From CHAT to Coded CoNLL-U: A Reproducible Pipeline for the Syntactic Annotation and Querying of Child Language Data

What it is about

Stein presents an open pipeline that converts CHILDES transcripts in the CHAT format into CSV tables and CoNLL-U. The script childes.py tokenises, optionally tags with TreeTagger, parses with the UDPipe web API, and can apply Grew rewriting rules to fix systematic parser errors. A second script, dql.py, runs Grew graph queries over the parsed CoNLL-U and writes user-defined codings into the metadata and the CSV. The pipeline is shown on the French CHILDES corpus for the production of object clitics. The result is a developmental curve: the share of dative clitics in child speech starts near 9% against about 26% in adult input and rises toward the adult level.

Ideas

• I1. en/src/convert.rs and en/src/tok.rs: a CHAT reader for English CHILDES that keeps the speaker, age and utterance IDs, checked by a round trip (utterance count and token count equal to the source, per corpus).
• I2. A query layer in Rust over the mova dialect: named patterns (for example a do-support or particle-verb pattern) that write codings into MISC, measured by match counts per 1k sentences on EWT test (test only, licence permitting).
• I3. en tagger and parser as the parser of the pipeline: the Grew-style rewrite rules of this paper, re-expressed as en/seeds rules, measured by the change in LAS on CHILDES test after each rule.

For Mova

The paper is a pipeline for corpora, not a model, so the useful parts are the coding layer and the CHAT reader. Mova already has the converters in en/src/convert.rs. The coding layer belongs next to prag and treeagree. Relevant chapters are languages/en/37-register-and-variation and 22-phrases. Measure first: the round-trip counts on one CHILDES corpus before any parser change.

Topics: Universal Dependencies

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.63317/498zo5heasd5

@inproceedings{stein2026chat,
  title = {{From CHAT to Coded CoNLL-U: A Reproducible Pipeline for the Syntactic Annotation and Querying of Child Language Data}},
  author = {Stein, Achim},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/498zo5heasd5},
  url = {https://doi.org/10.63317/498zo5heasd5}
}

References

26 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anne Abeillé and Lionel Clément. 2006. Manuel d’annotation pour les corpus du projet PERCEO. https://repository.ortolang.fr/api/content/perceo-/1/Manuel-annotation.pdf. ↗
2. Anna Cardinaletti, Sara Cerutti, and Francesca Volpato. 2021. On the acquisition of third person dative clitic pronouns in Italian. Lingue e linguaggio, (2):311–341. find
3. Hélène Delage, Stephanie Durrleman, and Ulrich H. Frauenfelder. 2016. Disentangling sources of difficulty associated with the acquisition of accusative clitics in French. Lingua, 180:1–24. ↗
4. Gaël Guibon, Marine Courtin, Kim Gerdes, and Bruno Guillaume. 2020. When Collaborative Treebank Curation Meets Graph Grammars. In LREC 2020 - 12th Language Resources and Evaluation Conference, Marseille, France. find
5. Bruno Guillaume. 2019. Graph matching for corpora exploration. In JLC 2019 - 10èmes Journées Internationales de La Linguistique de Corpus, Grenoble, France. find
6. Brian MacWhinney. 2000. The Childes Project. 1: Transcription Format and Programs. Erlbaum, Mahwah. find
7. Brian MacWhinney. 2019. CHAT Manual. find
8. Mihaela Pirvulescu and Nelleke Strik. 2014. The acquisition of object clitic features in French: A comprehension study. Lingua, 144:58–71. ↗
9. Beth Randall, Anthony Kroch, and Beatrice Santorini. 2004. CorpusSearch 2 users guide. find
10. Kenji Sagae, Eric Davis, Alon Lavie, Brian MacWhinney, and Shuly Wintner. 2007. High-accuracy annotation and parsing of CHILDES transcripts. In Proceedings of the Workshop on Cognitive Aspects of Computational Language Acquisition, pages 25–32, Prague, Czech Republic. Association for Computational Linguistics. ↗
11. Helmut Schmid. 1994. Probabilistic Part-of-Speech Tagging using Decision Trees. Stuttgart. Institut für maschinelle Sprachverarbeitung. find
12. Milan Straka. 2018. UDPipe 2.0 prototype at CoNLL 2018 UD shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 197–207, Brussels, Belgium. Association for Computational Linguistics. 8. find our card
13. Language Resource References Bassano, Dominique. 2005. CHILDES French Pauline Corpus. TalkBank. PID https://doi.org/10.21415/T5VS3W. ↗
14. Champaud, Christian. 2004. CHILDES French Champaud Corpus. TalkBank. PID https://doi.org/10.21415/T5M88F. ↗
15. Guillaume, Bruno. 2025. Grew. Graph Rewriting for NLP. PID https://grew.fr. ↗
16. Jisa, Harriet. 2004. PhonBank French Lyon Corpus. TalkBank. PID https://doi.org/10.21415/T5M02D. ↗
17. Le Normand, Marie-Thérèse. 2014. CHILDES French MTLN Corpus. TalkBank. PID https://doi.org/10.21415/T58S3M. ↗
18. Leveillé, Madeleine. 2004. CHILDES French Leveillé Corpus. TalkBank. PID https://doi.org/10.21415/T5RK5N. ↗
19. Morgenstern, Aliyah and Parisse, Christophe. 2009. PhonBank French Paris Corpus. TalkBank. PID https://doi.org/10.21415/T5PS3B. ↗
20. Palasis, Katerina. 2020. CHILDES French Palasis Corpus. TalkBank. PID https://doi.org/10.21415/T5SW4P. ↗
21. Plunkett, Bernadette. 2004. CHILDES French York Corpus. TalkBank. PID https://doi.org/10.21415/T5NK63. ↗
22. Rasetti, Lucienne. 2004. CHILDES French Geneva Corpus. TalkBank. PID https://doi.org/10.21415/T5F898. Schmid, Helmut. TreeTaggger. PID https://cis.uni-muenchen.de/ schmid/tools/TreeTagger/. ↗
23. Stein, Achim. 2026. French CHILDES Pipeline, version 3.0. GitHub. PID https://github.com/michaniets/french-childes. ↗
24. Straka, Milan. 2023. UDPipe, version 2.1.0. PID https://github.com/ufal/udpipe. ↗
25. Vion, Monique. 2004. CHILDES French Vion/Colas Corpus. TalkBank. PID https://doi.org/10.21415/T5N300. ↗
26. Yamaguchi, Naomi. 2015. PhonBank French Yamaguchi Corpus. TalkBank. PID https://doi.org/10.21415/T5ZP45. ↗

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/bank-2026-stein-chat-coded-conll-u-reproducible/