@lokascript/semantic 1.0.0 → 1.1.0

package/dist/index.d.cts

@@ -139,6 +139,15 @@ interface SemanticMetadata {
     readonly sourceText?: string;
     readonly sourcePosition?: SourcePosition;
     readonly patternId?: string;
+    /**
+     * Confidence score for the parse (0-1).
+     * Higher values indicate more certain matches.
+     * - 1.0: Exact match with all roles captured
+     * - 0.8-0.99: High confidence with minor uncertainty (stem matching, optional roles)
+     * - 0.6-0.8: Medium confidence (morphological normalization, defaults applied)
+     * - <0.6: Low confidence (may need fallback to traditional parser)
+     */
+    readonly confidence?: number;
 }
 interface SourcePosition {
     readonly start: number;
@@ -286,6 +295,10 @@ interface ExtractionRule {
     readonly transform?: (raw: string) => SemanticValue;
     /** Default value if not found (for optional roles) */
     readonly default?: SemanticValue;
+    /** Static value extraction (for event handler wrapped commands) */
+    readonly value?: string;
+    /** Extract value from a pattern role by name */
+    readonly fromRole?: string;
 }
 /**
  * Additional constraints on pattern applicability.
@@ -925,7 +938,7 @@ interface PossessiveConfig {
  * Complete language profile for pattern generation.
  */
 interface LanguageProfile {
-    /** ISO 639-1 language code */
+    /** ISO 639-1 or BCP 47 language code (e.g., 'es' or 'es-MX') */
     readonly code: string;
     /** Human-readable language name */
     readonly name: string;
@@ -964,18 +977,33 @@ interface LanguageProfile {
      * Individual keywords can override this via KeywordTranslation.form
      */
     readonly defaultVerbForm?: VerbForm;
+    /**
+     * Base language code to extend (for regional variants).
+     * When set, this profile inherits from the base and overrides specific fields.
+     * Example: 'es-MX' profile with extends: 'es' inherits from Spanish base.
+     */
+    readonly extends?: string;
 }
 /**
  * Configuration for event handler pattern generation.
- * Used by simple SVO languages that don't need hand-crafted patterns.
+ * Supports both SVO and SOV/VSO languages.
  */
 interface EventHandlerConfig {
-    /** Primary event keyword (e.g., 'on', 'bei', 'sur') */
-    readonly keyword: KeywordTranslation$1;
+    /** Primary event keyword (e.g., 'on', 'bei', 'sur') for SVO */
+    readonly keyword?: KeywordTranslation$1;
     /** Source filter marker (e.g., 'from', 'von', 'de') */
-    readonly sourceMarker: RoleMarker;
+    readonly sourceMarker?: RoleMarker;
     /** Conditional keyword (e.g., 'when', 'wenn', 'quand') */
     readonly conditionalKeyword?: KeywordTranslation$1;
+    /** Event marker for SOV/VSO languages (e.g., で (Japanese), 할 때 (Korean), da (Turkish), عند (Arabic)) */
+    readonly eventMarker?: RoleMarker;
+    /** Temporal/conditional markers that can optionally appear with events */
+    readonly temporalMarkers?: string[];
+    /**
+     * Negation marker for expressing negated events (e.g., Arabic عدم = "not/lack of").
+     * Used in patterns like: عند عدم التركيز = "when not focusing" = "on blur"
+     */
+    readonly negationMarker?: RoleMarker;
 }
 /**
  * Verb form used for command keywords.
@@ -1229,6 +1257,13 @@ declare class KoreanTokenizer extends BaseTokenizer {
      * Korean time units attach directly without whitespace.
      */
     private extractKoreanNumber;
+    /**
+     * Try to split a temporal event suffix from a word token.
+     * This handles compact forms like 클릭할때 → 클릭 + 할때
+     *
+     * @returns Split tokens if a suffix is found, null otherwise
+     */
+    private trySplitTemporalSuffix;
 }
 /**
  * Singleton instance.
@@ -1289,6 +1324,14 @@ declare class TurkishTokenizer extends BaseTokenizer {
     constructor();
     tokenize(input: string): TokenStream;
     classifyToken(token: string): TokenKind;
+    /**
+     * Try to match multi-word phrases that function as single units.
+     * Multi-word phrases are included in profileKeywords and sorted longest-first,
+     * so they'll be matched before their constituent words.
+     *
+     * Examples: "üzerine gelme" (hover), "fare üzerinde" (mouseover)
+     */
+    private tryMultiWordPhrase;
     /**
      * Extract a Turkish word.
      * Uses morphological normalization to handle verb conjugations.
@@ -2023,6 +2066,7 @@ declare function isLanguageSupported(code: string): boolean;
 
 /**
  * Try to get a profile, returning undefined if not registered.
+ * Supports fallback: if 'es-MX' is not registered, falls back to 'es'.
  */
 declare function tryGetProfile(code: string): LanguageProfile | undefined;
 /**

package/dist/index.d.ts

@@ -139,6 +139,15 @@ interface SemanticMetadata {
     readonly sourceText?: string;
     readonly sourcePosition?: SourcePosition;
     readonly patternId?: string;
+    /**
+     * Confidence score for the parse (0-1).
+     * Higher values indicate more certain matches.
+     * - 1.0: Exact match with all roles captured
+     * - 0.8-0.99: High confidence with minor uncertainty (stem matching, optional roles)
+     * - 0.6-0.8: Medium confidence (morphological normalization, defaults applied)
+     * - <0.6: Low confidence (may need fallback to traditional parser)
+     */
+    readonly confidence?: number;
 }
 interface SourcePosition {
     readonly start: number;
@@ -286,6 +295,10 @@ interface ExtractionRule {
     readonly transform?: (raw: string) => SemanticValue;
     /** Default value if not found (for optional roles) */
     readonly default?: SemanticValue;
+    /** Static value extraction (for event handler wrapped commands) */
+    readonly value?: string;
+    /** Extract value from a pattern role by name */
+    readonly fromRole?: string;
 }
 /**
  * Additional constraints on pattern applicability.
@@ -925,7 +938,7 @@ interface PossessiveConfig {
  * Complete language profile for pattern generation.
  */
 interface LanguageProfile {
-    /** ISO 639-1 language code */
+    /** ISO 639-1 or BCP 47 language code (e.g., 'es' or 'es-MX') */
     readonly code: string;
     /** Human-readable language name */
     readonly name: string;
@@ -964,18 +977,33 @@ interface LanguageProfile {
      * Individual keywords can override this via KeywordTranslation.form
      */
     readonly defaultVerbForm?: VerbForm;
+    /**
+     * Base language code to extend (for regional variants).
+     * When set, this profile inherits from the base and overrides specific fields.
+     * Example: 'es-MX' profile with extends: 'es' inherits from Spanish base.
+     */
+    readonly extends?: string;
 }
 /**
  * Configuration for event handler pattern generation.
- * Used by simple SVO languages that don't need hand-crafted patterns.
+ * Supports both SVO and SOV/VSO languages.
  */
 interface EventHandlerConfig {
-    /** Primary event keyword (e.g., 'on', 'bei', 'sur') */
-    readonly keyword: KeywordTranslation$1;
+    /** Primary event keyword (e.g., 'on', 'bei', 'sur') for SVO */
+    readonly keyword?: KeywordTranslation$1;
     /** Source filter marker (e.g., 'from', 'von', 'de') */
-    readonly sourceMarker: RoleMarker;
+    readonly sourceMarker?: RoleMarker;
     /** Conditional keyword (e.g., 'when', 'wenn', 'quand') */
     readonly conditionalKeyword?: KeywordTranslation$1;
+    /** Event marker for SOV/VSO languages (e.g., で (Japanese), 할 때 (Korean), da (Turkish), عند (Arabic)) */
+    readonly eventMarker?: RoleMarker;
+    /** Temporal/conditional markers that can optionally appear with events */
+    readonly temporalMarkers?: string[];
+    /**
+     * Negation marker for expressing negated events (e.g., Arabic عدم = "not/lack of").
+     * Used in patterns like: عند عدم التركيز = "when not focusing" = "on blur"
+     */
+    readonly negationMarker?: RoleMarker;
 }
 /**
  * Verb form used for command keywords.
@@ -1229,6 +1257,13 @@ declare class KoreanTokenizer extends BaseTokenizer {
      * Korean time units attach directly without whitespace.
      */
     private extractKoreanNumber;
+    /**
+     * Try to split a temporal event suffix from a word token.
+     * This handles compact forms like 클릭할때 → 클릭 + 할때
+     *
+     * @returns Split tokens if a suffix is found, null otherwise
+     */
+    private trySplitTemporalSuffix;
 }
 /**
  * Singleton instance.
@@ -1289,6 +1324,14 @@ declare class TurkishTokenizer extends BaseTokenizer {
     constructor();
     tokenize(input: string): TokenStream;
     classifyToken(token: string): TokenKind;
+    /**
+     * Try to match multi-word phrases that function as single units.
+     * Multi-word phrases are included in profileKeywords and sorted longest-first,
+     * so they'll be matched before their constituent words.
+     *
+     * Examples: "üzerine gelme" (hover), "fare üzerinde" (mouseover)
+     */
+    private tryMultiWordPhrase;
     /**
      * Extract a Turkish word.
      * Uses morphological normalization to handle verb conjugations.
@@ -2023,6 +2066,7 @@ declare function isLanguageSupported(code: string): boolean;
 
 /**
  * Try to get a profile, returning undefined if not registered.
+ * Supports fallback: if 'es-MX' is not registered, falls back to 'es'.
  */
 declare function tryGetProfile(code: string): LanguageProfile | undefined;
 /**