MSR-WS 2018 - Workshop on Multilingual Surface Realization (MSR-WS)

Natural Language Generation (NLG) is in the ascendant both as a stand-alone data-to-text or text-to-text task and as part of downstream applications (see, e.g., abstractive summarization, dialogue-based interaction, question answering, etc.). Only in 2017, three “deep” NLG shared tasks that focused on language generation from abstract semantic representations have been organized (although for English only): WebNLG, SemEval Task 9 , E2E. However, when compared to, e.g., parsing or machine translation, NLG still lags behind in terms of theoretical advances. Thus, while recent years witnessed a shift of the processing paradigm in these areas from traditional supervised machine learning techniques to deep learning techniques, NLG did not arrive there fully yet. Similarly, NLG still does not make full use of the available resources in the way, e.g., parsing does. For instance, the multilingual Universal Dependencies (UD) dataset has already been used for the CoNLL'17 parsing shared task. This dataset, which currently consists of 102 treebanks covering about 60 languages and can be downloaded freely, facilitates the development of large scale applications that work potentially across all of the UD treebank languages in a uniform fashion.
MSR-WS aims to change the situation and put NLG, and, in particular, surface generation, onto the main stream research agenda of Computational Linguistics, bringing together communities that hardly collaborated so far. It will provide a forum for the presentation of the results of the currently open multilingual Surface Realization Shared Task 2018 (SR’18) and of high quality papers on surface realization and related topics. SR’18 focuses on multilingual surface generation starting from UD treebanks. Since UDs are structures with a degree of abstraction that is targeted by state-of- the-art parsing, such that that the challenge to reverse neural network parsing algorithms for generation becomes a plausible research question, SR’18 solicits, apart from genuine generation approaches, contributions by the parsing community. SR’18 also aims to attract participants from other areas such as Computer Assisted Language Learning (and, in particular, grammatical error correction, since one of the tracks of the SR’18 is the generation of functional words such as bound prepositions and auxiliaries, whose correct introduction/omission is one of the primary challenges for language learners).
To complement the presentation of the SR’18 results, MSR-WS solicits contributions on all topics that are related to surface realization in NLG. Sought are presentations of cutting edge approaches that address problems of surface-oriented generation such as grammatical and/or information structure-driven word order determination, inflection, functional word determination, paraphrasing, etc. The presented works are expected to be a clear contribution to the progress in robust multilingual surface generation, i.e., be language-independent or easily portable from one language to another and clearly scalable. The topics of interest include, but are not limited to:
Linearization in NLG
Multilingual approaches to surface realization
Function word generation
Inflection in NLG
Joint generation from abstract representations
Surface-oriented text simplification
Surface-oriented spoken language generation
Application of surface realization for grammatical error correction
NLG in surface-oriented paraphrasing
Deep learning approaches to NLG