A 2nd Longitudinal Corpus for Children’s Writing with Enhanced Output for Specific Spelling Patterns

A 2nd Longitudinal Corpus for Children’s Writing with Enhanced Output for Specific Spelling Patterns Kay Berkling Cooperative State University Karlsruhe, Germany berkling@dhbw-karlsruhe.de Abstract This paper describes the collection of three longitudinal Corpora of German school children’s weekly writing in German, called H2 (H1 is available via LDC and contains some of the same students’ writing 2 years previously), E2 (E1 is not public), and ERK1. The texts were written within the normal classroom setting. Texts of children whose parents signed the permission to donate the texts to science were collected and transcribed. The corpus consists of the elicitation techniques, an overview of the data collected and the transcriptions of the texts both with and without spelling errors, aligned on a word by word basis. In addition, the hand-written texts were scanned in. The corpus is available for research via Linguistic Data Consortium (LDC). When using this Corpus, researchers are strongly encouraged to make additional annotations and improvements and return it to the public domain via LDC, especially since this effort was unfunded. Keywords: Orthography, Acquisition, Corpora, Elementary School, Writing, Digitization 1. Introduction Reading and Spelling are key skills acquired by children during their first four years of school. Unfortunately, it is not clearly understood how and why performance for stan- dardized test populations may degrade or improve. They may be due to teaching style or any other factor. Use- ful studies might look at how writing changes in longi- tudinal studies or as a function of particular training pro- grams, thereby lending insight into quality of school books or teaching philosophies. However, very little of this kind of validation is done on a larger scale or open to compara- tive research with open corpora. Two comparative standardized exams highlight that we have a serious problem that needs addressing without delay. The IQB study looks at reading, mathematics and spelling ability and has been performed in 2015 (9th grade), 2016 (4th grade) and will look at science in 2018. It represents a regional addition to the international PISA test and looks at skills across the different German states. Over the last few years, results for Baden-W¨urttemberg have fallen dramati- cally (Stanat et al., 2016). Another study, VERA in 2017 (Blank and Schult, 2017), clearly demonstrates that student competence in orthogra- phy is much worse than expected. Instead of the predicted 35%, an actual 64% of students place in the lowest 2 out of 5 competency levels in orthography. The skill of reading is less dramatic with around 40% of students expected and actually measured at the lower 2 competency levels. How- ever, both skills show scandalous results when looking at children who are not speaking German in their homes (an estimated 20% of the total number of students): Students tested into the lower 2 levels contain 70% (instead of 34% for German speakers) in reading and 80% (instead of 60% for German speakers) in orthography. Since we started collecting corpora to support research in the area of orthography acquisition, the problem …

Content truncated to 500 words