0 is a copyright dataset owned by LDC which we cannot re-distribute. '
        f'Please apply for a licence from LDC (https://catalog.ldc.upenn.edu/LDC2016T13) '
        f'then download it to {path_from_url(_CTB9_HOME)}'
    )

make_ctb(_CTB9_HOME)

View on GitHub (pinned to ddb1299bdd)

Solutions

1. Apply for an LDC licence (LDC2016T13) and download CTB9
2. Copy/extract the corpus to the path shown in the message (path_from_url(_CTB9_HOME) — typically your HANLP data dir) and re-run
3. Alternatively use an open Chinese treebank/dataset supported by HanLP if licensing CTB is not an option

Example fix

# shell: after obtaining CTB9 from LDC
mkdir -p ~/hanlp/ctb9
cp -r /path/to/ctb9/* ~/hanlp/ctb9/
# then re-run your script

from hanlp.utils.io_util import path_from_url
from os.path import exists
ctb_home = path_from_url('http://nlp.ee.polyu.edu.hk/releases/ctb9/')  # or reuse _CTB9_HOME
if not exists(ctb_home):
    raise SystemExit(f'CTB9 corpus missing at {ctb_home}. Obtain LDC2016T13 and place it there.')

try:
    import hanlp.datasets.parsing.ctb9  # noqa
except FileNotFoundError as e:
    print(e)  # show licence/placement instructions
    sys.exit(1)

When it happens

Trigger: import hanlp.datasets.parsing.ctb9 (or a pipeline that pulls it in) without the CTB9 corpus already at the expected local path; the download attempt gets an HTTP error instead of data.

Common situations: Fresh machines/environments; assuming HanLP auto-downloads all datasets like it does for open ones; new users unaware CTB is paid LDC data.

Related errors

• Got average f{average}, expected one of None, 'token', or 'b
• alpha must be float, list of float, or torch.FloatTensor, {}
• Only supports floating point dtypes.
• Does not support dtype " + str(dtype)
• activation must be callable: type={}

AI-assisted analysis of hankcs/HanLP@ddb1299bdd (2026-08-27). Data as JSON: /api/errors/c23b9799b47383f5. Report an issue: GitHub.