Multi-Alignment Templates Induction

Abstract

This paper examins approaches for translation between English and morphology-rich languages. Experiment with English-Russian and English-Lithuanian revels that "pure" statistical approaches on 10 million word corpus gives unsatisfactory translation. Then, several Web-available linguistic resources are suggested for translation. Syntax parsers, bilingual and semantic dictionaries, bilingual parallel corpus and monolingualWeb-based corpus are integrated in one comprehensive statistical model. Multi-abstraction language representation is used for statistical induction of syntactic and semantic transformation rules called multi-alignment templates. The decodingmodel is described using the feature functions, a log-linear modeling approach and A * search algorithm. An evaluation of this approach is performed on the English-Lithuanian language pair. Presented experimental results demonstrates that the multi-abstraction approach and hybridization of learning methods can improve quality of translation.

Semi-Automatic Ontological Alignment of Digitized Books Parallel Corpora

“…These models can robustly perform alignments on the bilingual corpus which usually represents a technical (literal) translation. But when we use a corpus of fiction books, these algorithms frequently give sparse and erroneous alignments (Laukaitis & Vasilecas, 2008;Laukaitis et al, 2011).…”

Sentence Level Alignment of Digitized Books Parallel Corpora

“…Nevertheless, we found that precision and recall of existing algorithms are too low in order to consider them practical when it comes to an alignment of translated fiction books. As shown in Laukaitis and Vasilecas (2008), Laukaitis et al (2011), the accuracy of these methods decreases drastically when we try to align a text that contains discrepancies, e.g. some book page layout segmentation strings, missing sentences and frequent one-to-many alignments.…”

Structured Data Extraction from Template-Generated Web Pages

“…Similar problem is studied by (Paradauskas, Laurikaitis 2006) were they analysed the process of enterprise knowledge extraction from relational database and source code of legacy information systems. (Laukaitis, Vasilecas 2008) presented a system that leverages extracted information from an online linguistic resource (Wordnet 10 ) to induce syntactic and semantic transformation rules. These rules are then used to improve machine learning based language translation.…”