# Compositional Coding Capsule Network with K-Means Routing for Text Classification

arXiv: https://arxiv.org/abs/1810.09177 · HTML: https://mlsanity.com/paper/1810.09177/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Oct 2018.

## Built on by (most cited first, 5 of 5)

- [Graph Routing between Capsules](https://mlsanity.com/paper/2106.11531/) (2021)
- [IDoFew: Intermediate Training Using Dual-Clustering in Language Models for Few Labels Text Classification](https://mlsanity.com/paper/2401.04025/) (2024)
- [CNN-Trans-Enc: A CNN-Enhanced Transformer-Encoder On Top Of Static BERT representations for Document Classification](https://mlsanity.com/paper/2209.06344/) (2022)
- [A Dual-Directional Context-Aware Test-Time Learning for Text Classification](https://mlsanity.com/paper/2503.15469/) (2025)
- [ADE: Adaptive Dictionary Embeddings -- Scaling Multi-Anchor Representations to Large Language Models](https://mlsanity.com/paper/2604.24940/) (2026)

## Abstract (arXiv)

Text classification is a challenging problem which aims to identify the category of texts. In the process of training, word embeddings occupy a large part of parameters. Under the limitation of limited computing resources, it indirectly limits the ability of subsequent network designs. In order to reduce the number of parameters, the compositional coding mechanism has been proposed recently. Based on this, this paper further explores compositional coding and proposes a compositional weighted coding method. And we apply capsule network to model the relationship between word embeddings, a new routing algorithm, which is based on k-means clustering theory, is proposed to fully mine the relationship between word embeddings. Combined with our compositional weighted coding method and the routing algorithm, we design a neural network for text classification. Experiments conducted on eight challenging text classification datasets show that the proposed method achieves competitive accuracy compared to the state-of-the-art approach with significantly fewer parameters.