--- title: 'Telling-Making-Enacting: Narrative Pipeline' url: https://www.emergentmind.com/topics/telling-making-enacting-tme type: topic --- # Telling-Making-Enacting: Narrative Pipeline The Telling–Making–Enacting (TME) paradigm constitutes a pipeline architecture for transforming deep story representations into dialogic conversational narratives. Within the context of computational storytelling, TME formally decomposes the narrative generation process into “Telling” (deep semantic encoding), “Making” (systematic content transformation), and “Enacting” (persona-driven surface realization). The architecture is instantiated in the M2D (“Monolog to Dialog”) system, which algorithmically converts monologic retellings into multi-speaker dialog while incorporating personality models and explicit dialogic variation [1708.07476]. ## 1. Deep Representation (“Telling”) The Telling phase composes the underlying semantic content of a story using two hierarchical structures: the Story Intention Graph (SIG) and Deep Syntactic Structures (DsyntS). - **Story Intention Graph (SIG):** Formally, the SIG is a directed graph $$ G = (V_E \,\cup\, V_C,\, E_T \,\cup\, E_C \,\cup\, E_A) $$ where - $V_E = \{e_1, ..., e_n\}$ are event nodes (atomic story actions), - $V_C = \{c_1, ..., c_m\}$ are character/prop nodes, - $E_T \subset V_E \times V_E$ encode temporal precedence ($e_i \to e_j$ if $e_i$ precedes $e_j$), - $E_C \subset V_E \times V_C \cup V_C \times V_E$ are participation edges (event-character/prop links), - $E_A \subset V_E \times \{\textit{goal}, \textit{plan}, \textit{affect}\} \times V_C$ annotate intentions and affects. - **Deep Syntactic Structures (DsyntS):** Each text sentence $s_k$ is represented as a tree $$ T_k = (N_k,\, \mathit{lab}_k,\, \mathit{rel}_k) $$ with node set $N_k$, node–label function $\mathit{lab}_k: N_k \to \{\texttt{noun}, \texttt{verb}, ...\}$, and labeled arcs $\mathit{rel}_k \subset N_k \times N_k \times \{\texttt{subject}, \texttt{object}, ...\}$. The coupling of the SIG (“fabula”) and DsyntS (“discourse skeleton”) comprises the narrative’s deep structure in the TME system [1708.07476]. ## 2. Systematic Conversion Pipeline (“Making”) The Making phase constitutes the algorithmic transformation from deep story structures into dialogic form. This phase is realized by the M2D system as a multi-stage process, summarized both in descriptive steps and as high-level pseudocode. - **Content Allocation:** Parameter $\alpha \in [0,1]$ determines the fraction of content assigned to Speaker 1, with $1-\alpha$ for Speaker 2. - **Aggregation/De-aggregation:** If a DsyntS $T_k$ exceeds a node threshold $L_{\max}$, split; if consecutive trees are both below $L_{\min}$ and share participant, merge. - **Pronominalization & Coreference:** Entities previously mentioned and now available for first/second-person realization are replaced with pronouns. - **Content Elaboration:** Question–answer (QA) pairs are generated by pruning—with probability $p_{QA}$, a selected node is replaced with a suitable wh-word, dropping its subtree to form the question, and the answer is the removed subtree. - **Pragmatic Marker Insertion:** Style-weight vector $w_i \in \mathbb{R}^F$ over $F$ features controls the insertion of dialogic markers (e.g., hedges, exclamations); probability for marker $f$ is $$ P_i(f) = \frac{\exp(w_{i,f})}{\sum_{f'} \exp(w_{i,f'})} $$ - **Lexical Choice:** Content words can be substituted by synonyms $y$ if $\mathit{freq}(y) \leq \beta \cdot \mathit{freq}(x)$ or to diversify. - **Morphosyntactic Post-processing:** Grammatical contractions, possessives, and final punctuation are processed before realization. - **Surface Realization:** RealPro renders DsyntS trees into grammatical English dialog turns. Table: Core Pipeline Stages and Parameters | Stage | Key Parameter(s) | Operation Summary | |----------------------------|------------------|-----------------------------------------| | Content Allocation | $\alpha$ | Split turns between speakers | | Aggregation/De-aggregation | $L_{\max}, L_{\min}$ | Merge/split DsyntS trees | | Pronominalization | -- | Replace with pronoun when contextually licensed | | Pragmatic Markers | $w_i$ | Insert dialogic feature by softmax | | Lexical Choice | $\beta$, $p_{swap}$ | Synonym replacement | A high-level pseudocode implementation of the full pipeline is included as an explicit procedure (see original source for details) [1708.07476]. ## 3. Persona and Personality Control (“Enacting”) Enacting operationalizes speaker-level variation in dialogic realization by controlling personality traits along a reduced Big-Five axis. - **Trait Vector:** Each speaker $i$ is given a trait $t_i \in \{−1,+1\}$, with $t_i=+1$ (extravert) and $t_i=−1$ (introvert). - **Style-Weight Vector:** For dialog-feature set $F = \{f_1, ..., f_m\}$, two prototypes ($w^{(+)}$ for extraverts, $w^{(−)}$ for introverts) are defined, with $$ w_i = \frac{1 + t_i}{2} w^{(+)} + \frac{1 - t_i}{2} w^{(−)} $$ The probability of feature $f_j$ in speaker $i$’s turn is then: $$ P_i(f_j) = \mathrm{softmax}_j(w_{i, j}) = \frac{\exp(w_{i, j})}{\sum_{k=1}^m \exp(w_{i, k})} $$ - **Content Allocation Bias:** Extraverts are assigned more dialog content: $$ N_1 = \mathrm{round}\left( \frac{1 + t_1}{2} \cdot N \right),\quad N_2 = N - N_1 $$ Personality instantiation directly modulates dialogic markers, degree of content allocation, and dialogic act selection in realization. ## 4. Key Mathematical Formulations - **Dialogue-Act Selection:** For candidate dialog act $a$ in context $c$, score for speaker $i$ is $$ S_i(a \mid c) = \theta^\top \phi(c,a) + \gamma\, t_i\, a_{\mathrm{extro}} $$ where $\phi(c,a)$ encodes local context and act, $a_{\mathrm{extro}}$ indicates interactivity, and $t_i$ biases extraverted behavior. Select $$ a^* = \arg\max_a S_i(a \mid c) $$ - **Language Model Adaptation:** Basic language model $$ P(w_k \mid w_{