Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Margin Prototypical Network for Few-shot Relation Classification with Fine-grained Features

Miao Fan, Yeqi Bai, Mingming Sun, Ping Li

TL;DR

This paper considers few-shot learning is of great practical significance to RC and thus improves a modern framework of metric learning for few- shot RC, and adopts the large-margin ProtoNet with fine-grained features, expecting they can generalize well on long-tail relations.

Abstract

Relation classification (RC) plays a pivotal role in both natural language understanding and knowledge graph completion. It is generally formulated as a task to recognize the relationship between two entities of interest appearing in a free-text sentence. Conventional approaches on RC, regardless of feature engineering or deep learning based, can obtain promising performance on categorizing common types of relation leaving a large proportion of unrecognizable long-tail relations due to insufficient labeled instances for training. In this paper, we consider few-shot learning is of great practical significance to RC and thus improve a modern framework of metric learning for few-shot RC. Specifically, we adopt the large-margin ProtoNet with fine-grained features, expecting they can generalize well on long-tail relations. Extensive experiments were conducted by FewRel, a large-scale supervised few-shot RC dataset, to evaluate our framework: LM-ProtoNet (FGF). The results demonstrate that it can achieve substantial improvements over many baseline approaches.

Large Margin Prototypical Network for Few-shot Relation Classification with Fine-grained Features

TL;DR

This paper considers few-shot learning is of great practical significance to RC and thus improves a modern framework of metric learning for few- shot RC, and adopts the large-margin ProtoNet with fine-grained features, expecting they can generalize well on long-tail relations.

Abstract

Relation classification (RC) plays a pivotal role in both natural language understanding and knowledge graph completion. It is generally formulated as a task to recognize the relationship between two entities of interest appearing in a free-text sentence. Conventional approaches on RC, regardless of feature engineering or deep learning based, can obtain promising performance on categorizing common types of relation leaving a large proportion of unrecognizable long-tail relations due to insufficient labeled instances for training. In this paper, we consider few-shot learning is of great practical significance to RC and thus improve a modern framework of metric learning for few-shot RC. Specifically, we adopt the large-margin ProtoNet with fine-grained features, expecting they can generalize well on long-tail relations. Extensive experiments were conducted by FewRel, a large-scale supervised few-shot RC dataset, to evaluate our framework: LM-ProtoNet (FGF). The results demonstrate that it can achieve substantial improvements over many baseline approaches.
Paper Structure (12 sections, 5 equations, 2 figures, 3 tables)

This paper contains 12 sections, 5 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Proposed Framework
  3. Fine-grained Features
  4. Triplet Loss for Large Margin ProtoNet
  5. Experiments
  6. Benchmark Dataset
  7. Experimental Settings
  8. Result Analysis
  9. Ablation Studies
  10. Fine-grained Features v.s. CNN-based Embedding
  11. Triplet Loss v.s. Softmax Cross-entropy
  12. Conclusion

Figures (2)

  • Figure 1: The framework of LM-ProtoNet (FGF) which is composed of two modules: fine-grained features for instance embedding (on the left) and triplet loss for ProtoNet (on the right). In this case, LM-ProtoNet is addressing a $3$-way (classes: $A$, $B$, and $C$) -$3$-shot RC, and $A_c$ is the center of class $A$.
  • Figure 2: a $7$-way-$40$-shot scenario of RC where the embeddings of the instances in the support set are acquired by ProtoNet on the top and LM-ProtoNet (FGF) at the bottom. The embeddings are mapped into the same 2D metric space by the technique of t-SNE.