Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TransferNet: An Effective and Transparent Framework for Multi-hop Question Answering over Relation Graph

Jiaxin Shi, Shulin Cao, Lei Hou, Juanzi Li, Hanwang Zhang

TL;DR

TransferNet addresses multi-hop QA over relation graphs by learning to transfer entity scores along attention-weighted relation scores across hops, applicable to both label-form (knowledge graphs) and text-form (textual relations). It maintains an entity score vector $\mathbf{a}^t$ and a relation-score matrix $\mathbf{W}^t$ at each step, with question-aware attention producing $q^t$ and a differentiable final aggregation yielding $a^*$. The approach achieves state-of-the-art results on MetaQA (including 100% accuracy for 2- and 3-hop questions) and strong gains on WebQSP and CompWebQ, with substantial improvements in text-form QA as well, while providing interpretable intermediate reasoning paths. Its general, transparent design supports mixed forms of relations and facilitates visualization of the decision process, promising broad applicability to real-world QA over heterogeneous relation graphs.

Abstract

Multi-hop Question Answering (QA) is a challenging task because it requires precise reasoning with entity relations at every step towards the answer. The relations can be represented in terms of labels in knowledge graph (e.g., \textit{spouse}) or text in text corpus (e.g., \textit{they have been married for 26 years}). Existing models usually infer the answer by predicting the sequential relation path or aggregating the hidden graph features. The former is hard to optimize, and the latter lacks interpretability. In this paper, we propose TransferNet, an effective and transparent model for multi-hop QA, which supports both label and text relations in a unified framework. TransferNet jumps across entities at multiple steps. At each step, it attends to different parts of the question, computes activated scores for relations, and then transfer the previous entity scores along activated relations in a differentiable way. We carry out extensive experiments on three datasets and demonstrate that TransferNet surpasses the state-of-the-art models by a large margin. In particular, on MetaQA, it achieves 100\% accuracy in 2-hop and 3-hop questions. By qualitative analysis, we show that TransferNet has transparent and interpretable intermediate results.

TransferNet: An Effective and Transparent Framework for Multi-hop Question Answering over Relation Graph

TL;DR

TransferNet addresses multi-hop QA over relation graphs by learning to transfer entity scores along attention-weighted relation scores across hops, applicable to both label-form (knowledge graphs) and text-form (textual relations). It maintains an entity score vector and a relation-score matrix at each step, with question-aware attention producing and a differentiable final aggregation yielding . The approach achieves state-of-the-art results on MetaQA (including 100% accuracy for 2- and 3-hop questions) and strong gains on WebQSP and CompWebQ, with substantial improvements in text-form QA as well, while providing interpretable intermediate reasoning paths. Its general, transparent design supports mixed forms of relations and facilitates visualization of the decision process, promising broad applicability to real-world QA over heterogeneous relation graphs.

Abstract

Multi-hop Question Answering (QA) is a challenging task because it requires precise reasoning with entity relations at every step towards the answer. The relations can be represented in terms of labels in knowledge graph (e.g., \textit{spouse}) or text in text corpus (e.g., \textit{they have been married for 26 years}). Existing models usually infer the answer by predicting the sequential relation path or aggregating the hidden graph features. The former is hard to optimize, and the latter lacks interpretability. In this paper, we propose TransferNet, an effective and transparent model for multi-hop QA, which supports both label and text relations in a unified framework. TransferNet jumps across entities at multiple steps. At each step, it attends to different parts of the question, computes activated scores for relations, and then transfer the previous entity scores along activated relations in a differentiable way. We carry out extensive experiments on three datasets and demonstrate that TransferNet surpasses the state-of-the-art models by a large margin. In particular, on MetaQA, it achieves 100\% accuracy in 2-hop and 3-hop questions. By qualitative analysis, we show that TransferNet has transparent and interpretable intermediate results.

Paper Structure

This paper contains 21 sections, 13 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Preliminary
  5. TransferNet
  6. Training
  7. Additional Modules
  8. Relation Score Computation
  9. Label Form
  10. Text Form
  11. Experiments
  12. Datasets
  13. Baselines
  14. Implementations
  15. Results
  16. ...and 6 more sections

Figures (4)

  • Figure 1: Answering a multi-hop question over the relation graph. The relations are constrained predicates in the label form (i.e., knowledge graph) while free texts in the text form. The reasoning process has been marked in the graph, where the correspondence between relations and question words has been highlighted in the same color.
  • Figure 2: The framework of TransferNet (top) and example of reasoning process (bottom).
  • Figure 3: Reasoning process of 3-hop questions. The top is in label form, where the suffix "_rev" means reverse relation. The bottom is in text form, where "mask" in blue means the language mask. We show the relation scores in purple and highlight the activated entities and relations (score > 0.8) and words (score > 0.05) in red.
  • Figure 4: Comparison of data efficiency (left) and convergency speed (right) on label-formed MetaQA.