PuSH - Publikationsserver des Helmholtz Zentrums München

Zugarini, A.* ; Rothenbacher, T.* ; Klede, K.* ; Ernandes, M.* ; Eskofier, B.M. ; Zanca, D.*

Die Rätselrevolution: Automated German Crossword Solving.

In: (9th Italian Conference on Computational Linguistics, CLiC-it 2023, 30 November-2 December 2023, Venice, Italy). Aachen: RWTH, 2023. (CEUR Workshop Proc. ; 3596)
Verlagsversion
Open Access Gold (Paid Option)
Crossword puzzles are popular word games played in various languages around the world, with diverse styles across different countries. For this reason, automated crossword solvers designed for a language, may not work well on others. In this paper, we extend Webcrow, an automatic crossword solver, to German, making it the first program for crossword solving in the German language. To address the lack of large clue-answer crossword pairs data, Webcrow combines multiple modules, known as experts, which retrieve potential answers from various resources, including the web, knowledge graphs, and linguistic rules. The system is evaluated on a collection of crosswords from variegate sources, where it is able to solve perfectly 67% of them. Additional analysis reveals that while our solver achieved commendable results, puzzles with poorly constrained schemas and original clues still presented significant hurdles. These findings shed light on the complexity of the crossword-solving problem and emphasize the need for future research to address and overcome these particular challenges effectively.
Weitere Metriken?
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Konferenzbeitrag
Korrespondenzautor
Schlagwörter Automated Crosswords Solving ; German Crosswords ; Question Answering
ISSN (print) / ISBN 1613-0073
Konferenztitel 9th Italian Conference on Computational Linguistics, CLiC-it 2023
Konferzenzdatum 30 November-2 December 2023
Konferenzort Venice, Italy
Quellenangaben Band: 3596 Heft: , Seiten: , Artikelnummer: , Supplement: ,
Verlag RWTH
Verlagsort Aachen
Nichtpatentliteratur Publikationen
Institut(e) Institute of AI for Health (AIH)