Linguistic form and meaning in a computational analysis of learner language -- On the integration of morpho-syntactic and semantic analysis

本研究项目探索并开发了一种新的计算架构，用于分析学习者的语言，特别是第二语言学习者所写的德语语言。目的是将形式和意义的分析结合起来，使学习者为了完成给定的语言学习任务而需要表达的意义的信息成为可能，从而为形态句法分析提供信息。模块化体系结构将冲突信息的反向通道与高效的搜索策略相结合，使得通过各级处理获得的信息来影响词法句法分析成为可能。计算语义学传统中溯因推理和演绎推理组件的整合在系统中起着突出的作用。新的综合分析体系的成功与否可以从德语阅读理解练习语料库（CREG）中教师批注的学习者答案来全面评估。这种对学习者语言的形式和意义进行分析的新综合方法在广泛的背景下具有相关性，例如智能语言辅导系统的反馈生成，语言测试中语言能力的自动评估，或学习者语料库的注释，为语言发展的语言学研究提供信息。就更广泛的影响而言，该项目因此解决了如何将自然语言话语的语言或额外语言语境用于分析和解释的一般问题。

This research project explores and develops a new computationalarchitecture for analyzing learner language, specifically the languagewritten by second language learners of German. The goal is tointegrate the analysis of form and meaning in a way that makes itpossible to inform the morpho-syntactic analysis on the basis ofinformation about the meaning that the learner needs to express inorder to complete a given language learning task. The modulararchitecture combines backchannels for conflict information withefficient search strategies to make it possible to influence themorpho-syntactic analysis through information obtained at all levelsof processing. The integration of abductive and deductive reasoningcomponents in the tradition of computational semantics plays aprominent role in the system.The success of the new integrated analysis architecture can thoroughlybe evaluated on the basis of the teacher-annotated learner answersfrom the Corpus of Reading Comprehension Exercises in German (CREG).The new integrated approach to the analysis of the form and meaning oflearner language is of relevance in a wide range of contexts, such asthe generation of feedback for intelligent language tutoring systems,the automatic assessment of language proficiency in language testing,or the annotation of learner corpora informing linguistic researchinto language development. In terms of broader impact, the projectthus addresses the general question how the linguistic orextra-linguistic context of natural language utterances can be usedfor their analysis and interpretation.