We are happy to announce a talk by Ruby Sleeman(Institute for Mediterranean Studies (IMS-FORTH) Rethymno) and Nicolas Lamoure (University of Frankfurt) in the Historical Linguistics Colloquium. The talk will take place online and in English. Date: July 16th, 2026 Time: 2pm - 4pm c.t. Place: https://uni-frankfurt.zoom-x.de/j/68018980251?pwd=VpaWfCKq2CqU94b0KSzJEfefblCtfs.1 Title: Facilitating efficient manual curation of automatically extracted webcorpus data: A case study of microvariation in German "Kein ein" and Dutch "Geenéén" Abstract: This talk presents a reproducible workflow for extracting, filtering, and curating rare syntactic constructions from large web corpora, using German kein ein and Dutch geen één as a case study. Example: 1a. %Er hat keinen einen Fehler gemacht. 1b. Hijheeft geen één fout gemaakt. he has no one mistake made 'He hasn't made a single mistake.' The construction combines a negative determiner with an element corresponding to 'one', with the meaning 'not a single N', and shows interesting micro- and mesovariation: while Dutch appears to allow the construction more generally (Broekhuis & Den Dikken 2015), German displays variation across speakers and contexts. In order to further investigate such constructions which are accepted to varying degreesby different speakers, we present a workflow for extracting and curating corpus data. As simple string searches produce many false positives, while fully automatic syntactic filtering remains too noisy, the workflow combines computational extraction with manual linguistic evaluation. Candidate examples are retrieved from the German and Dutch TenTen corpora (Jakubí?ek et al. 2013) via the Sketch Engine API, parsed with Stanza (Qi et al. 2020), and filtered for dependency configurations in which negative and indefinite determiners attach to the same nominal head. The resulting data are exported in a structured, spreadsheet-readable format and further evaluated in a dedicated web-based curation tool. The tool supports collaborative annotation through individual user accounts, separately stored judgements, resumable sessions, and exportable results. The talk focuses on the practical advantages and limitations of this human-in-the-loop approach, including parser errors, false positives, corpuscoverage, and the move from corpus evidence to later acceptability judgements. Broekhuis, Hans & Dikken, Marcel den. (2015). 5.1.5.3. The syntactic distribution of (noun phrases containing) geen. Taalportaal. Retrieved from _https://taalportaal.org/taalportaal/topic/pid/topic-14286558606894032_ <https://taalportaal.org/taalportaal/topic/pid/topic-14286558606894032>. (accessed 15 June 2026). Jakubí?ek, M., Kilgarriff, A., Ková?, V., Rychlý, P., & Suchomel, V. (2013). The TenTen corpus family. In 7th International Corpus Linguistics Conference CL (pp. 125-127). Qi, Peng, Yuhao Zhang, Yuhui Zhang, Jason Boltonand Christopher D. Manning. (2020). Stanza: A Python Natural Language Processing Toolkit for Many Human Languages. In Association for Computational Linguistics (ACL) System Demonstrations.