RFP for Technology Development in Indian Languages

1
RFP for Technology Development in Indian
Languages
BANASTHALI VIDYAPITH-RAJASTHAN Ph.
01438-228647/228648 Fax
01438-228365 website www.banasthali.org/
www.ITinHindi.org
2
Banasthali Vidyapith, an institution deemed to be
university is engaged in the projects related to
Technology Development in Indian Languages for
more than last two decades wherein its
contribution has been marked by
Banasthali Vidyapith and IT in Hindi

• Content development in Hindi
• Tools Applications development for Computing in
  Hindi
• Human Resource development for TDIL
• Capacity building for IT enabled services through
  following projects
• Dissemination by holding IT Localization Clinics

3
Over these years Banasthali Vidyapith has
built sufficient infrastructure and manpower to
emerge as a Centre for Research Development in
Computing in Indian Languages - Certificate
Course on IT Localisation - Masters degree
courses including M.Tech.(Comp.Sc./Software
Engg.), M.Tech.(VLSI) - Research
4
Banasthali Vidyapith proposes

• the following two related projects
• to develop evaluation metrics for testing
  various technologies developed thus far in
• - Machine Translation(MT), English to
  Indian Languages
• - Cross lingual Information Retrieval
  (CLIR) between English and Indian languages.
• to design and develop few modules to augment
  Machine Translation and CLIR system from
  English to Indian languages
• - A Sense Disambiguator for English-Hindi
  translator
• - Bilingual Phrasal term dictionary
  (English-Hindi)

5

• 1.Title to develop evaluation metrics for
  testing various technologies developed thus far
  in
• Machine Translation(MT) English to Indian
  Languages
• Cross lingual Information Retrieval (CLIR)
  between English and Indian languages.
• Proposer Prof. Aditya Shastri
• Institution Banasthali Vidyapith
• Rajasthan-304 022
• Language/Language pair English-Hindi
• Components to Full test suits for
• be implemented the above two
• 
  
  .Contd

6
Evaluation metrics for MT and CLIR technologies

• Metric design for
• - Software (architecture, functionality)
• - Translation output in Target language (syntax,
  semantics, pragmatics) in conformity with ISO and
  W3C recommendations
• 
  evaluation metrics

7
W3C Internationalization model (118N)
Globalization
Localization
User interface testing
Localized functionality testing
International sufficiency testing
Global functionality testing
L10N
G11N
Localizability testing
evaluation metrics
8
Standard Evaluation measures for Machine
translation

• to obtain ratings on same scale for aspects of
  quality such as intelligibility and fidelity.
• to count the number of errors, typically by
  simply counting the number of corrections made by
  post-editor(s).
• to classify errors according to pre established
  classification scheme under the assumption that
  some errors are more important than others.

evaluation metrics
9

• The techniques to be used
• Metrics to be designed for
• Black box testing to test performance
• at lexical level
• at syntactic level
• at semantics level
• at pragmatic level

10

• (b) White box testing to evaluate
• Software architecture (Monotethic, Modular
  component based, Object oriented)
• Language dependability(embedded/separate)
• Expandability (documentation)
• Hardware implementability (e.g. VLSI design)
• Usage requirements (interfaces, post editing)

11

• The research out come will be
• a generalised CASE tool for evaluation of MT
  between a language pair
• English Indian language
• (ii) A database of English-Hindi test cases to
  test the translation at all the above four levels.

12

• The CASE tool design and development shall
  require
• The use of the technology to be tested as the
  process
• alongwith its lexical resources such as
  dictionaries, rule base etc.

test case input
MT/CLIR The technology to be tested
Translated output from the technology
difference
Evaluation
Compare with desired output
metrics for test criteria