A Cognitive Stimulation Dialogue System with Multi-source Knowledge Fusion for Elders with Cognitive Impairment

Jiyue Jiang,Sheng Wang,Qintong Li,Lingpeng Kong,Chuan Wu

2023-05-15

Abstract:When communicating with elders with cognitive impairment, cognitive stimulation (CS) help to maintain the cognitive health of elders. Data sparsity is the main challenge in building CS-based dialogue systems, particularly in the Chinese language. To fill this gap, we construct a Chinese CS conversation (CSConv) dataset, which contains about 2.6K groups of dialogues with CS principles and emotional support strategy labels. Making chit chat while providing emotional support is overlooked by the majority of existing cognitive dialogue systems. In this paper, we propose a multi-source knowledge fusion method for CS dialogue (CSD), to generate open-ended responses guided by the CS principle and emotional support strategy. We first use a progressive mask method based on external knowledge to learn encoders as effective classifiers, which is the prerequisite to predict the CS principle and emotional support strategy of the target response. Then a decoder interacts with the perceived CS principle and emotional support strategy to generate responses. Extensive experiments conducted on the CSConv dataset demonstrate the effectiveness of the proposed method, while there is still a large space for improvement compared to human performance.

Computation and Language

What problem does this paper attempt to address?

The main focus of this paper is to design a Cognitive Stimulation Dialogue System (CSD) for older adults with cognitive impairments, aiming to maintain their cognitive health through conversation. The main challenge encountered in the study is data sparsity, especially in constructing a Chinese dialogue system. To address this issue, the authors constructed a Chinese Cognitive Stimulation Conversation Dataset called CSConv, which includes approximately 2600 dialogues following cognitive stimulation principles and emotion support strategies.

The paper proposes a multi-source knowledge fusion approach to generate open-ended responses that follow cognitive stimulation principles and emotion support strategies. Firstly, they use a progressive masking approach of external knowledge to train the encoder to predict the CS principles and emotion support strategies of the target response. Then, the decoder interacts with perceived CS principles and emotion support strategies to generate the response.

Experiments conducted on the CSConv dataset demonstrate the effectiveness of the proposed method, albeit with room for improvement compared to human performance. Additionally, they design a multi-source interaction mechanism to appropriately combine emotion support strategies and cognitive stimulation therapy to generate responses beneficial for psychological health.

In summary, the contributions of this paper include:
1. Construction of a Chinese Cognitive Stimulation Conversation Dataset (CSConv).
2. Proposing a progressive masking method to enhance the generalization ability of emotion knowledge and applicability in conversations with older adults.
3. Designing a multi-source interaction approach to simulate the interaction between the encoder, decoder, and external knowledge.
4. Conducting extensive experiments to validate the effectiveness of the CSD.