--- title: 'CSConv: Cognitive Stimulation Dialogue Dataset' url: https://www.emergentmind.com/topics/cognitive-stimulation-csconv type: topic --- # CSConv: Cognitive Stimulation Dialogue Dataset Cognitive Stimulation CSConv denotes a Chinese CS conversation dataset and the associated CSD framework introduced for dialogue with elders with cognitive impairment. The work addresses data sparsity in CS-based dialogue systems, particularly in the Chinese language, by constructing CSConv and by proposing a multi-source knowledge fusion method for CS dialogue that generates open-ended responses guided by the CS principle and emotional support strategy [2305.08200]. In the reported formulation, cognitive stimulation is coupled with emotional support rather than treated as a purely task-oriented prompting problem, and the system architecture explicitly predicts CS principle, emotion, and support strategy before response generation [2305.08200]. ## 1. Dataset origin and annotation protocol CSConv was constructed from approximately 2,600 one-to-one “elder ↔ therapist” dialogue groups [2305.08200]. The sources were Cognitive Stimulation Therapy videos in Cantonese translated to Mandarin, with a \(2:1\) ratio, and Mandarin dialogues manually authored from the CST handbook, which contributed 18 principles. The corpus was cleaned by truncating overly long turns and removing repeated punctuation, and it was formatted with manual speaker tags distinguishing SPEAKER, representing the elder, from LISTENER, representing the therapist or smart-speaker. Annotation proceeded by automated initialization of three labels per utterance via BERT classifiers, followed by two rounds of manual review [2305.08200]. The three label families were CS principle, emotion, and support strategy. This combination reflects the paper’s premise that “making chit chat while providing emotional support is overlooked by the majority of existing cognitive dialogue systems,” and the annotation scheme operationalizes that premise at utterance level. A plausible implication is that the dataset was designed not only for response generation but also for multi-task supervision, since the same utterances support classification of CS principle, emotion, and support strategy as prerequisites for controlled generation. ## 2. Label ontology and corpus statistics The CS principle labels comprise seven types: None, Inquiry, Respect, Reminiscence, Expression, Enjoyment, and Comfort [2305.08200]. The emotion labels comprise eight types: none, disgust, sadness, fear, surprise, like, happiness, and anger. The support strategy labels also comprise seven types: None, Question, Reflection of Feelings, Self-disclosure, Providing Suggestions, Information, and Others. The corpus statistics reported for CSConv are as follows: | Statistic | Value | |---|---:| | Conversations | 2,643 | | Total utterances | 16,845 | | SPEAKER turns | 8,363 | | LISTENER turns | 8,480 | | Avg. tokens per conv. | 60.39 | | Avg. utterances per conv. | 6.37 | | Avg. tokens per utt. | 9.48 | These statistics indicate a relatively short-turn conversational regime, with 6.37 utterances per conversation and 9.48 tokens per utterance [2305.08200]. This suggests that the dataset emphasizes compact, locally coherent exchanges rather than long-horizon dialogue management. Because the labels are attached at utterance level and the conversations are short, the benchmark is especially suited to studying fine-grained conditioning of response style and content under limited context. ## 3. Progressive-mask encoders and utterance-level classification The encoder component begins from a standard BERT encoder, specifically `bert-base-chinese`, and is enriched via two external-knowledge dictionaries: an emotion dictionary from Jiagu sentiment extraction and a keyword dictionary from Jiagu keyword extraction [2305.08200]. During pre-training, the model randomly masks spans of size 1–4 tokens or full sentences, mixing ratios \(\lambda_i\) for 1–4-token masks, a sentence-level mask ratio \(\lambda_5\), and 5% standard BERT-style `[MASK]`/random/unchanged tokens. At layer \(l\), the encoder is defined by the standard residual-attention and residual-FFN updates \[ \begin{aligned} h_i^{(l)} &= \text{LN}\bigl(x_i^{(l-1)} + \mathrm{MHAtt}(x_i^{(l-1)})\bigr)\\ x_i^{(l)} &= \text{LN}\bigl(h_i^{(l)} + \mathrm{FFN}(h_i^{(l)})\bigr)\,, \end{aligned} \] where \(\mathrm{MHAtt}\) is multi-head self-attention and \(\mathrm{FFN}\) is a two-layer feed-forward network [2305.08200]. After \(L\) layers, the contextualized token embeddings are denoted \(\mathbf{C_u}=[\tilde x_0,\ldots,\tilde x_m]\). The model flattens an \(M\)-turn context \(\mathcal U=[\mathrm{CLS},x_1,\dots,x_m]\) and feeds it into three parallel classifiers for CS principle, emotion, and strategy [2305.08200]. For CS principle, the paper gives \[ \ell_{cs} \;=\; \mathrm{CNN}_{\{2,3,4\}\times 256}\bigl(\mathbf{C_u}\bigr)\,\in\mathbb R^{7}, \] with analogous heads for the 8-way emotion task and the 7-way strategy task. Each classifier is trained with cross-entropy loss; for example, \[ \mathcal L_{cs} =\;-\frac1N\sum_{i=1}^N\sum_{k=1}^7 y^{(i)}_{cs,k}\,\log p_{cs,k}^{(i)}\,. \] The role of these encoders is not merely representational. The paper states that learning encoders as effective classifiers is the prerequisite to predicting the CS principle and emotional support strategy of the target response [2305.08200]. This places the classification subsystem upstream of generation and makes the predicted labels an explicit control signal. ## 4. Decoder conditioning and multi-source knowledge fusion At generation time, the decoder appends three special tokens corresponding to the predicted emotion, CS, and strategy labels, \((t_{emo},t_{cs},t_{str})\), to the end of the decoder input sequence: \[ \mathcal Y = [y_1,\dots,y_j,\;t_{emo},t_{cs},t_{str}]\,. \] Each token is embedded through word and positional embeddings \(\mathbf E_W(y_j)\) and \(\mathbf E_P(y_j)\in\mathbb R^d\) [2305.08200]. An extra encoder processes \(\mathcal Y\) to yield hidden states \(\{he_j\}\), after which the hidden states at the three label positions are replaced with the corresponding outputs from the three pre-trained classification encoders [2305.08200]. In the decoder’s cross-attention, these states attend to the conversation context. At decoder layer \(l\), the paper writes \[ he_j^{(l)} = \mathrm{LN}\bigl(y_j^{(l-1)} + \mathrm{MHAtt}(y_j^{(l-1)})\bigr)\,, \] with the label positions swapped as described above. The model further introduces two auxiliary MSE losses to bias generation toward high-intensity emotion words and keywords [2305.08200]. The emotion-intensity is derived from Chinese EmoBank, the keyword-intensity from Jiagu scores, and both are matched against self-attention weights \(a_j\) in decoder cross-attention: \[ \mathcal L_{emo}^{att} = \frac1{|\mathcal Y|}\sum_{j} \bigl(\eta_{emo}(y_j)-a_j\bigr)^2,\quad \mathcal L_{kw}^{att} = \frac1{|\mathcal Y|}\sum_{j} \bigl(\eta_{kw}(y_j)-a_j\bigr)^2. \] Response generation itself is formulated autoregressively: \[ \prod_{t=1}^n P(y_t\mid y_{