Our Vision

In India, every year, more than 1 crore students appear in CBSE, ICSE, and state board class 10th and 12th exams. Out of these students, more than 65% students study in non-English medium schools (link). Often, such students when join higher education institutes where the primary language of instruction is English, face language barriers. Lack of textbooks and other educational resources in native languages, add more to the issue for these students. Even when they wish to understand the content from supplementary resources such as the internet or trying to keep up with contemporary research, they face the similar issue. We are leaving behind such potential minds.


The lack of native language instruction also has a significant economic impact. Of the top twenty countries in GDP per capita, all of them use the language of the common people for higher and technical education. Most of these countries do not use English. In the bottom twenty economies, most of the countries are those which persist with a colonial language, not the mother tongue of the majority of people. The transition to mother tongue education requires significant investment in language infrastructure.


To succeed in this initiative, we must have textbooks resources with the entire engineering and medical curriculum translated into the many Indian languages. In addition to this, we must set up a strong ecosystem which can help access the content in native languages.

Goals

  • Web Platform with complete translation workbench
  • Machine Learning models for translation and post editing to augment human translators
  • Domain specific Linguistic resource (Dictionaries, translation memory) for efficient and consistent quality translation
  • To setup parallel corpus in technical domain (>30,00,000 words) to speed up the translation and post editing research
  • To translate and publish 500 books in native languages for engineering curriculum and make it accessible in cost effective manner
  • Active ecosystem of seed translators, reviewer, proof readers around the platform
  • Related Projects

    Akshar Anveshini

    Optical Character Recognition (OCR) and Post-editing system for the Sanskrit language.

    Inhouse OCR(Optical Character Recognition) model for the digitization of Sanskrit books.It uses state of the art technology that gives us 95.14% character level accuracy for Sanskrit text.

    IndicOCR

    Optical Character Recognition for Indian Texts

    End-to-end framework for Error Detection and Corrections in Indic-OCR

    More Information

    Credentials to use the software can be provided upon acknowledgement/request.

    DECILE

    Data Efficient Machine Learning

    Data Efficient Learning with Less Data State of the art AI and Deep Learning are very data hungry. This comes at significant cost including larger resource costs (multiple expensive GPUs and cloud costs), training times (often times multiple days), and human labeling costs and time. Decile attempts to solve this by answering the following question. Can we train state of the art deep models with only a sample (say 5 to 10\%) of massive datasets, while having neglibible impact in accuracy? Can we do this while reducing training time/cost by an order of magnitude, and/or significantly reducing the amount of labeled data required?

    MALTA

    Multi-Modal And Multi-Lingual Temporal Sentence Alignment

    For localizing sentences/captions in videos that leverages both audio and video modalities and that can generalize to new and possibly low-resource language settings.Moreover, it is a rich new dataset, whose annotation is driven by both audio and visual modalities and which is richer in the audio modality than previous datasets.

    RUDDER

    cRoss lingUal viDeo anD tExt Retrieval.

    RUDDER consists of 492(to be updated) videos, with an average length of 80 seconds and around 7 sentences describing every video. A video is present in multiple languages namely HINDI, TAMIL, MALAYALAM, URDU, KANNADA. (Note: All videos do not have audio in every language mentioned earlier)

    Publications

    Publications related to the projects.

    Gardua Udaan

    Gardua Prakashan

    GLISTER: Generalization based Data Subset Selection for Efficient and Robust Learning

    Krishnateja Killamsetty, Durga Sivasubramanian, Ganesh Ramakrishnan, Rishabh Iyer

    In Proceedings of The 35th AAAI Conference on Artificial Intelligence (AAAI 2021).

    Meta-Learning for Effective Multi-task and Multilingual Modelling

    Ishan Tarunesh, Sushil Khyalia, Vishwajeet Kumar, Ganesh Ramakrishnan, Preethi Jyothi

    In Proceedings of The 16th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2021)

    Caption Alignment for Low Resource Audio-Visual Data

    Vighnesh Reddy Konda, Mayur Warialani, Rakesh Prasanth Achari, Varad Bhatnagar, Jayaprakash Akula, Ganesh Ramakrishnan, Pankaj Singh, Gholamreza Haffari and Preethi Jyothi

    In Proceedings of The 21st INTERSPEECH Conference (Interspeech 2020), Shanghai, China.

    OCR On-the-Go: Robust End-to-end Systems for Reading License Plates & Street Signs

    Rohit Saluja, Ayush Maheshwari, Ganesh Ramakrishnan, Parag Chaudhuri and Mark Carman

    In Proceedings of The 15th International Conference on Document Analysis and Recognition (ICDAR 2019), Sydney, Australia.

    StreetOCRCorrect: An Interactive Framework forOCR Corrections in Chaotic Indian Street Videos

    Pankaj Singh, Bhavya Patwa, Rohit Saluja, Ganesh Ramakrishnan, Parag Chaudhuri and Mark Carman

    In Proceedings of The 2nd International Workshop on Open Services and Tools for Document Analysis, associated with the 15th International Conference on Document Analysis and Recognition (ICDAR-OST 2019), Sydney, Australia.

    Learning From Less Data: Diversified Subset Selection and Active Learning in Image Classification Tasks

    Vishal Kaushal, Rishabh Iyer, Anurag Sahoo, Khoshrav Doctor, Narasimha Raju, Ganesh Ramakrishnan

    Accepted paper at the 7th IEEE Winter Conference on Applications of Computer Vision (WACV), 2019, Hawaii, USA.

    Sub-word Embeddings for OCR Corrections in Highly Fusional Indic Languages

    Rohit Saluja, Mayur Punjabi, Mark Carman, Ganesh Ramakrishnan and Parag Chaudhuri

    In Proceedings of The 15th International Conference on Document Analysis and Recognition (ICDAR 2019), Sydney, Australia.

    Improving the learnability of classifiers for Sanskrit OCR corrections

    Devaraja Adiga, Rohit Saluja, Vaibhav Agrawal, Ganesh Ramakrishnan, Parag Chaudhuri, K. Ramasubramanian and Malhar Kulkarni

    Proceedings of the 17th World Sanskrit Conference, Vancouver, 2018.

    Time Aggregation Operators for Multi-label Audio Event Detection

    Pankaj Joshi, Digvijaysingh Gautam, Ganesh Ramakrishnan and Preethi Jyothi

    In Proceedings of The 21st INTERSPEECH Conference (Interspeech 2018), Shanghai, China.

    Synthesis of Programs from Multimodal Datasets

    Shantanu Thakoor, Simoni Shah, Ganesh Ramakrishnan, Amitabha Sanyal

    In Proceedings of the 32nd AAAI Conference on Artificial Intelligence (AAAI-18), New Orleans, Louisiana, USA

    Error Detection and Corrections in Indic OCR using LSTMs

    Rohit Saluja, Devaraj Adiga, Parag Chaudhuri, Ganesh Ramakrishnan and Mark Carman

    International Conference on Document Analysis and Recognition (ICDAR) 2017, Kyoto, Japan.

    A Framework for Document Specific Error Detection and Corrections in Indic OCR

    Rohit Saluja, Devaraj Adiga, Ganesh Ramakrishnan, Parag Chaudhuri and Mark Carman

    1st International Workshop on Open Services and Tools for Document Analysis (ICDAR- OST) 2017, Kyoto, Japan.

    A Framework for Error Detection and Corrections in Sanskrit

    Rohit Saluja, Devaraj Adiga, Parag Chaudhuri, Ganesh Ramakrishnan and Mark Carman

    Research and Innovation Symposium in Computing (RISC) 2017 (Most Admiring Poster Presentation Award), IIT-Bombay, India.

    Building Compact Lexicons for Cross-Domain SMT by mining near-optimal Pattern Sets

    Pankaj Singh, Ashish Kulkarni, Himanshu Ojha, Vishwajeet Kumar, Ganesh Ramakrishnan

    In Proceedings of the 20th Pacific Asia Conference on Knowledge Discovery and Data Mining (PAKDD) 2016

    Summarizing Multi-Document Topic Hierarchies using Submodular Mixtures

    Ramakrishna Bairi, Rishabh Iyer, Ganesh Ramakrishnan and Jeff Bilmes

    In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL), Beijing, China, July - 2015

    A Machine Assisted Human Translation System for Technical Documents.

    Kumar, V., Kulkarni, A., Singh, P., Ramakrishnan, G., & Arnaal, G.

    In Proceedings of the 8th International Conference on Knowledge Capture (p. 33). ACM.

    Enquiry

    Contact Us

    SI-A418, KReSIT, IIT-Bombay, Mumbai, Maharastra