Changes between Version 2 and Version 3 of en/MachineTranslation


Ignore:
Timestamp:
Jun 5, 2014, 3:16:21 PM (7 years ago)
Author:
xkocinc
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • en/MachineTranslation

    v2 v3  
    66 * Statistical machine translation
    77 * Extension of translation memories
    8  * Domain-speci
    9 c machine translation
     8 * Domain-specic machine translation
    109 * Machine translation between close languages
    1110 * Sub-word level machine translation
     
    1918 * baseline SMT available for everyone
    2019 * languages with high number of wordforms need special treatment
    21  * '''dělám, děláš, dělal, dělajícímu, dělaje, děláním, ...'''
     20 * '''dělám, děláš, dělal, dělajícímu, dělaje, děláním, ...'''
    2221
    2322   '''magas, magasabb, legmagasabb, legeslegmagasabb, ...'''
     
    3029
    3130[[Image(/trac/research/raw-attachment/wiki/en/MachineTranslation/word_matrix1.png)]]
     31
     32
    3233[[Image(/trac/research/raw-attachment/wiki/en/MachineTranslation/word_matrix2.png)]]
     34
     35
    3336[[Image(/trac/research/raw-attachment/wiki/en/MachineTranslation/word_matrix3.png)]]
    3437
    3538
    36 == Domain-speci
    37 c machine translation ==
     39== Domain-specic machine translation ==
    3840 * straightforward way of increasing quality of MT
    39  * domain-speci
    40 c corpora can be downloaded on demand
     41 * domain-specic corpora can be downloaded on demand
    4142 * separate models for each domain: ''' sports, cooking, gardening'''
    4243 * one sense per domain: '''bat'''
     
    4647 * translations of
    4748   * product details, product descriptions in e-shops,
    48    * manuals, warranty certi
    49 cates,
     49   * manuals, warranty certicates,
    5050   * user interface localizations, ...
    5151
     
    5454 * West and South Slavic languages: Czech, Slovak, Polish, Serbian, Croatian, Slovene
    5555 * MT mainly on word level, structure is very similar
    56  * di
    57 erences can be described systematically by rules:
    58    '''hraje na klavíri''' <-> '''hraje na klavír'''
     56 * dierences can be described systematically by rules:
     57   '''hraje na klavíri''' <-> '''hraje na klavír'''
    5958 * billion-word corpora available for these languages
    6059 * dictionaries can be generated semi-automatically
     
    7978== Conclusions ==
    8079 * generating new segments for translation memories
    81  * domain-speci
    82 c translation
     80 * domain-specic translation
    8381 * translation between close languages
    8482 * sub-word level translation