Bangla
PulseAugur coverage of Bangla — every cluster mentioning Bangla across labs, papers, and developer communities, ranked by signal.
12 day(s) with sentiment data
-
AI translation errors found across multiple languages on dev.to sites
A software development team discovered over a thousand mistranslations across their websites, where words from one language appeared within sentences of another. These errors, generated by automated translation models, …
-
Deep learning model detects fabricated Bangla news photocards
Researchers have developed a deep learning model to detect fabricated news photocards in Bangla, a common format for localized misinformation. The study created a dataset of authentic and fabricated photocards, finding …
-
New dataset and methods improve depression detection in Bengali social media
Researchers have developed BanglishDepYT, a new dataset and methodology for detecting depression in native Bengali social media comments. The approach, EPH (Emotion-Preserving Hybrid preprocessing), effectively handles …
-
Gated Knowledge Distillation Fails to Improve Low-Resource Bangla Generation
A new research paper explores gated knowledge distillation techniques for low-resource Bangla language generation, finding that these methods do not significantly improve upon standard knowledge distillation or cross-en…
-
New augmentation method boosts Bangla disease prediction accuracy
Researchers have developed a new method for improving disease prediction in Bangla by using structure-preserving augmentation techniques. This approach, tested on a benchmark of 725 profiles across 85 diseases, showed s…
-
New Benchmark Reveals Widespread Fabrication of Islamic Scripture by Open Language Models
A new benchmark called BILAL-BENCH has been developed to evaluate the ability of open-source language models to fabricate Islamic scripture, specifically in Bengali. The benchmark includes 240 questions covering Quranic…
-
Bengali coreference resolution research questions multilingual model assumptions
A new research paper on arXiv challenges common assumptions about multilingual transformer models, particularly for coreference resolution in low-resource languages like Bengali. The study found that a small amount of h…
-
New Bengali Speech Corpus 'Lipi-Ghor' Released with YouTube Data
Researchers have developed Lipi-Ghor, a substantial new corpus of Bengali speech data designed to address the under-resourced nature of the language in spoken language research. This corpus contains 885 hours of audio f…
-
KrishokTech system enhances Bengali agricultural advice with safety controls
A new agricultural advisory system named KrishokTech has been developed to provide advice in Bengali, focusing on safety and reliability. The system incorporates explicit authorization controls to manage potentially uns…
-
New Bengali Speech Corpus 'Lipi-Ghor' Released with Scalable Pipeline · 2 sources tracked
Researchers have developed Lipi-Ghor, a substantial new corpus of Bengali speech data, comprising 885 hours from nearly 1000 YouTube videos across 17 domains. This corpus includes speaker diarization and long-form audio…
-
Researcher explores quantization for multilingual LLMs on personal devices
A researcher is investigating methods to make multilingual language models more practical for personal devices by exploring quantization techniques. The initial focus is on the Tiny Aya model from Cohere, with a particu…
-
New LoWAgent benchmark reveals performance gaps in AI web agents for low-resource languages
A new benchmark called LoWAgent has been introduced to evaluate the performance of web agents across low-resource languages, addressing a gap in current testing methodologies. The benchmark includes six localized websit…
-
AI framework streamlines Bengali theatre production planning
Researchers have developed ThetreStageAI, a retrieval-augmented generation framework designed to assist in the structured planning of Bengali theatre productions. This system utilizes a multi-view approach, maintaining …
-
Bengali idioms reveal literal lock-in in decision models
A new research paper titled "Literal Lock-In: How Test Language Shifts Literal Attraction in Typed Decision Models on Bengali Idioms" explores the behavior of decision models, which are non-generative scorers used for t…
-
CoDeFont framework enables one-shot font generation across scripts
Researchers have developed CoDeFont, a novel framework for one-shot font generation that can synthesize target glyphs by transferring the style of a single reference glyph. This method addresses the challenge of capturi…
-
AI system aids dementia patients with Bangla speech analysis and sentence completion
Researchers have developed a novel system to assist individuals with dementia by analyzing their Bangla speech and text. The system utilizes a transformer-based model for speech-to-text transcription and evaluates vario…
-
New research tackles LLM hallucinations with advanced detection methods · 5 sources tracked
Researchers are developing new methods to detect and mitigate hallucinations in large language models (LLMs). One approach, Cross-Layer Attention Probing (CLAP), processes LLM activations across its entire residual stre…
-
New BANGLAPLAN benchmark tests LLM planning in Bangla
A new benchmark called BANGLAPLAN has been developed to evaluate the constraint-based planning and decision-making capabilities of large language models in the Bangla language. The benchmark consists of 1,000 problems a…
-
LLMs show varied performance in English vs. Bengali summarization
A new study published on arXiv evaluates the performance of eight large language models (LLMs) on multilingual abstractive summarization, focusing on English and Bengali. The research found significant cross-lingual per…
-
New Bengali cyberbullying corpus incorporates conversational context
Researchers have developed BenCoDiC, a new corpus designed to address the limitations of existing Bengali cyberbullying datasets. This large-scale resource comprises over 50,000 instances collected from Bengali politica…