狠狠撸

狠狠撸Share a Scribd company logo

Transzaar - CAT Tool for Indian Languages including English Arabic

?
?
Rashid Ahmad
Rashid Ahmad

Transzaar is an AI-powered computer-aided translation (CAT) tool that aims to empower human translators. It offers features like integrated machine translation and translation memory to help translators work more efficiently. Transzaar also provides linguistic analysis tools and lets translators customize resources. Currently it is being used by over 65 translators and helps improve their productivity by more than 2 times. The creators plan to expand it to support factory-style translation workflows in the future.

1 of 13
Download to read offline
Transzaar: Empowers Human Translators Rashid Ahmad, Priyank Gupta, Nagaraju Vuppala, Sanket Kumar Pathak, Ashutosh Kumar, Gagan Soni, Sravan Kumar, Manish Shrivastava, Avinash K Singh, Arbind K Gangwar, Pawan Kumar, Mukul K Sinha IIIT Hyderabad INDIA, eBhasha Setu INDIA The 18th International Conference on Computational Science and Its Applications (ICCSA 2018), Melbourne, Monash University, Australia 2-5-July-2018 For more details visit at http://www.transzaar.com
Outline ? Introduction ? Related Work ? Background and Motivation ? Transzaar: Problems, Solutions and Features ? Evaluation and Usage ? Conclusion and Future Work
Introduction ? Digital Content on Internet Accessible using devices (Desktop/Mobile/Tablet) through telecom data services provider. Understandability is big issue as content is not available to people in their native language. ? Content Creation (Native/local language) – Writing the fresh content – By translating the existing content into local languages. ? Transzaar - An AI powered CAT tool – It offers Computer Aided Translation (CAT) functionality. – Using Transzaar translators can post-edit MT output easily while delivering with better turn-around-time. 3
Related Work ? ALPS system developed in 1981 it was first CAT Tool designed for personal computer. ? CASMACAT is a modular, web-based translation workbench that offers advanced functionalities for computer-aided translation. ? PET is a post-editing tool, which can be used for evaluating the quality of translations in terms of how much effort these translations require in order to be fixed. ? MateCat is a tool whose objective is to improve the integration of machine translation (MT) and human translation within the so-called computer aided translation (CAT) framework. ? Anubis, provides detailed description of one of the state-of-art CAT systems. ? Some of the commercial available CAT tools are - SDL Trados, Wordfast, Memsource, DejaVu and SmartCat etc. 4
Background and Motivation ? MT system does not produce Publishable Quality Content. ? Hence, MT as technology, far from replacing humans, should be considered as a tool in the hands of human translators. ? This observation motivated our team to build this new tool Transzaar whose prime function would be -- to assist human translators. 5
Transzaar CAT Tool ? Transzaar - An AI powered CAT tool – It has linguistic tools/resources and feedback loop – It aids the process of post-editing thereby increasing the productivity of a human translator by 2 to 3 folds. – It collects user feedback continuously, which helps the MT system and tool to further learn and improve periodically, with the additional new dataset generated. 6
Transzaar: Translator Problems ? Mechanical Tasks – Search into the lexical databases and Translation Management. ? Repetitive Tasks – Name Entity and TERMINOLOGY. ? Cognitive Tasks – Translation from the scratch is an Recall . For human brain “Recognition is easy; Recall is hard”. 7
Transzaar : Solution Techniques ? Search – It provides one click search into multiple lexical resources for a selected/highlighted word or term. ? Visual Alerts – While the human translator is post editing the machine translated content all NER, TERMINOLOGY are flagged in the editing pane by visual alerts. ? Recall to Recognition – It is easier for the translator to select from the most likely correct results (or formulate from an approximate result) than to provide a solution on a blank slate. 8
Transzaar Features ? Editor Kit – Cut/copy/paste, find/replace, predictive typing, segmentation etc. ? Resources Integration – Multiple MT, TM and various resources integration provision is available. ? Concordance Lookup – Concordance lookup in the source/target language parallel aligned corpus. ? Web-based – Access from anywhere, anytime, platform independent, for each user. 9
Transzaar Distinct Features ? Text Analysis (Pre-Processing) – Extraction of Collocations, Terminologies, NEs, MWEs, etc. ? Transliteration – Transliteration of foreign language terms (mostly proper nouns), or technical terms into target language script. ? Spell Checker – Spell checking tool for Indian major languages ? Customization – Inject individual creativity with customized lexical resources or tools. 10
Evaluation and Usage ? Transzaar CAT Tool is being currently used by 65+ users. ? It is being used by a language service startup company providing translation service among Indian languages ? For an experienced translator with higher skill set, the tool helps to improve his productivity by more than 2 fold. ? As Tool have integrated its own MT systems for some language pairs (accuracy > 83%) at the back-end. We see higher improvement in productivity for the such language pairs. 11
Conclusion and Future Work ? We have introduced Transzaar, a next-generation AI powered tool that offers computer aided translation function to human translators. It helps to improve the productivity of translators. ? It also helps a human translator to improve the fluency and the accuracy of the machine translated content to match the naturalness of a native speaker. ? In future, our team plans to extend this tool to support factory mode translation to support multiple kinds of translation work-flow. 12
Thank You Questions? Questions and Queries rashid101b@gmail.com For more details visit at http://www.transzaar.com

More Related Content

Transzaar - CAT Tool for Indian Languages including English Arabic

  • 1. Transzaar: Empowers Human Translators Rashid Ahmad, Priyank Gupta, Nagaraju Vuppala, Sanket Kumar Pathak, Ashutosh Kumar, Gagan Soni, Sravan Kumar, Manish Shrivastava, Avinash K Singh, Arbind K Gangwar, Pawan Kumar, Mukul K Sinha IIIT Hyderabad INDIA, eBhasha Setu INDIA The 18th International Conference on Computational Science and Its Applications (ICCSA 2018), Melbourne, Monash University, Australia 2-5-July-2018 For more details visit at http://www.transzaar.com
  • 2. Outline ? Introduction ? Related Work ? Background and Motivation ? Transzaar: Problems, Solutions and Features ? Evaluation and Usage ? Conclusion and Future Work
  • 3. Introduction ? Digital Content on Internet Accessible using devices (Desktop/Mobile/Tablet) through telecom data services provider. Understandability is big issue as content is not available to people in their native language. ? Content Creation (Native/local language) – Writing the fresh content – By translating the existing content into local languages. ? Transzaar - An AI powered CAT tool – It offers Computer Aided Translation (CAT) functionality. – Using Transzaar translators can post-edit MT output easily while delivering with better turn-around-time. 3
  • 4. Related Work ? ALPS system developed in 1981 it was first CAT Tool designed for personal computer. ? CASMACAT is a modular, web-based translation workbench that offers advanced functionalities for computer-aided translation. ? PET is a post-editing tool, which can be used for evaluating the quality of translations in terms of how much effort these translations require in order to be fixed. ? MateCat is a tool whose objective is to improve the integration of machine translation (MT) and human translation within the so-called computer aided translation (CAT) framework. ? Anubis, provides detailed description of one of the state-of-art CAT systems. ? Some of the commercial available CAT tools are - SDL Trados, Wordfast, Memsource, DejaVu and SmartCat etc. 4
  • 5. Background and Motivation ? MT system does not produce Publishable Quality Content. ? Hence, MT as technology, far from replacing humans, should be considered as a tool in the hands of human translators. ? This observation motivated our team to build this new tool Transzaar whose prime function would be -- to assist human translators. 5
  • 6. Transzaar CAT Tool ? Transzaar - An AI powered CAT tool – It has linguistic tools/resources and feedback loop – It aids the process of post-editing thereby increasing the productivity of a human translator by 2 to 3 folds. – It collects user feedback continuously, which helps the MT system and tool to further learn and improve periodically, with the additional new dataset generated. 6
  • 7. Transzaar: Translator Problems ? Mechanical Tasks – Search into the lexical databases and Translation Management. ? Repetitive Tasks – Name Entity and TERMINOLOGY. ? Cognitive Tasks – Translation from the scratch is an Recall . For human brain “Recognition is easy; Recall is hard”. 7
  • 8. Transzaar : Solution Techniques ? Search – It provides one click search into multiple lexical resources for a selected/highlighted word or term. ? Visual Alerts – While the human translator is post editing the machine translated content all NER, TERMINOLOGY are flagged in the editing pane by visual alerts. ? Recall to Recognition – It is easier for the translator to select from the most likely correct results (or formulate from an approximate result) than to provide a solution on a blank slate. 8
  • 9. Transzaar Features ? Editor Kit – Cut/copy/paste, find/replace, predictive typing, segmentation etc. ? Resources Integration – Multiple MT, TM and various resources integration provision is available. ? Concordance Lookup – Concordance lookup in the source/target language parallel aligned corpus. ? Web-based – Access from anywhere, anytime, platform independent, for each user. 9
  • 10. Transzaar Distinct Features ? Text Analysis (Pre-Processing) – Extraction of Collocations, Terminologies, NEs, MWEs, etc. ? Transliteration – Transliteration of foreign language terms (mostly proper nouns), or technical terms into target language script. ? Spell Checker – Spell checking tool for Indian major languages ? Customization – Inject individual creativity with customized lexical resources or tools. 10
  • 11. Evaluation and Usage ? Transzaar CAT Tool is being currently used by 65+ users. ? It is being used by a language service startup company providing translation service among Indian languages ? For an experienced translator with higher skill set, the tool helps to improve his productivity by more than 2 fold. ? As Tool have integrated its own MT systems for some language pairs (accuracy > 83%) at the back-end. We see higher improvement in productivity for the such language pairs. 11
  • 12. Conclusion and Future Work ? We have introduced Transzaar, a next-generation AI powered tool that offers computer aided translation function to human translators. It helps to improve the productivity of translators. ? It also helps a human translator to improve the fluency and the accuracy of the machine translated content to match the naturalness of a native speaker. ? In future, our team plans to extend this tool to support factory mode translation to support multiple kinds of translation work-flow. 12
  • 13. Thank You Questions? Questions and Queries rashid101b@gmail.com For more details visit at http://www.transzaar.com