Attention-based joint learning system and method for intent detection and slot filling
MYPI2024007739A0Pending Publication Date: 2026-06-30UNIVERSITI TEKNOLOGI MALAYSIA
0 Cites 0 Cited by
Patent Information
- Application Number
- MYPI2024007739
- Authority / Receiving Office
- MY · MY
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2026-06-30
Abstract
The present invention relates to a system and method for joint intent detection and slot filling in natural language processing (NLP) tasks, utilizing deep learning techniques. The system integrates Bidirectional Long Short-Term Memory (BiLSTM) networks, scalar and vectorial attention mechanisms, and Convolutional Neural Networks (CNNs) to simultaneously predict both the intent of an input sequence and the slot labels for each token. The system comprises an embedding layer for transforming input dialogue into word embeddings, a BiLSTM layer for processing the sequence in both forward and backward directions, and an attention mechanism that assigns importance to words for both intent detection and slot filling. The CNN layer captures local patterns in the sequence, and an output layer predicts a probability distribution over possible intents and generates slot labels. The method includes transforming the input into word embeddings, processing it with BiLSTM and attention mechanisms, passing it through a CNN layer, and outputting intent and slot predictions. The present invention enhances the accuracy and efficiency of natural language understanding in applications such as dialogue systems and virtual assistants. Figure 1
Need to check novelty before this filing date? Find Prior Art
