539 research outputs found
Sort by
Deepfix: Statistical Post-editing of Statistical Machine Translation Using Deep Syntactic Analysis
No full textDeepfix is a statistical post-editing system for improving
the quality of statistical machine
translation outputs.
It attempts to correct errors in verb-noun valency using deep syntactic
analysis and a simple probabilistic model of valency.
On the English-to-Czech translation pair, we show that statistical post-editing of
statistical machine translation
leads to
an improvement
of the translation quality when helped by deep linguistic knowledge
Translation of "It" in a Deep Syntax Framework
Get PDFWe present a novel approach to the translation of the English personal pronoun it to Czech. We conduct a linguistic analysis on how the distinct categories of it are usually mapped to their Czech counterparts. Armed with these observations, we design a discriminative translation model of it, which is then integrated into the TectoMT deep syntax MT framework. Features in the model take advantage of rich syntactic annotation TectoMT is based on, external
tools for anaphoricity resolution, lexical co-occurrence frequencies measured on a large parallel corpus and gold coreference annotation. Even though the new model for it exhibits no improvement in terms of BLEU, manual evaluation shows that it outperforms the original solution in
8.5% sentences containing it
No Free Lunch in Factored Phrase-Based Machine Translation
No full textFactored models have been successfully used in many language pairs to improve translation quality in various aspects. In this work, we analyze this paradigm in an attempt at automating the search for well-performing machine translation systems. We examine the space of possible factored systems, concluding that a fully automatic search for good configurations is not feasible. We demonstrate that even if results of automatic evaluation are available, guiding the search is difficult due to small differences between systems, which are further blurred by randomness in tuning. We describe a heuristic for estimating the complexity of factored models. Finally, we discuss the possibilities of a "semi-automatic" exploration of the space in several directions and evaluate the obtained systems
Tools for Machine Translation Quality Inspection
No full textThis report describes Addicter, tool for automatic error detection and evaluation, providing its user also with graphical interface useful for browsing through the dataset
Naučíme počítač (cizí) řeči?
No full textA popularizing talk on computational linguistics and machine translation given at Science Café in Kladno
Chimera – Three Heads for English-to-Czech Translation
No full textThis paper describes our WMT submissions CU-BOJAR and CU-DEPFIX, the latter
dubbed "chimera" because it combines on three diverse approaches: TectoMT, a system
with transfer at the deep syntactic level of representation, factored
phrase-based translation using Moses, and finally automatic rule-based correction
of frequent grammatical and meaning errors.
We do not use any off-the-shelf system-combination method
Utilizing Source Context in Statistical Machine Translation
No full textCurrent methods for statistical machine translation typically utilize only a
limited context in the input sentence. Many language phenomena thus remain out
of their reach, for example long-distance agreement in morphologically rich
languages or lexical selection often require information from the whole source
sentence. In this work, we present an overview of approaches for including wider
context in SMT and describe our first experiments
Strojový překlad lepší než Google
No full textA brief description of our machine translation system that performed better than Google Translate in an English-to-Czech translation task
PhraseFix: Statistical Post-Editing of TectoMT
No full textWe present two English-to-Czech systems that took part in the WMT 2013 shared task: TectoMT and PhraseFix. The former is a deep-syntactic transfer-based system, the latter is a more-or-less standard statistical post-editing (SPE) applied on top of TectoMT. In a brief survey, we put SPE in context with other system combination techniques and evaluate SPE vs. another simple system combination technique: using synthetic parallel data from TectoMT to train a statistical MT system (SMT). We confirm that PhraseFix (SPE) improves the output of TectoMT, and we use this to analyze errors in TectoMT. However, we also show that extending data for SMT is more effective
Two Case Studies on Translating Pronouns in a Deep Syntax Framework
Get PDFWe focus on improving the translation of the English pronoun it and English reflexive pronouns in an English-Czech syntax-based machine translation framework. Our evaluation both from intrinsic and extrinsic perspective shows that adding specialized syntactic and coreference-related features leads to an improvement in trans-
lation quality