Parallel text

A parallel text is a text placed alongside its translation or translations. Parallel text alignment is the identification of the corresponding sentences in both halves of the parallel text. The Loeb Classical Library and the Clay Sanskrit Library are two examples of dual-language series of texts. Reference Bibles may contain the original languages and a translation, or several translations by themselves, for ease of comparison and study; Origen's Hexapla (Greek for "sixfold") placed six versions of the Old Testament side by side. The most famous example is the Rosetta Stone.

Large collections of parallel texts are called parallel corpora (see text corpus). Alignments of parallel corpora at sentence level are prerequisite for many areas of linguistic research. During translation, sentences can be split, merged, deleted, inserted or reordered by the translator. This makes alignment a non-trivial task.

Four main corpora types can be distinguished.

A noisy parallel corpus contains bilingual sentences that are not perfectly aligned or have poor quality translations. Nevertheless, most of its contents are bilingual translations of a specific document.

A comparable corpus is built from non-sentence-aligned and untranslated bilingual documents, but the documents are topic-aligned.

A quasi-comparable corpus includes very heterogeneous and non-parallel bilingual documents that may or may not be topic-aligned.

The rarest parallel corpora are corpora that contain translations of the same document into two or more languages, aligned at the sentence level at least.

Large corpora used as training sets for machine translation algorithms are usually extracted from large bodies of similar sources, such as databases of news articles written in the first and second languages describing similar events.

However, extracted fragments may be noisy, with extra elements inserted in each corpus. Extraction techniques can differentiate between bilingual elements represented in both corpora and monolingual elements represented in only one corpus in order to extract cleaner parallel fragments of bilingual elements. Comparable corpora are used to directly obtain knowledge for translation purposes. High-quality parallel data is difficult to obtain, however, especially for under-resourced languages.

...
Wikipedia