nlp.petamem.com
  Powered by PetaMem technology
Advertisement 

Wednesday 08 Feb 2012
MyNLP - guest
Register
Forgot password?
NLP Services & Tools
Dictionary
Spellchecker
Chatbot
Language Identification
Diacritisation
Number <=> Text
Special characters
Main menu
 Language 
 Theme 
Articles
News
Discussions
Advertisement

dot
Help page

Text Translation Module

Overview and Functionality

This module provides access to many very versatile text translation facilities. The required steps are very easy:

  • Choose the method of translation
  • Optionally - if you selected a MT-based method - choose the desired quality
  • Choose the target language
  • Input the text to be translated

You can choose between three major methods of translation:

  1. Machine Translation This option will cause the text you submit to be translated by a computer. This is the cheapest and fastest option to have your text translated. Depending on the underlying MT engine, you can also choose various levels of translation quality. See below for their explanation.
  2. Human-edited Machine Translation A very good compromise between cost and quality. Your text is translated by an MT system as with the option described above, but subsequently edited by a human professional translator, where glitches that MT might have left are corrected. Please read below about the extent of the editing works.
  3. Human Translation The highest quality you can get is by letting your text translate a professional translator. While this option is more expensive and slower than the previous two, our implementation and workflow guarantee the fastest and cheapest online translation you'll be able to find on the internet.

Depending on the underlying MT engine, you can also choose various levels of translation quality:

  1. Basic This is a very basic form - hardly MT. Basically it is a simple, word-by-word first-fit translation. No morphosyntactical processing or any other sophisticated methods. Using this method is very cheap and its usage domain could be to possibly grasp the meaning of the original text.
  2. Simple This mode ads morphological analysis and 2-word phrase lookahead functionality as well as a disambiguation based on statistical occurence data. There is no syntactical reorder however. Using this method is quite cost efficient, and the results should be more than sufficient to grasp the meaning of the original text. This quality level should be used as minimum if a human-edit is requested.
  3. Average Adds syntax reorder and n-word phrase lookahead to the simple mode. This quality grade provides output comparable with most modern MT systems and could be used in cases where users can accept contemporary MT quality.
  4. Good Adds elaborated disambiguation methods and the processing of procedural words. In real-world texts, results are slightly better than the "Average" quality and could be very well used in intranets where providing multilingual information at lowest cost is neccessary.
  5. Quality Adds semantic inference, named entity recognition and anaphora resolution. If these terms mean nothing to you, just be aware, that this is the bleeding edge of todays MT technology and will deliver results hitherto unseen from MT systems. The results are of publishable quality for casual content and need only very few human intervention to become perfect.
  6. Human-Grade The system proactively searches the internet for relevant texts and dictionaries in addition to all its local features, to provide also good translations for uncovered areas. The system extracts informations from these sources and learns the relevant facts about the topic - thus improving its own knowledge base.

You can either input (or cut & paste) the text to be processed into the textarea in the form, or upload a file from your local computer. In case you input both text and upload a file, the text is discarded and only the file is processed.

Examples

Restrictions