MT and Overcoming Language Barrier

Brief History of Machine Translation

Anusaaraka: a tool to reduce language barrier

Anusaaraka: A better approach to Machine Translation

Some examples MT systems

TAUM METEO

• High quality
• Fully automatic
• very limited domain (Weather bulletins from English to French)

SYSTRAN

• Practically being used by Google, altavista, etc.
• Domain independent
• translation is unreliable (leading to accidents!)

MT efforts in India at a glance

          DOMAIN SPECIFIC

• Mantra system (C-DAC) Translation of govt. appointment letters -- Tree Adjoining Grammar
• Public health compaign documents -- Angla Bharati approach (ERDCI & IIT Kanpur)

APPLICATION SPECIFIC

• Matra (Human aided MT) (NCST,now C-DAC)

GENERAL PURPOSE MT

• Angla Bharati approach (IIT Kanpur)
• UNL based MT (IIT Bombay)
• Shiva: EBMT (IIIT Hyderabad/IISc Bangalore)
• Shakti: English-Hindi MT system (IIIT Hyderabad)

Difficulties in Machine Translation

• Why only limited success in Machine Translation efforts?
  

The Machine Translation efforts are aiming at reducing the barrier to ZERO!
( Aiming at building a machine that can walk like a human!). This is an impossible task!

Problems in analysis of the source language text:

• Ambiguity 

  • Raama phala khaataa hai.
    Machine has difficulty in identifying kartaa and karma.
    Human being uses world knowledge to resolve this.

  • Time flies like an arrow.
    

    Possible parses:

    • a)

      Time	flies	like	an	arrow
      N	V	Prep	Det	N

    • b)

      Time	flies	like	an	arrow
      N	N	V	Det	N

    • c)

      Time	flies	like	an	arrow
      V	N	Prep	Det	N

      (flies are like an arrow)
       

    • d)

      Time	flies	like	an	arrow
      V	N	Prep	Det	N

      (manner of timing)
      Clue: World Knowledge and Context

  • He saw a man on the hill with a telescope.
    Possible readings:
     
    • saw with a telescope
    • man with a telscope
    • hill with a telescope
       

    Clue: World Knowledge and Context

  • Patient had a stiff neck and fever
    • patient had stiff neck and fever
       
    • Patient had stiff neck and fever

    Clue: World Knowledge

  • Table the resolution
    U.S.A.: To postpone
    U.K. : To present before the committee
    Clue: British Versus American Usages
  • Dereferencing:
    • The farmer's wife sold the cow because she needed money.
      The farmer's wife sold the cow because she was not giving milk.
      Clue: World Knowledge
    • The mother with babies under four ...
      The mother with babies under forty ...
      Clue: Common Sense
    • William James cited Mozart's discussion on his composition.
      Marvin Minsky cited McCorduck's discussion on his reasearch.
      Clue: Historical Background
       
    • Even Human beings commit mistakes!
      Misinterpretation of Delphic Oracle by King Croesus:
      If Croesus attacks the Persians, he will destroy a mighty empire.
      
      Croesus's interpreted:
      a mighty empire: a mighty empire of Persians
      
      Another reading which he overlooked:
      a mighty empire: a mighty empire of his own!
       
• Ill Formedness
  • Incomplete/ungrammatical sentences
    Doctors's diagnosis notes:
    • stiff neck and fever.
    • Brain scan negative.