Neural Generation for Czech: Data and Baselines
This work addresses the under-explored area of non-English natural language generation, specifically for Czech, which is incremental as it provides foundational data and baselines for a new language domain.
The authors tackled the problem of neural text generation for Czech, a morphologically rich language, by creating the first Czech NLG dataset in the restaurant domain and developing baseline models that address inflection challenges, achieving results with BLEU scores up to 0.65 and human evaluation ratings around 4.0 out of 5.
We present the first dataset targeted at end-to-end NLG in Czech in the restaurant domain, along with several strong baseline models using the sequence-to-sequence approach. While non-English NLG is under-explored in general, Czech, as a morphologically rich language, makes the task even harder: Since Czech requires inflecting named entities, delexicalization or copy mechanisms do not work out-of-the-box and lexicalizing the generated outputs is non-trivial. In our experiments, we present two different approaches to this this problem: (1) using a neural language model to select the correct inflected form while lexicalizing, (2) a two-step generation setup: our sequence-to-sequence model generates an interleaved sequence of lemmas and morphological tags, which are then inflected by a morphological generator.