Coarse2Fine: Fine-grained Text Classification on Coarsely-grained Annotated Data

Abstract

Existing text classification methods mainly focus on a fixed label set, whereas many realworld applications require extending to new fine-grained classes as the number of samples per label increases. To accommodate such requirements, we introduce a new problem called coarse-to-fine grained classification, which aims to perform fine-grained classification on coarsely annotated data. Instead of asking for new fine-grained human annotations, we opt to leverage label surface names as the only human guidance and weave in rich pretrained generative language models into the iterative weak supervision strategy. Specifically, we first propose a label-conditioned finetuning formulation to attune these generators for our task. Furthermore, we devise a regularization objective based on the coarse-fine label constraints derived from our problem setting, giving us even further improvements over the prior formulation. Our framework uses the fine-tuned generative models to sample pseudo-training data for training the classifier, and bootstraps on real unlabeled data for model refinement. Extensive experiments and case studies on two real-world datasets demonstrate superior performance over SOTA zeroshot classification baselines.

Large-scale fine-grained semantic indexing of biomedical literature based on weakly-supervised deep learning

“…These results confirm that CO is a strong heuristic for the task of finegrained semantic indexing, outperforming again dictionary-based heuristics. In addition, the weak labelling enhancement approach (ALO 3 ) achieved an overall maF 1 about 5pp higher than CO alone, which is a statistically significant difference 35 . The performance of LM 3 was not significantly lower than that of ALO 3 , as the two approaches achieved practically the same F 1 score for most of the 70 labels.…”

“…In contrast to the broader problem of hierarchical text classification, coarsegrained class labels are already available in the biomedical indexing problem that we study. Recently, Mekala et al [35] focused on this special case of fine-grained indexing, where coarse-grained labels are available, proposing the Coarse2Fine (C2F ) method. C2F relies on the literal occurrence of fine-grained labels to create an initial weakly labeled dataset.…”

“…The second method is Coarse2Fine (C2F ) [35], which extends coarse-grained categories into fine-grained ones, as explained in section 2. The basic difference of C2F to the proposed method is that it does not use semantic information, as does the concept-occurrence heuristic.…”

“…34 Detailed results with scores per descriptor are available in the supplementary file S4. 35 Wilcoxon signed-rank test, H 1 : F 1 CO < F 1 ALO 3 , p-value = 0.000001. 36 Wilcoxon signed-rank test, H 1 : F 1 CO < F 1 eDBMn, p-value = 0.0065.…”

Intermediate Training on Question Answering Datasets Improves Generative Data Augmentation

“…Recent advances in NLP demonstrate exceptional capabilities of generative language models in generating text (Radford and Narasimhan, 2018;Radford et al, 2019;Lewis et al, 2020;Brown et al, 2020). Prior work (Kumar et al, 2020;Anaby-Tavor et al, 2020;Mekala et al, 2021)…”