A Conceptual Framework For Trie-Augmented Neural Networks (TANNS)

TL;DR

Trie-Augmented Neural Networks (TANNs) address interpretability and efficiency by embedding neural subnetworks at trie nodes to create hierarchical, modular decision processes. The paper provides a formal definition, validates the approach on XOR and AND/OR benchmarks, and demonstrates applicability to text classification on 20 Newsgroups and SMS Spam Collection, comparing with RNNs and FNNs. Key findings show that TANNs can match or slightly improve performance relative to traditional architectures while offering traceable decision paths, with performance influenced by trie depth, segmentation strategy, and computational cost. The work lays a conceptual foundation and outlines practical directions for optimization, scaling, and deployment in structured data tasks.

Abstract

Trie-Augmented Neural Networks (TANNs) combine trie structures with neural networks, forming a hierarchical design that enhances decision-making transparency and efficiency in machine learning. This paper investigates the use of TANNs for text and document classification, applying Recurrent Neural Networks (RNNs) and Feed forward Neural Networks (FNNs). We evaluated TANNs on the 20 NewsGroup and SMS Spam Collection datasets, comparing their performance with traditional RNN and FFN Networks with and without dropout regularization. The results show that TANNs achieve similar or slightly better performance in text classification. The primary advantage of TANNs is their structured decision-making process, which improves interpretability. We discuss implementation challenges and practical limitations. Future work will aim to refine the TANNs architecture for more complex classification tasks.

Figures (9)

• Figure 1: Illustration of the TANN architecture
• Figure 2: Convolutional Neural Network in Trie-Augmented Neural Network
• Figure 3: TANN Decision-Making
• Figure 4: Training Loss for the XOR Benchmark
• Figure 5: Training Loss for the AND/OR Benchmark