Emotion Support Conversation Emotion Support Conversation

insert image description here

This blog mainly analyzes some of the latest situation of relevant papers under the scene of emotional support, that is, psychological counseling, the links of the papers and the main ideas. The motivation, scheme, and experimental results of each article will be introduced in detail later.

1、Hide:

Towards Emotional Support Dialog Systems

A raw dataset of emotionally underpinned conversations.

Detailed blog post reading

2、MISC:

MISC: A MIxed Strategy-Aware Model Integrating COMET for Emotional Support Conversation

The main idea of ​​the article is 1) use trial knowledge to model the user's emotional state and mine relevant information; 2) use mixed strategies to make better strategy predictions and assist emotional support responses.

Detailed blog post reading

3、GLHG:

Control Globally, Understand Locally: A Global-to-Local Hierarchical Graph Network for Emotional Support Conversation

The main idea of ​​the article is derived from the basic principle of emotional support, that is, it is hoped that the dialogue system can perform psychological consultation like humans. Specifically, 1) the system needs to explore the causes of help-seekers’ emotional problems (global); 2) analyze their implicit psychological intentions according to the current dialogue (local) to guide the provision of supportive responses.

Detailed blog post reading

4、MultiESC:

Improving Multi-turn Emotional Support Dialogue Generation with Lookahead Strategy Planning

The main idea of ​​the article is 1) a more reasonable strategic arrangement can effectively provide emotional support, and understanding the user's status can also help to better respond; 2) and how to dynamically model the user's status is also the subject of this article A core research point.

Detailed blog post reading

5、FADO:

FADO: Feedback-Aware Double COntrolling Network for Emotional Support Conversation

The main idea of ​​the article is 1) The user's feedback information is a more important factor, because it can reflect the user's status of the current dialogue, such as satisfaction or dissatisfaction; 2) The current dialogue system only considers the context to Policy flow, ignoring the process from policy to context flow.

Detailed blog post reading

other

For more interesting MRC articles, see: Machine Reading Comprehension Using Reverse Thinking.
Related Literature
Bi-directional Cognitive Thinking Network for Machine Reading Comprehension Paper Reading
Evidence Reasoning Network.
Hybrid Curriculum Learning for Emotion Recognition in Conversation
BERT for text classification methods

Guess you like

Origin blog.csdn.net/ganxiwu9686/article/details/127678201