Human language technologies
Advancing how machines understand and generate human language across speech, text, and translation.
Meeting Intelligence: Task Rephrasing
Exploratory project to ‘decontextualize’ tasks and to-do items identified in meeting transcriptions, and rewrite each of them in a single sentence to appear in a separate to-do list. We build upon pretrained seq2seq transformer models,…
Odia Speech Data and Model
As part of this release, Navana Tech and Microsoft Research India are open-sourcing 1648 hours of validated Odia speech dataset and a baseline model for Odia speech recognition. The speech dataset consists of recordings in…
LiST (Lite Self-Training)
We present a new method LiST for efficient fine-tuning of large pre-trained language models (PLMs) in few-shot learning settings. LiST significantly improves over recent methods that adopt prompt fine-tuning using two key techniques. The first…
LITMUS Predictor
LITMUS Predictor provides support for simulating performance in ~100 languages given training observations of the desired task-model. Each training observation specifies the finetuning-datasize + test-performance in different languages. Further, the tool provides support for constructing…
Stochastic Mixture-of-Experts
This PyTorch package implements Taming Sparsely Activated Transformer with Stochastic Experts.
Deep Neural Machine Translation
This PyTorch package implements Very Deep Transformers for Neural Machine Translation, to stabilize the large scale language model and neural machine translation training, as described in: Very deep transformers for neural machine translation