Recommended 【論文紹介】Approximate Bayesian Image Interpretation Using Generative Probabilisti...
[DL輪読会]Relational inductive biases, deep learning, and graph networks
[DL輪読会]Learning to Simulate Complex Physics with Graph Networks
日本ソフトウェア科学会第36回大会発表資料「帰納的プログラミングの初等教育の試み」西澤勇輝
[DL輪読会]Learning to Generalize: Meta-Learning for Domain Generalization
【文献紹介】Multi-modal Summarization for Asynchronous Collection of Text, Image, A...
【論文紹介】How Powerful are Graph Neural Networks?
Neural networks for Graph Data NeurIPS2018読み会@PFN
【DL輪読会】Data-Efficient Reinforcement Learning with Self-Predictive Representat...
[DL輪読会]Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs (CV...
[DL輪読会]Graph R-CNN for Scene Graph Generation
[DL Hacks]Model-Agnostic Meta-Learning for Fast Adaptation of Deep Network
A Novel Cascade Binary Tagging Framework for Relational Triple Extraction
【輪読会】Braxlines: Fast and Interactive Toolkit for RL-driven Behavior Engineeri...
[DL輪読会]Dream to Control: Learning Behaviors by Latent Imagination
2020 0906 acl_2020_reading_shared
[DL輪読会]Life-Long Disentangled Representation Learning with Cross-Domain Laten...
Incorporating syntactic and semantic information in word embeddings using gra...
Can language models be biomedical knowledge bases
A semantics aware-transformer_model_of_relation_linking_for_knowledge_base_qu...
More Related Content 【論文紹介】Approximate Bayesian Image Interpretation Using Generative Probabilisti...
[DL輪読会]Relational inductive biases, deep learning, and graph networks
[DL輪読会]Learning to Simulate Complex Physics with Graph Networks
日本ソフトウェア科学会第36回大会発表資料「帰納的プログラミングの初等教育の試み」西澤勇輝
[DL輪読会]Learning to Generalize: Meta-Learning for Domain Generalization
【文献紹介】Multi-modal Summarization for Asynchronous Collection of Text, Image, A...
【論文紹介】How Powerful are Graph Neural Networks?
What's hot (20)
Neural networks for Graph Data NeurIPS2018読み会@PFN
【DL輪読会】Data-Efficient Reinforcement Learning with Self-Predictive Representat...
[DL輪読会]Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs (CV...
[DL輪読会]Graph R-CNN for Scene Graph Generation
[DL Hacks]Model-Agnostic Meta-Learning for Fast Adaptation of Deep Network
A Novel Cascade Binary Tagging Framework for Relational Triple Extraction
【輪読会】Braxlines: Fast and Interactive Toolkit for RL-driven Behavior Engineeri...
[DL輪読会]Dream to Control: Learning Behaviors by Latent Imagination
2020 0906 acl_2020_reading_shared
[DL輪読会]Life-Long Disentangled Representation Learning with Cross-Domain Laten...
Incorporating syntactic and semantic information in word embeddings using gra...
More from Ace12358 (18) Can language models be biomedical knowledge bases
A semantics aware-transformer_model_of_relation_linking_for_knowledge_base_qu...
A semantics aware-transformer_model_of_relation_linking_for_knowledge_base_qu...
Camouflaged chinese spam_content_detection_with_semi-supervised_generative_ac...
The contribution of_stylistic_information_to_content-based_mobile_spam_filtering
Dual tkb a-dual_learning_bridge_between_text_and_knowledge_base
Knowledge_graph_alignment_with_entity-pair_embedding
Low-dimensional hyperbolic knowledge graph embeddings
Neural relation extraction for knowledge base enrichment introduced by Yoshia...
Emnlp2015 reading festival_lstm_cws
Naacl2015unsupervised morph
DeepLearning論文紹介@Ace12358
1. I Know What You Asked:
Graph Path Learning using
AMR for Commonsense
Reasoning
Jungwoo Lim∗, Dongsuk Oh∗, Yoonna Jang, Kisu Yang, Heuiseok Lim
COLING 2020 , @論文読み会, 紹介者: Yoshiaki Kitagawa
2. Summary
Task
CommonsenseQA
事前定義された知識を使用した常識的な
推論によって正解を予測するタスク
先行研究と比べてどこがすごいか？
常識的な推論のプロセスをモデルにうま
く組み込んでいるところ
技術や手法のキモ
Graph Integrating and Pruning
AMR と CN グラフの統合と枝刈り
Language Encoder and Graph Path
Learning Module
QA sentence のエンコーダと path 学習の
枠組み
どうやって有効だと検証したか？
選択式のQAデータセットでの Accuracy を比較
議論はあるか？
難しい選択肢: 正解と同じ relation を持っているもの
は間違えてしまう
次に読むべき論文
Abstract Meaning Representation (AMR) (Banarescu et
al., 2013)
5. Graph Integrating and Pruning (1/2)
AMR を拡張して、ConceptNet のサブグラフを付与
ARG0, ARG1と繋がっていない node を削除 (a) -> (b)
7. Language Encoder and Graph Path Learning
Module
indicates the shortest path of the relation between two nodes.
between concepts i and j is the concatenation of the final hidden states from the forward and backward GRU networks
attention score considering the concepts and their relations
12. Summary
Task
CommonsenseQA
事前定義された知識を使用した常識的な
推論によって正解を予測するタスク
先行研究と比べてどこがすごいか？
常識的な推論のプロセスをモデルにうま
く組み込んでいるところ
技術や手法のキモ
Graph Integrating and Pruning
AMR と CN グラフの統合と枝刈り
Language Encoder and Graph Path
Learning Module
QA sentence のエンコーダと path 学習の
枠組み
どうやって有効だと検証したか？
選択式のQAデータセットでの Accuracy を比較
議論はあるか？
難しい選択肢: 正解と同じ relation を持っているもの
は間違えてしまう
次に読むべき論文
Abstract Meaning Representation (AMR) (Banarescu et
al., 2013)
