arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1803.09074cs.CLcs.AIcs.NE
收藏

面向机器理解的多范围推理

Multi-range Reasoning for Machine Comprehension

Nanyang Technological University(南洋理工大学);Institute for Infocomm Research(信息通信研究院)

Yi Tay, Luu Anh Tuan, Siu Cheung Hui

摘要:

我们提出了MRU(多范围推理单元),一个新的快速组合编码器的机器理解(MC)。我们提出的MRU编码器的特点是多范围门控,执行一系列参数化的收缩和扩展层,用于学习从长期和短期依赖性中受益的门控向量。我们的方法的目标如下:(1)学习表示,同时知道长期和短期的上下文,(2)建模文档内块之间的关系和(3)快速和有效的序列编码。我们表明,我们提出的编码器作为一个独立的编码器,以及作为一个互补的积木展示了有前途的结果。我们在三个具有挑战性的MC数据集上进行了广泛的实验,即RACE,SearchQA和NarrativeQA,在所有数据集上都取得了极具竞争力的性能。在RACE基准测试中,我们的模型在不使用任何递归层或卷积层的情况下,性能优于DFN(动态融合网络)1.5%-6%。同样,我们在SearchQA基准测试中实现了相对于AMANDA的竞争性能,在NarrativeQA基准测试中实现了相对于Biocom的竞争性能,而没有使用任何LSTM/GRU层。最后,将MRU编码器与标准BiLSTM架构相结合,进一步提高了性能,实现了最先进的结果。

英文摘要:

We propose MRU (Multi-Range Reasoning Units), a new fast compositional encoder for machine comprehension (MC). Our proposed MRU encoders are characterized by multi-ranged gating, executing a series of parameterized contract-and-expand layers for learning gating vectors that benefit from long and short-term dependencies. The aims of our approach are as follows: (1) learning representations that are concurrently aware of long and short-term context, (2) modeling relationships between intra-document blocks and (3) fast and efficient sequence encoding. We show that our proposed encoder demonstrates promising results both as a standalone encoder and as well as a complementary building block. We conduct extensive experiments on three challenging MC datasets, namely RACE, SearchQA and NarrativeQA, achieving highly competitive performance on all. On the RACE benchmark, our model outperforms DFN (Dynamic Fusion Networks) by 1.5%-6% without using any recurrent or convolution layers. Similarly, we achieve competitive performance relative to AMANDA on the SearchQA benchmark and BiDAF on the NarrativeQA benchmark without using any LSTM/GRU layers. Finally, incorporating MRU encoders with standard BiLSTM architectures further improves performance, achieving state-of-the-art results.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑