539 research outputs found
Sort by
A Dataset Comparison for an Indonesian-English Statistical Machine Translation System
No full textIn this paper, we study the effect of incorporating morphological information on an Indonesian (id) to English (en) Statistical Machine Translation (SMT) system as part of a preprocessing module. The linguistic phenomenon that is being addressed here is Indonesian cliticized words. The approach is to transform the text by separating the correct clitics from a cliticized word to simplify the word alignment. We also study the effect of applying the preprocessing on different SMT systems trained on different kinds of text, such as spoken language text. The system is built using the state-of-the-art SMT tool, MOSES. The Indonesian morphological information is provided by MorphInd. Overall the preprocessing improves the translation quality, especially for the Indonesian spoken language text, where it gains 1.78 BLEU score points of increase
Using Parallel Features in Parsing of Machine-Translated Sentences for Correction of Grammatical Errors
No full textIn this paper, we present two dependency
parser training methods appropriate for parsing outputs of statistical machine translation (SMT), which pose problems to standard
parsers due to their frequent ungrammaticality. We adapt the MST parser by exploiting
additional features from the source language,
and by introducing artificial grammatical errors in the parser training data, so that the
training sentences resemble SMT output.
We evaluate the modified parser on DEPFIX, a system that improves English-Czech
SMT outputs using automatic rule-based corrections of grammatical mistakes which requires parsed SMT output sentences as its input. Both parser modifications led to improvements in BLEU score; their combination was evaluated manually, showing a statistically significant improvement of the translation quality
Selecting Data for English-to-Czech Machine Translation
No full textWe provide a few insights on data selection for
machine translation. We evaluate the quality
of the new CzEng 1.0, a parallel data source
used in WMT12. We describe a simple technique
for reducing out-of-vocabulary rate after
phrase extraction. We discuss the benefits
of tuning towards multiple reference translations
for English-Czech language pair. We
introduce a novel approach to data selection
by full-text indexing and search: we select
sentences similar to the test set from a large
monolingual corpus and explore several options
of incorporating them in a machine translation
system. We show that this method can
improve translation quality. Finally, we describe
our submitted system CU-TAMCH-BOJ
IDENTIC Corpus: Morphologically Enriched Indonesian-English Parallel Corpus
No full textThis paper describes the creation process of an Indonesian-English parallel corpus (IDENTIC). The corpus contains 45,000 sentences collected from different sources in different genres. Several manual text preprocessing tasks, such as alignment and spelling correction, are applied to the corpus to assure its quality. We also apply language specific text processing such as tokenization on both sides and clitic normalization on the Indonesian side. The corpus is available in two different formats: ‘plain’, stored in text format and ‘morphologically enriched’, stored in CoNLL format. Some parts of the corpus are publicly available at the IDENTIC homepage
Manually Classified Errors in Czech-Slovak Translation
No full textOutputs of five Czech-Slovak machine translation systems (Česílko, Česílko 2, Google Translate and Moses with different settings) for first 50 sentences of WMT 2010 testing set. The translations were manually processed and the errors were marked and classified according to the scheme by Vilar et al. (David Vilar, Jia Xu, Luis Fernando D’Haro, Hermann Ney: Error Analysis of Statistical Machine Translation Output, Proceedings of LREC-2006, 2006
Probes in a Taxonomy of Factored Phrase-Based Models
No full textWe introduce a taxonomy of factored phrase based
translation scenarios and conduct a
range of experiments in this taxonomy. We
point out several common pitfalls when designing
factored setups. The paper also describes
our WMT12 submissions CU-BOJAR
and CU-POOR-COMB
Improving Word Alignment by Exploiting Adapted Word Similarity
No full textThis paper presents a method to improve a word alignment model in a phrase-based Statistical Machine Translation system for a low resourced language using a string similarity approach. Our method captures similar words that can be seen as semi-monolingual across languages, such as numbers, named entities, and adapted/loan words. We use several string similarity metrics to measure the monolinguality of the words, such as Longest Common Subsequence Ratio (LCSR), Minimum Edit Distance Ratio (MEDR), and we also use a modified BLEU Score (modBLEU).
Our approach is to add intersecting alignment points for word pairs that are orthographically similar, before applying a word alignment heuristic, to generate a better word alignment.
We demonstrate this approach on Indonesian-to-English translation task, where the languages share many similar words that are poorly aligned given a limited training data.
This approach gives a statistically significant improvement by up to 0.66 in terms of BLEU score
Improving SMT by Using Parallel Data of a Closely Related Language
No full textThe amount of training data in statistical machine translation is critical for translation quality. In this paper, we demonstrate how to increase translation quality for one language pair by bringing in parallel data from a closely related language. In particular, we improve en→sk translation using a large Czech–English
parallel corpus and a shallow (rule-based) MT system for cs→sk. Several setup options are explored in order to identify the best possible configuration
Announcing Prague Czech-English Dependency Treebank 2.0
No full textWe introduce a substantial update of the Prague Czech-English Dependency Treebank, a parallel corpus manually annotated at the deep
syntactic layer of linguistic representation. The English part consists of the Wall Street Journal (WSJ) section of the Penn Treebank.
The Czech part was translated from the English source sentence by sentence. This paper gives a high level overview of the underlying
linguistic theory (the so-called tectogrammatical annotation) with some details of the most important features like valency annotation,
ellipsis reconstruction or coreference