On-device Speech-to-Intent engine powered by deep learning

RCal is a standalone, interactive Python CLI application that calculates the optimal "Pró-labore" (administrator salary) and taxes for a Brazilian tech company operating under the Simples Nacional tax regime (Anexo III, Exporting Services, subject to the "Fator R" rule).

How knowledge of possible values can improve spoken slot filling without fine-tuning speech recognition models? Research on public transport data in Czech. Novel architecture using Whisper + additional decoder layer that toggles trie-constrained decoding. Includes dataset loading, model architecture, training, inference and evaluation.

Pytorch implementation of JointBERT: "BERT for Joint Intent Classification and Slot Filling"

Chinese Spoken Language Understanding, a sequence-labeling task based on RNN/LSTM/GRU and Transformer.

Open source code for EMNLP-19 Paper "A Stack-Propagation Framework with Token-Level Intent Detection for Spoken Language Understanding".

Intent Classification using the Fluent Speech Commands Dataset

EMNLP-2020: Cross-lingual Spoken Language Understanding with Regularized Representation Alignment

语义理解/口语理解，项目包含有词法分析：中文分词、词性标注、命名实体识别；口语理解：领域分类、槽填充、意图识别。

Dynamic Time-Aware Attention to Speaker Roles and Contexts for Spoken Language Understanding