信息抽取的几个challenges

  • 模型层面:
    • 一个字符一个标签的假定无法处理重叠或嵌套块(例如中国教育部部长,部长)
    • 以词为基本单位的设定难以应用到无自然边界的语言(如汉语),而以字为基本单位有歧义
  • 数据层面:
    • 有监督学习依赖于高成本的人工标注数据
  • 优化层面:
    • 与分类任务不同,信息抽取任务的类别分布有显著差异:正例样本远远少于负例样本;错误和歧义通常集中于某几个特定的类别对之间

Papers read

TinyBERT: Distilling BERT for Natural Language Understanding

【EMNLP-2019】Patient Knowledge Distillation for BERT Model Compression

【NAACL-HLT-2019】BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

【2017NIPS】Attention Is All You Need

【ACL-2018】Universal Language Model Fine-tuning for Text Classification

【ACL-2019】Deep Unknown Intent Detection with Margin Loss

【ACL-2019】A Novel Bi-directional Interrelated Model for Joint Intent Detection and Slot Filling

【AAAI-2019】Unsupervised Transfer Learning for Spoken Language Understanding in Intelligent Agents

【EMNLP-2018】Zero-shot User Intent Detection via Capsule Neural Networks

【ACL-2019】Joint Slot Filling and Intent Detection via Capsule Neural Networks

【NAACL-HLT-2018】Slot-Gated Modeling for Joint Slot Filling and Intent Prediction

【CoRR-2017】Multi-Domain Adversarial Learning for Slot Filling in Spoken Language Understanding

【INTERSPEECH-2016】Multi-Domain Joint Semantic Frame Parsing using Bi-directional RNN-LSTM

【SIGDIAL-2013】Deep Neural Network Approach for the Dialog State Tracking Challenge

【EMNLP-15】A Model of Zero-Shot Learning of Spoken Language Understanding

【NAACL-18】A Bi-model based RNN Semantic Frame Parsing Model for Intent Detection and Slot Filling

【INTERSPEECH-16】Attention-Based Recurrent Neural Network Models for Joint Intent Detection and Slot Filling

【IJCAI-16】A Joint Model of Intent Determination and Slot Filling for Spoken Language Understanding