Development of English to Malayalam Machine Aided Translation System Using AnglaBharati Technology

1
Development of English to Malayalam Machine Aided
Translation System Using AnglaBharati
Technology

• Language Technology Centre
• C-DAC, Thiruvananthapuram

2
Proposer

• R. Ravindra Kumar
• Senior Director Head
• Language Technology Centre
• C-DAC, Thiruvananthapuram

Institution
Centre For Development of Advanced
Computing Vellayambalam Thiruvananthapuram
695 033
3
Language Pair English-Malayalam

• Component that will be implemented
• Text Generator for AnglaBharati Engine

4
Text Generator

• The text generator converts the pseudo code
  in PLIL (Pseudo Lingua for Indian Languages)
  generated by the AnglaBharati engine into the
  target language- Malayalam.
• Generator (For full sentences)
• Partial Generator ( For partial sentences)
• Post Generator (Deals with exceptional cases)

5

• Techniques Used
• Uses Paninian FrameWork
• Grammatical Rules as per Paninian Framework will
  applied for effecting translation
• Post-positions handling
• The post positions depends upon the nature of
  the verb and the Karak role of the preceding
  word/word group in the sentence. So for the
  categorization of words rules and heuristics are
  need to be devised.
• Sandhi handling
• The agglutinative nature of Malayalam makes
  sandhi handling a very significant part in the
  text generator. In this module we have to
  implement language rules for handling sandhi. The
  lexical resources needed in this module are
  sandhi rules and post position rules.

6
Modification needed in AnglaBharati Engine for
Implementing AnglaMalayalam

• Substitution of Hindi words/phrases with
  corresponding Malayalam words/phrases in PLIL and
  tables used by the following modules of the
  AnglaBharati engine
• Preprocessor
• Morphological Analyzer
• Example Base
• Rule Base
• Changes in Transliterate module
• The transliterate module transliterates the
  unknown acronyms and named entities that are not
  found in the lexical database and other
  resources.

7

• Performance of these techniques in other
  languages
• The AnglaHindi MAT system based on AnglaBharati
  technology generates approximately 90 acceptable
  translation in case of simple, compound and
  complex sentences up to a length of 20 words.
• Estimated Time
• 18 months

8
Schedule
9

• Preferred Domain
• Will be decided after identifying a user.
• Other Evaluation Metrics
• Intelligibility
• Accuracy
• Error Analysis

10
Angla-Bharati Features

• Exploits the structural similarity of ILs
• Inter-lingua based approach
• Uses a mixture of some Example Base translation
  methodology
• Provision for augmenting its abstracted example
  base specific to an application domain.
• Effort of 30 is required for a new English to
  Indian Language translator

11

• Thank you