Generating High Quality Proposition Banks for Multilingual Semantic Role Labeling

Abstract

Semantic role labeling (SRL) is crucial to natural language understanding as it identifies the predicate-argument structure in text with semantic labels. Unfortunately, resources required to construct SRL models are expensive to obtain and simply do not exist for most languages. In this paper, we present a two-stage method to enable the construction of SRL models for resourcepoor languages by exploiting monolingual SRL and multilingual parallel data. Experimental results show that our method outperforms existing methods. We use our method to generate Proposition Banks with high to reasonable quality for 7 languages in three language families and release these resources to the research community.

Multi-source annotation projection of coreference chains: assessing strategies and testing opportunities

“…There have been some influential work on annotation projection for different NLP tasks which performed quite well cross-lingually, e.g. for semantic role labelling (Akbik et al, 2015) or syntactic parsing (Lacroix et al, 2016). At the same time, several recent studies on annotation projection for coreference have proven it to be a more difficult task than POS tagging or syntactic parsing, which is hard to be tackled by projection algorithms.…”

Towards Semi-Automatic Generation of Proposition Banks for Low-Resource Languages

“…For the purpose of comparison to previous work on high-resource languages, we replicate 1 Common problems for non-experts that we observe in our initial experiments involve ambiguities caused by implicit or causal role-predicate relationships, as well as figurative usage and hypotheticals. earlier evaluation practice and English preprocessing steps (Akbik et al, 2015). After projection, we randomly select 100 sentences for each target language and pass them to a curation step by 2 nonexperts.…”

“…For this reason, previous work defined lexical and syntactic constraints to increase projection quality. These include verb filters to allow only verbs to be labeled as frames (Van der Plas et al, 2011), heuristics to ensure that only heads of syntactic constituents are labeled as arguments (Padó and Lapata, 2009) and the use of verb translation dictionaries (Akbik et al, 2015) to constrain frame mappings. Adaptation to low-resource languages.…”

Translate and Label! An Encoder-Decoder Approach for Cross-lingual Semantic Role Labeling

“…We compare the increase in F1 score when training models with different amounts of additional data. We also add a comparison of the improvement achieved when adding the same amount of sentences produced by the labeled projection method of Akbik et al (2015). We see in Table 6 that adding our German data shows improvement in F1 score, despite the fact that the CoNLL-09 la- 13 The label distribution is given in the Supplement, A.3.…”