Bengali.AI Competitions

Bengali.AI is constantly pushing the boundaries of Bengali language technologies by hosting exciting research competitions in partnership with national and international academic institutions, industries, and government collaborators. These competitions are designed to expedite the creation of resources and solutions that will revolutionize the field. Competitors are encouraged to experiment with various approaches to surpass our hidden test-sets and document their discoveries in research papers. All solutions are open-sourced to ensure continuity of research and support the field. And let’s not forget about the prize money! It’s a win-win situation for everyone involved.

Your work in this competition could have an impact beyond speech recognition improvements for one of the world's most popular, yet low-resource languages. You could also provide a much-needed push towards solving one of speech recognition's major challenges, out-of-distribution generalization.

An image segmentation task. Layout object (paragraph, text-box, image and table) segmentation from a large dataset with samples from the last 150 years. A great opportunity to the development of Bangla Optical Character Recognition(OCR) systems.

The objective of the task is to detect substrings that contain grammatical or spelling errors. The participating teams were presented with a comprehensive dataset of 100k sentences that were meticulously annotated at the word-level.

The first competition on Bengali ASR modeling. This competition won $5000 from Kaggle as a top community competition for that month.

For this competition, you’re given the image of a handwritten Bengali grapheme and are challenged to separately classify three constituent elements in the image: grapheme root, vowel diacritics, and consonant diacritics. ~500k samples in the dataset.

In this competition, your goal is to correctly identify digits from NumtaDB. Essentially a Bengali MNIST, but harder due to the presence of different domains. The competition was a special track for ICBSLP 2018 conference.