Federated Non-negative Matrix Factorization for Short Texts Topic Modeling with Mutual Information

TL;DR

This paper introduces FedNMF, a federated learning framework for topic modeling on short texts that preserves privacy, and enhances performance with mutual information maximization, outperforming existing federated models.

Contribution

The paper proposes FedNMF+MI, a novel federated NMF approach that incorporates mutual information to improve topic modeling accuracy on heterogeneous short text data.

Findings

FedNMF+MI outperforms FedLDA and standard FedNMF in coherence and F1 score.

Mutual information maximization mitigates performance degradation in federated topic modeling.

The framework effectively preserves data privacy while maintaining high-quality topic extraction.

Abstract

Non-negative matrix factorization (NMF) based topic modeling is widely used in natural language processing (NLP) to uncover hidden topics of short text documents. Usually, training a high-quality topic model requires large amount of textual data. In many real-world scenarios, customer textual data should be private and sensitive, precluding uploading to data centers. This paper proposes a Federated NMF (FedNMF) framework, which allows multiple clients to collaboratively train a high-quality NMF based topic model with locally stored data. However, standard federated learning will significantly undermine the performance of topic models in downstream tasks (e.g., text classification) when the data distribution over clients is heterogeneous. To alleviate this issue, we further propose FedNMF+MI, which simultaneously maximizes the mutual information (MI) between the count features of local texts and their topic weight vectors to mitigate the performance degradation. Experimental results show that our FedNMF+MI methods outperform Federated Latent Dirichlet Allocation (FedLDA) and the FedNMF without MI methods for short texts by a significant margin on both coherence score and classification F1 score.