Let MT simplify and speed up your Alignment for TM creation
2894Authors
nan
Abstract
AbstractLarge quantities of multilingual legal documents are waiting to be regularly aligned and used for future translations. For reasons of time, effort and cost, manual alignment is not an option. Automatically aligned segments are suitable for concordance search but are unreliable for fuzzy search and pretranslation. MT-based alignment could be the key to improving the results.
Keywords
Full Content
Use MT to simplify and speed up your alignment for TM creation Judith Klein STAR Group Wiesholz 35, 8262 Ramsen Switzerland judith.klein@star-group.net Giorgio Bernardinello STAR Group Wiesholz 35, 8262 Ramsen Switzerland giorgio.bernardinello @star-group.net Abstract Large quantities of multilingual legal documents are waiting to be regularly aligned and used for future translations. Due to time restraints and the effort and cost required, manual alignment is not an option. Automatically aligned segments are suitable for concordance search but are unreliable for fuzzy search and pre- translation. MT-based alignment could be the key to improving the results. 1 Why align if there is MT? Today, in the translation sector, many texts and documents in different sectors and areas and in many different language combinations are trans- lated using MT or with MT support. Why go to the effort of using alignment to create a Transla- tion Memory (TM) for a CAT tool from existing translations? This is because precisely the exist- ing verified translations are required. MT can probably provide precisely these translations by chance, but not reliably. 2 What data and for what purpose? The Systematic Collection of Legislation (SR)1 of the Swiss Federal Administration (Bund) is currently automatically aligned and provided as a TM2 four times a year. It consists of over 5000 MS Word documents in German, French and Italian, each with over two million segments or almost 20 million words3. The final version of © 2020 The authors. This article is licensed under a Creative Commons 3.0 licence, no derivative works, attri- bution, CC-BY-ND. 1 https://www.admin.ch/gov/de/start/bundesrecht/sy 2 stematische-sammlung.html Previously in MultiTrans. 3 Version: March 2020 the legal text is synchronised or translated into all three languages in the original document. Since the legal texts have been automatically aligned and not translated in the CAT tool, there is no TM with reliable segment alignment. The TM can therefore be used via concordance search4 but is not suitable for fuzzy search or pretranslation. 3 STAR's translation technology From 2020, the Bund will use STAR's translation technology as its general translation solution with the translation memory tool Transit as the core system. An alignment tool is integrated into Transit as standard. The MT interface makes it easy to use machine translations from different MT systems. 4 Interactive alignment in Transit In alignment projects, the document pairs are imported in the respective languages and segmented independently of each other. The segment alignment then does not match, for example, if two sentences in one document correspond to just one sentence in the other document, if a sentence has no equivalent in the other text, or if the sentence order is different. As usual with alignment tools, formal (e.g. sentence length, numbers, formatting) and lexical parameters (e.g. dictionary entries or unchanged words, such as company names) are used to cal- culate the segment alignment. 4 In Transit, the user can open the aligned document pair in the Transit editor directly from the concordance re- sult and copy the corresponding section of text. The same applies …
Content truncated to 500 words
Labeled Text
AbstractLarge quantities of multilingual legal documents are waiting to be regularly aligned and used for future translations. For reasons of time, effort and cost, manual alignment is not an option. Automatically aligned segments are suitable for concordance search but are unreliable for fuzzy search and pretranslation. MT-based alignment could be the key to improving the results.
Publication Details
- Paper ID:
- 2894
- Year:
- 2020
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan