论文笔记A Neural Conversational Model

Abstract

1.Previous approaches often restrict to specific domains and require handcrafted rules.
2.Use the Sequence to Sequence frame work.
3.The strength of the model that it can be trained end-to-end and thus requires much fewer hand-crafted rules.
4. The model’s weakness is the lack of consistency is a common failure mode of our model.

Introduction

1. Neural network can be used to map complicated structures to other complicated structures.
2. Mapping a sequence to another sequence which has direct applications in NLU. (Sutskever et al., 2014)

The advantage of mapping a seq to another seq

4. requires little feature engineering and domain specificity, meanwhile matching the best answer of the input.
5. Allow researchers to work on tasks for domain knowledge, and too hard to design rules.
6. This approach can do surprisingly well on generating fluent and accurate replies to conversations.

The disadvantage of seq to seq

7. Due to the complexity of this mapping, conversational modeling has been designed to be very narrow in domain.

Datasets

8. IT helpdesk datasets of conversations
9. A noisy datasets of movie subtitles

Experiments

10. can hold a natural conversation and sometimes perform simple forms of common sense reasoning.
11. the recurrent nets obtain better performance compared to the n-gram model
12. capture important long-range correlations.

Related Work

14. seq2seq的应用
    （1）neural machine translation and achieves improvements on the English-French and English-German translation tasks from WMT’14 dataset (Luong et al.,2014; Jean et al., 2014)
    （2）parsing(Vinyals et al., 2014a)
    （3）image captioning (Vinyals et al., 2014b)

Difference from other conventional systems

15. lack domain knowledge