Blurb-Refined Inference from Crowdsourced Book Reviews using Hierarchical Genre Mining with Dual-Path Graph Convolutions

TL;DR

HiGeMine tackles hierarchical, multi-label book genre classification under noisy, crowd-sourced reviews by jointly leveraging authoritative blurbs and a zero-shot filtered set of reviews. The approach uses a two-stage, coarse-to-fine pipeline where Level-1 fiction/non-fiction is determined via dual-path GCNs over blurb- and review-derived tokens, then Level-2 classifiers predict fine-grained genres while modeling inter-genre dependencies through a label co-occurrence graph. A dedicated vocabulary, genre-aware word embeddings, and adaptive fusion weights enhance robustness to noise, with a new hierarchical dataset enabling targeted evaluation. Empirical results show HiGeMine outperforms strong baselines across hierarchical classification, language models, and both open- and closed-source LLMs, highlighting the value of principled content filtering and structured representation learning for real-world, noisy text data.

Abstract

Accurate book genre classification is fundamental to digital library organization, content discovery, and personalized recommendation. Existing approaches typically model genre prediction as a flat, single-label task, ignoring hierarchical genre structure and relying heavily on noisy, subjective user reviews, which often degrade classification reliability. We propose HiGeMine, a two-phase hierarchical genre mining framework that robustly integrates user reviews with authoritative book blurbs. In the first phase, HiGeMine employs a zero-shot semantic alignment strategy to filter reviews, retaining only those semantically consistent with the corresponding blurb, thereby mitigating noise, bias, and irrelevance. In the second phase, we introduce a dual-path, two-level graph-based classification architecture: a coarse-grained Level-1 binary classifier distinguishes fiction from non-fiction, followed by Level-2 multi-label classifiers for fine-grained genre prediction. Inter-genre dependencies are explicitly modeled using a label co-occurrence graph, while contextual representations are derived from pretrained language models applied to the filtered textual content. To facilitate systematic evaluation, we curate a new hierarchical book genre dataset. Extensive experiments demonstrate that HiGeMine consistently outperformed strong baselines across hierarchical genre classification tasks. The proposed framework offers a principled and effective solution for leveraging both structured and unstructured textual data in hierarchical book genre analysis.

Figures (9)

• Figure 1: Overall framework of HiGeMine
• Figure 2: Review filtering and vocabulary generation
• Figure 3: Detailed framework of HiGeMine. Best viewed in color.
• Figure : (a) Past book genre classifiers
• Figure : (a) Past book genre classifiers