Skit.ai

kaldi-serve

Server framework for Kaldi ASR Toolkit

speech-to-intent-dataset

Dataset Release for Intent Classification from Speech

emotion-tts-dataset

Dataset release for Emotional TTS in Indian Accent

slu-prosody

Code repository for the paper "Improving End-to-End SLU performance with Prosodic Attention and Distillation" accepted at Interspeech 2023.

vcore

Common, utility packages for Go

dialogy

Language understanding toolkit for human dialogs.

Map-Mix

The official implementation of the method discussed in the paper Improving Spoken Language Identification with Map-Mix(work accepted at ICASSP-2023)

phone-number-entity-dataset

Dataset Release for Phone Number Entity capture task

several-people-are-talking

Audio archive of reading club discussions at Vernacular.ai

tech

Skit's tech website

ccb

Slack bot for virtual coffee pairing

eevee

Evaluation utilities

tts-middleware

Middleware module for our speech synthesis systems

dialogy-template-simple-transformers

Dialogy template using simple transformers.

handbook

How Skit operates and other notes

skit-pipelines

Reusable workflows for ml using kubeflow components.

skit-auth

Authorization for skit.ai's platform.

skit-labels

Command line interface for interacting with Label studio and (deprecated) tog data server

Improving-Numeric-ASR-via-Cues

Code and datasets for paper "Improving Alphanumeric Automatic Speech Recognition Via Cues"

onboarding

Artifacts from onboarding sessions

sandbox

Slackbot for generating curated, pre-tagged datasets, using the Outbound Dialler API

skit-calls

Library to fetch calls from a given environment.

Diversity-of-ASR-Hypotheses-SLU

Code for baselines and public datasets for upcoming paper "ASR-Robust Spoken Language Understanding via Self-Knowledge Distilation"

faqs

go-grpc-prometheus

Prometheus monitoring for your gRPC Go servers.

gosip

SIP in Go

indic-num2words

Modules to convert Numbers to Words for all Indian Languages.

johnny

faster audio downloads & processing.

RWKV-LM

Fork of RWKV-LM to finetune the LLM on conversational and API data

whisper