workflow somflow-guided-cluster-exploration template: loop(build-model -> define-unit (assign cells) -> contextualise (grid layout) -> characterise (cell measures) -> visualise (SOM view) -> characterise (annotate) -> abstract (interpret) -> define-unit (record run) -> visualise (provenance flow) -> abstract (reflect) -> assess -> generate-knowledge (strategy) -> define-unit (partition/retain)) -> generate-knowledge (synthesise) description: "Assumed composite workflow (synthesised from the exploratory analysis usage scenario) for guided iterative cluster exploration with self-organizing maps; a map is trained on the current data subset, cells are analysed for quality, interestingness, and partitioning recommendations, the analyst adapts computations or partitions the data while retaining every analysis step in a branching analytic-provenance graph that is continuously visualised and used for reflection and comparison across the exploration" # ============================================================================ # INPUT # ============================================================================ artifact D_utterances : entities origin: given internal structure: elementary embedment: set features: - id: f_pitch_contour value structure: vector value type: numeric description: "Time series values representing the item's temporal profile (e.g., pitch contour over time)" - id: f_metadata value structure: vector value type: categorical description: "Associated categorical metadata attributes (e.g., speaker characteristics, utterance type, repetition)" description: "Collection of time series items with associated categorical metadata, forming the dataset to be explored; also serves as the current working subset for the drill-down loop" artifact S_analysis_strategy : specification origin: given representation form: "configuration schema" description: "Initial choices for data preprocessing, distance metric, self-organizing map parameters, and criteria for partitioning the current data subset" artifact D_som_runs : entities origin: given internal structure: elementary embedment: relational structure description: "Initially empty collection of retained SOM analysis runs, to be populated with the initial SOM and all subsequently created, retrained, filtered, or partitioned SOMs; a cluster may correspond to a single cell or a set of similar neighbouring cells" artifact F_run_lineage : feature(D_som_runs) origin: given value structure: relational configuration representation form: "directed parent-child analysis flow graph" description: "Initially empty parent-child relations between SOM runs, indicating retraining, filtering, splitting, or selection-based partitioning operations" # ============================================================================ # DRILL-DOWN EXPLORATION LOOP # ============================================================================ loop L_drilldown: purpose: "Iteratively train, analyse, and refine self-organizing maps on successive data subsets, retaining every analysis step in a branching provenance graph, adapting computations or partitioning data to progressively explore clusters and their relations to metadata" until: "Clusters and metadata relations are sufficiently understood across the retained analysis graph, and no further adaptation or partitioning is warranted" body: # ======================================================================== # Analyze: train map and compute cell-level views # ======================================================================== transform T_train_som : intent: build-model manner: "train self-organizing map providing joint vector quantization and topology-preserving two-dimensional embedding" input: D_utterances, S_analysis_strategy output: M_som actor: machine description: "Train a self-organizing map on the current data subset according to the current preprocessing, metric, and parameter configuration" artifact M_som : model(D_utterances, S_analysis_strategy) model type: "clustering and dimensionality-reduction model" representation form: "grid of prototype vectors with topological neighborhood structure (self-organizing map)" description: "Trained model partitioning the current data subset into topologically ordered prototype cells" transform T_assign_cells : intent: define-unit manner: "assign items to best-matching grid cells" input: D_utterances, M_som output: D_cells, F_cell_assignment actor: machine description: "Define grid cells as analysis units and assign each item to its best-matching cell" artifact D_cells : entities internal structure: group/cluster embedment: relational structure features: - id: f_cell_size value structure: atomic value type: numeric description: "Number of items assigned to the cell" description: "Grid cells of the self-organizing map, each an unordered group of similar items; cells are related through the map's neighborhood topology" artifact F_cell_assignment : feature(D_utterances) value structure: atomic value type: categorical description: "Best-matching cell identifier for each item" transform T_grid_layout : intent: contextualise manner: "position cells according to the map's topology-preserving embedding" input: D_cells, M_som output: D_grid_space, A_cell_layout actor: machine description: "Place grid cells within a two-dimensional reference space derived from the trained map's topology" artifact D_grid_space : entities internal structure: elementary features: - id: f_dimensions value structure: atomic value type: numeric description: "Number of spatial dimensions (2)" description: "Two-dimensional grid space serving as reference frame for cell positions" artifact A_cell_layout : arrangement(D_cells) context: D_grid_space principle: "topology-preserving grid placement from self-organizing map training" description: "Positions of grid cells reflecting their similarity-based neighborhood topology" transform T_compute_cell_measures : intent: characterise manner: "compute quality, interestingness, similarity, and partitioning recommendations" input: D_cells, F_cell_assignment, D_utterances, M_som output: F_cell_quality, F_cell_interestingness, F_partition_recommendations actor: machine description: "Compute per-cell quantization error from assigned items and SOM prototypes, metadata-interestingness scores, similar-neighbour candidates, and derived refinement and partitioning action recommendations" artifact F_cell_quality : feature(D_cells) value structure: atomic value type: numeric description: "Quantization error per cell, indicating how precisely the prototype represents its assigned items" artifact F_cell_interestingness : feature(D_cells) value structure: vector value type: numeric description: "Per-metadata-attribute interestingness scores per cell, indicating potential relations between cell content and metadata" artifact F_partition_recommendations : feature(D_cells) value structure: vector value type: {numeric, categorical} description: "Per-cell and map-level recommendation scores and action categories for refinement, similar-cell selection, metadata splits, and cluster-based splits" transform T_visualise_som : intent: visualise manner: "SOM grid view with prototype and item time series, topology overlays, metadata colouring, and ranked recommendation cues" input: A_cell_layout, D_cells, D_utterances, M_som, F_cell_assignment, F_cell_quality, F_cell_interestingness, F_partition_recommendations output: V_som actor: machine description: "Render SOM cells with prototype and assigned-item time series, topology and quality overlays, metadata colouring, and ranked recommendations for further analysis" artifact V_som : visualisation(A_cell_layout, D_cells, D_utterances, M_som, F_cell_assignment, F_cell_quality, F_cell_interestingness, F_partition_recommendations) layout: "two-dimensional SOM grid" form: "cell glyphs with prototype and item line charts, quality or metadata overlays, and recommendation thumbnails" encoding: "cell position from A_cell_layout; prototype line and training history from M_som; assigned item traces and bandwidth from D_utterances and F_cell_assignment; cell shading from F_cell_quality or metadata colouring; density from f_cell_size; ranked thumbnails from F_cell_interestingness; partitioning cues from F_partition_recommendations" description: "Interactive SOM view supporting inspection of cluster structure, prototype quality, metadata relations, and recommended next actions" # ======================================================================== # Adapt (data enrichment): manual annotation # ======================================================================== transform T_annotate_cells : intent: characterise manner: "interactive assignment of user-defined metadata labels" input: D_cells, F_cell_assignment, V_som output: F_user_annotations actor: human description: "Assign analyst-defined labels, such as interesting, uninteresting, or measurement-error, to cells or their member items" artifact F_user_annotations : feature(D_utterances) value structure: atomic value type: categorical description: "User-defined item annotations propagated from selected SOM cells, available for later filtering and recommendation computation" transform T_interpret : intent: abstract manner: "perception and interpretation of cell topology, quality overlays, metadata colouring, and user annotations" input: V_som, F_user_annotations output: P_patterns actor: human description: "Identify salient regions, candidate groupings, and apparent relations between map structure and metadata attributes" artifact P_patterns : pattern(D_cells, F_cell_interestingness) representation form: "categorized patterns with descriptions" description: "Identified salient map regions, candidate groupings, and apparent relations between cluster structure and metadata attributes" # ======================================================================== # Reflect: retain the current run and compare against the provenance graph # ======================================================================== transform T_record_analysis_step : intent: define-unit manner: "retain current SOM result as a provenance-graph node" input: D_som_runs, F_run_lineage, M_som, D_utterances, S_analysis_strategy output: D_som_runs_r, F_run_lineage_r actor: machine description: "Add the current SOM analysis result and its relation to predecessor runs to the retained analytic-provenance graph" artifact D_som_runs_r : entities internal structure: elementary embedment: relational structure description: "Updated collection of retained SOM analysis runs" artifact F_run_lineage_r : feature(D_som_runs_r) value structure: relational configuration representation form: "directed parent-child analysis flow graph" description: "Updated provenance relations between retained SOM runs" assign: D_som_runs := D_som_runs_r F_run_lineage := F_run_lineage_r transform T_visualise_flow : intent: visualise manner: "analytic-provenance flow graph with embedded SOM views" input: D_som_runs, F_run_lineage output: V_som_flow actor: machine description: "Display retained SOM analyses as graph nodes connected by data-flow and partitioning relationships, supporting navigation, comparison, annotation, and provenance inspection" artifact V_som_flow : visualisation(D_som_runs, F_run_lineage) layout: "directed flow graph" form: "graph nodes containing SOM thumbnails and directed links" encoding: "nodes represent retained SOM runs; links represent data-flow relations from F_run_lineage; link width represents the number of transferred items; linked colour overlays and annotations support comparison across runs" description: "Analytic-provenance graph showing the evolution and branching structure of SOM-based cluster exploration" transform T_reflect : intent: abstract manner: "compare retained SOM results and analysis branches via linked overlays and annotations" input: V_som_flow, D_som_runs, F_run_lineage, P_patterns output: P_overall_relations actor: human description: "Compare the current and earlier SOM analyses, branches, and data partitions to identify verified relations between cluster structure and metadata, informing ongoing exploration" artifact P_overall_relations : pattern(D_som_runs, P_patterns) representation form: "categorized patterns with descriptions" description: "Relations between cluster structure and metadata attributes verified by comparison across the retained analysis graph" # ======================================================================== # Assess: decide whether to adapt, partition, or stop # ======================================================================== transform T_assess : intent: assess manner: "judge whether current and accumulated analyses adequately reveal relevant structure, guided by quality, interestingness, and reflection" input: V_som, P_patterns, F_cell_quality, F_cell_interestingness, P_overall_relations output: K_assessment actor: human description: "Evaluate whether the current and accumulated SOM analyses adequately represent the data and relations of interest, or whether further adaptation or partitioning is warranted" artifact K_assessment : knowledge(D_cells, M_som, D_som_runs) representation form: "quality judgment" description: "Assessment of map, partition, and accumulated analysis adequacy, and whether further adaptation or partitioning is needed" if K_assessment indicates further analysis needed: then: # ==================================================================== # Adapt: decide computation or partitioning strategy # ==================================================================== transform T_decide_strategy : intent: generate-knowledge manner: "decide whether to adapt data preprocessing, metric, or parameters, or to define partitioning criteria, guided by identified patterns and partitioning recommendations" input: K_assessment, P_patterns, F_partition_recommendations, S_analysis_strategy output: S_analysis_strategy_r actor: human description: "Formulate updated analysis strategy: adjust preprocessing, metric, or self-organizing map parameters, and/or specify criteria for the next data partition, optionally guided by machine-derived recommendations" artifact S_analysis_strategy_r : specification representation form: "configuration schema" description: "Updated configuration of preprocessing, metric, parameters, and partition criteria for the next analysis step" assign: S_analysis_strategy := S_analysis_strategy_r # ==================================================================== # Partition: define the next data subset # ==================================================================== transform T_define_next : intent: define-unit manner: "retain the current subset for recomputation, or partition it into a subset of interest based on cell, cluster, metadata, or annotation criteria" input: D_utterances, D_cells, F_cell_assignment, F_user_annotations, S_analysis_strategy output: D_next actor: hybrid description: "Define the data subset to analyse in the next step, either by retaining the current subset under revised preprocessing or by selecting a subset based on cell, cluster, metadata, or noise-annotation criteria" artifact D_next : entities internal structure: elementary embedment: set features: - id: f_pitch_contour value structure: vector value type: numeric description: "Time series profile retained for each selected item" - id: f_metadata value structure: vector value type: categorical description: "Associated metadata retained for each selected item" description: "Selected or retained subset of the current data, preserving time series and metadata for the next SOM analysis" assign: D_utterances := D_next else: exit loop L_drilldown end loop L_drilldown # ============================================================================ # SYNTHESISE FINDINGS # ============================================================================ transform T_synthesise : intent: generate-knowledge manner: "formulate statements summarising discovered groupings and metadata relations across retained analysis branches" input: P_overall_relations, K_assessment, D_som_runs, F_run_lineage output: K_findings actor: human description: "Synthesise findings about the data's cluster structure, its relations to metadata attributes, and data quality issues uncovered across the retained SOM analysis graph" artifact K_findings : knowledge(P_overall_relations, K_assessment, D_som_runs, F_run_lineage) representation form: "statements and explanations" description: "Documented findings about discovered clusters, their relations to metadata attributes, and data quality issues identified through iterative, retained SOM-based exploration"