Automatic Text Simplification for German

Abstract

The article at hand aggregates the work of our group in automatic processing of simplified German. We present four parallel (standard/simplified German) corpora compiled and curated by our group. We report on the creation of a gold standard of sentence alignments from the four sources for evaluating automatic alignment methods on this gold standard. We show that one of the alignment methods performs best on the majority of the data sources. We used two of our corpora as a basis for the first sentence-based neural machine translation (NMT) approach toward automatic simplification of German. In follow-up work, we extended our model to render it capable of explicitly operating on multiple levels of simplified German. We show that using source-side language level labels improves performance with regard to two evaluation metrics commonly applied to measuring the quality of automatic text simplification.

LAIR - Language As Intermediate Representation

“…The LangugeTool 18 , for instance, uses a rule-based approach to check texts for violations of LS rules. For automatic text simplification in German, Suter et al (2016) present a rule-based approach, and Ebling et al (2022) introduce a sentence-based neural machine translation approach. SummAI 19 uses artificial intelligence to assist authors in translating existing texts into LS and creating new LS information.…”

Cognitive Aspects of Compound Translation: Insights into the Relation between Implicitation and Semantic Complexity from a Translation Process Perspective

“…To our knowledge, Säuberli et al (2020) were the first to adapt neural models to the features of German Easy Language. Their models were able to implement some specificities of Easy Language, such as choosing basic words or shortening sen-tences.…”

“…As their analysis also revealed that in most cases, the sentences were not significantly easier than the original sentences, they conclude that a larger parallel corpus is needed to successfully train an automatic text simplification system for German Easy Language. Spring et al (2021) expanded the corpus used by Säuberli et al (2020) and developed a sentencebased machine translation approach to automatically simplify standard German into different simplification levels of the Common European Framework of References for Languages (CEFR). To tackle the above-mentioned alignment problems, they used the Sentence Alignment Tools Evaluation Framework (SATEF), which allows for n:m alignments, meaning that one alignment segment can consist of a varying number of sentences in the source and target text.…”

“…However, both Säuberli et al (2020) and Spring et al (2021) showed that existing models tend to copy the source segments. The latter were able to reduce the copying behavior of the text simplification models by applying different pretraining and fine-tuning strategies and by adding copy labels.…”

Language Representation Models: An Overview

“…(Rios et al, 2021) investigated the usage of an adapted mBART version with Longformer attention (Beltagy et al, 2020) on Swiss newspaper articles. These results have been further improved with a sentence-based approach (Ebling et al, 2022). Most recently, the first detailed surveys about German text simplification have been released (Anschütz et al, 2023;Stodden et al, 2023;Schomacker et al, 2023).…”