tom-yx-wang/pychildes-morphosyntax

PyChildes Morphosyntax

Utterance-level morphosyntactic annotations derived from the CHILDES English Eng-NA/Eng-UK corpus via PyChildes.

Credit

When citing the use of TalkBank and CHILDES facilities, please use this reference to the last printed version of the CHILDES manual:

@book{macwhinney2000childes,
    author    = {MacWhinney, Brian},
    title     = {The CHILDES Project: Tools for Analyzing Talk},
    edition   = {3rd},
    year      = {2000},
    publisher = {Lawrence Erlbaum Associates},
    address   = {Mahwah, NJ}
}

This tool is a part of the broader Trabank toolkit that we are actively developing. If you find this tool useful, please give us credit by citing:

@misc{trabank,
    title={Trabank: A Toolkit for Computational Developmental Studies in Language Models},
    author={Ma, Ziqiao and Wu, Shuyu and Wang, Yixuan and Chai, Joyce and Shi, Freda},
    howpublished={https://github.com/Mars-tin/PyChildes},
    year={2025}
}

Schema

Field	Type	Description
id_file	string	Corpus file path (e.g. Eng-UK/Edinburgh/eve1510)
id_utterance	uint64	Utterance index within the file
speaker	string	Speaker code (e.g. PAR0, CHI)
utterance	large_string	Raw utterance text
morph_lemma	list<string>	Lemma for each token
morph_pos	list<dict[string, uint8]>	Part-of-speech tag per token
morph_word_type	list<dict[string, uint8]>	Token surface type: word, compound, or punctuation
morph_metadata	large_list<large_string>	Per-token JSON with CHILDES morphological detail (word/clitic index, features, subpos, compound components)
syntax_arc	list<uint64>	Dependency head index per token (0 = ROOT)
syntax_rel	list<dict[string, uint8]>	Dependency relation label per token (e.g. ROOT, SUBJ, OBJ)

All per-token list fields are aligned: index i across morph_* and syntax_* fields refers to the same token.

Details

Morphological Metadata (morph_metadata)

Each element is a JSON string roughly structured as follows:

{
  "index_word": 1,
  "index_clitic": 1,
  "subpos": "per",
  "components": [
    { "pos": "n", "lemma": "ice" },
    { "pos": "n", "lemma": "berg" }
  ],
  "features": {
    "feature": ["3S"],
    "suffix": [],
    "prefix": [],
    "explanation": []
  }
}

subpos (subsidiary POS tag, like art for articles), components and features are only present if applicable.

POS Tags (morph_pos)

Available POS tags include:

['L2', 'adj', 'adv', 'aux', 'bab', 'chi', 'co', 'comp', 'conj', 'coord', 'cop', 'det', 'dia', 'fam', 'inf', 'meta', 'mod',
 'n', 'neg', 'neo', 'none', 'on', 'part', 'phon', 'post', 'prep', 'pro', 'punct', 'qn', 'sing', 'test', 'uni', 'unk', 'v', 'wplay']

Dependency Relations (syntax_rel)

Available dependency relation labels include:

['APP', 'AUX', 'BEG', 'BEGP', 'CJCT', 'CMOD', 'COBJ', 'COM', 'COMP', 'CONJ', 'COORD', 'CPOBJ', 'CPRED', 'CSUBJ', 'DATE',
 'DET', 'END', 'ENDP', 'ENUM', 'INCROOT', 'INF', 'JCT', 'LINK', 'LP', 'MOD', 'NAME', 'NEG', 'NJCT', 'OBJ', 'OBJ2', 'OM',
 'POBJ', 'POSTMOD', 'PQ', 'PRED', 'PUNCT', 'QUANT', 'ROOT', 'SRL', 'SUBJ', 'TOP', 'XJCT', 'XMOD']

Example Data

{'id_file': 'Eng-UK/Edinburgh/eve1510',
 'id_utterance': 11,
 'speaker': 'PAR0',
 'morph_lemma': ['look', 'it', 'be', 'a', 'bubble', 'iceberg', '.'],
 'morph_metadata': ['{"index_word": 1, "index_clitic": 1, "features": {"feature": [], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 2, "index_clitic": 1, "subpos": "per", "features": {"feature": [], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 2, "index_clitic": 2, "features": {"feature": ["3S"], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 3, "index_clitic": 1, "subpos": "art", "features": {"feature": [], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 4, "index_clitic": 1, "features": {"feature": [], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 5, "index_clitic": 1, "components": [{"pos": "n", "lemma": "ice"}, {"pos": "n", "lemma": "berg"}], "features": {"feature": [], "suffix": [], "prefix": [], "explanation": []}}',
  '{"index_word": 6, "index_clitic": 1}'],
 'morph_pos': ['co', 'pro', 'cop', 'det', 'n', 'n', 'punct'],
 'morph_word_type': ['word',
  'word',
  'word',
  'word',
  'word',
  'compound',
  'punctuation'],
 'syntax_arc': [3, 3, 0, 6, 6, 3, 3],
 'syntax_rel': ['COM', 'SUBJ', 'ROOT', 'DET', 'MOD', 'PRED', 'PUNCT'],
 'utterance': "look it's a bubble iceberg ."}