优化的前端UI

src/cognitive_agent/infrastructure/speech/understanding.py

@@ -36,9 +36,9 @@ from cognitive_agent.infrastructure.speech.vocabulary import (
 
 _CLAUSE_SPLIT = re.compile(r"[\s，,；;。！？!?…]+")
 _POSITIVE_SPLIT = re.compile(
-    rf"[\s，,；;]*({'|'.join(map(re.escape, POSITIVE_CONNECTORS))})[\s，,；;]*"
+    rf"[\s，,；;]*({'|'.join(map(re.escape, sorted(POSITIVE_CONNECTORS, key=len, reverse=True)))})[\s，,；;]*"
 )
-_COORDINATION_SPLIT = re.compile(r"\s*(?:以及|和|与|及|、)\s*")
+_COORDINATION_SPLIT = re.compile(r"\s*(?:还有|以及|和|跟|与|及|、)\s*")
 _CORRECTION_PATTERN = re.compile(
     rf"[\s，,；;]*({'|'.join(map(re.escape, sorted(CORRECTION_MARKERS, key=len, reverse=True)))}|不[，,]+是)[\s，,；;]*"
 )
@@ -62,6 +62,14 @@ _PERCENT = re.compile(r"(-?\d+(?:\.\d+)?)\s*[%％]")
 _CHINESE_PERCENT = re.compile(r"百分之([零〇一二两三四五六七八九十百点\d.]+)")
 _TERMINAL = re.compile(r"[。！？!?]+$")
 _CONTROL_SCAFFOLD = re.compile(r"^(?:把|将|请|给我|帮我)+")
+_STRONG_CONTINUATION_ACTION = re.compile(
+    r"打开|开启|启动|关闭|关掉|关上|停止|调整|调节|设置|设成|设为|设到|"
+    r"调到|调成|调为|变成|开到|向前|往前|朝前|向后|往后|朝后|"
+    r"向上|往上|向下|往下|升高|降低"
+)
+_NON_NUMERIC_ONE = re.compile(
+    r"(?:一下(?:子)?|一会儿?|一点(?:点|儿)?)(?![零〇一二两三四五六七八九])"
+)
 
 
 @dataclass(frozen=True)
@@ -109,12 +117,7 @@ class UnderstandingTranscriptProcessor:
             blocks = [part for part in blocks if part not in POSITIVE_CONNECTORS]
             for raw_block in blocks:
                 block = raw_block
-                if (
-                    shared_target is not None
-                    and not device_terms_in(block)
-                    and not _CORRECTION_PATTERN.search(block)
-                    and not _CANCELLATION_PATTERN.search(block)
-                ):
+                if shared_target is not None and self._should_inherit_shared_target(block):
                     block = self._inherit_shared_target(block, shared_target)
                 clauses.append(block)
                 (
@@ -269,7 +272,19 @@ class UnderstandingTranscriptProcessor:
                 reasons.append("correction_not_resolved")
             return active, cancelled, reasons, bool(new.commands), resolution_ms
 
-        analyses = [self._analyse_fragment(item) for item in self._plain_fragments(block)]
+        analyses: list[_FragmentAnalysis] = []
+        local_shared_target: SpeechCommandIntent | None = None
+        for item in self._plain_fragments(block):
+            fragment = item
+            if (
+                local_shared_target is not None
+                and self._should_inherit_shared_target(fragment)
+            ):
+                fragment = self._inherit_shared_target(fragment, local_shared_target)
+            analysis = self._analyse_fragment(fragment)
+            analyses.append(analysis)
+            if analysis.commands:
+                local_shared_target = analysis.commands[-1]
         resolution_ms += sum(item.resolution_ms for item in analyses)
         for index, analysis in enumerate(analyses):
             following = analyses[index + 1] if index + 1 < len(analyses) else None
@@ -329,6 +344,8 @@ class UnderstandingTranscriptProcessor:
             return application_analysis
         devices = device_terms_in(text)
         actions = action_terms_in(text)
+        if devices and not actions and self._has_explicit_set_value(text):
+            actions = [(0, 0, "set", "implicit_set")]
         mode = self.resolver.match_mode(text)
         if mode is not None and (devices or actions):
             term = (
@@ -738,11 +755,18 @@ class UnderstandingTranscriptProcessor:
     def _commands_for(
         self, text: str, term: DeviceTerm, action: str, negated: bool
     ) -> tuple[list[SpeechCommandIntent], list[str], float]:
+        implicit_driver_temperature = (
+            term.semantic_key == "hvac.power"
+            and self._numeric_value(text) is not None
+            and bool(re.search(r"摄氏度|度|温度", text))
+        )
         default_scope = (
             "passenger"
             if term.domain == "seat"
             else "front_left"
             if term.domain == "window"
+            else "driver"
+            if term.semantic_key == "hvac.temperature" or implicit_driver_temperature
             else "all"
             if term.semantic_key == "reading_light.power"
             else "vehicle"
@@ -1160,6 +1184,10 @@ class UnderstandingTranscriptProcessor:
 
     def _inherit_shared_target(self, fragment: str, command: SpeechCommandIntent) -> str:
         target = self._display_target(command)
+        if self._has_explicit_scope(fragment):
+            spoken_scope = str(command.qualifiers.get("spoken_scope", ""))
+            if spoken_scope and target.startswith(spoken_scope):
+                target = target[len(spoken_scope) :]
         if (
             command.domain == "seat"
             and not command.qualifiers.get("target_explicit")
@@ -1167,8 +1195,38 @@ class UnderstandingTranscriptProcessor:
         ):
             scope = "主驾" if command.target_scope in {"driver", "front_left"} else "副驾"
             target = f"{scope}{target}"
+        if not action_terms_in(fragment) and self._has_explicit_set_value(fragment):
+            return f"{target}设置为{fragment}"
         return f"{target}{fragment}"
 
+    @staticmethod
+    def _has_explicit_scope(text: str) -> bool:
+        return any(
+            alias in text
+            for aliases in SCOPE_ALIASES.values()
+            for alias in aliases
+        )
+
+    @classmethod
+    def _has_explicit_set_value(cls, text: str) -> bool:
+        if cls._percentage_value(text) is not None or color_in(text) is not None:
+            return True
+        return cls._numeric_value(text) is not None and bool(
+            re.search(r"摄氏度|百分比|度|档|级|[%％]", text)
+        )
+
+    @classmethod
+    def _should_inherit_shared_target(cls, fragment: str) -> bool:
+        if (
+            device_terms_in(fragment)
+            or _CORRECTION_PATTERN.search(fragment)
+            or _CANCELLATION_PATTERN.search(fragment)
+        ):
+            return False
+        return bool(_STRONG_CONTINUATION_ACTION.search(fragment)) or cls._has_explicit_set_value(
+            fragment
+        )
+
     @staticmethod
     def _cancel_commands(
         commands: Iterable[SpeechCommandIntent], *, negated: bool = False
@@ -1259,6 +1317,7 @@ class UnderstandingTranscriptProcessor:
 
     @staticmethod
     def _numeric_value(text: str) -> int | float | None:
+        text = _NON_NUMERIC_ONE.sub("", text)
         match = _NUMBER.search(text)
         if match:
             value = float(match.group(1))

src/cognitive_agent/infrastructure/speech/vocabulary.py

@@ -44,7 +44,23 @@ DEVICE_TERMS: tuple[DeviceTerm, ...] = (
         "seat",
         "seat.longitudinal_position",
         "seat_longitudinal_motor",
-        ("座椅", "坐椅", "座位"),
+        (
+            "座椅",
+            "坐椅",
+            "座位",
+            "驾驶座椅",
+            "驾驶座",
+            "驾驶席",
+            "司机座椅",
+            "司机座",
+            "主驾驶座椅",
+            "主驾驶座",
+            "乘客座椅",
+            "乘客座",
+            "副驾驶座椅",
+            "副驾驶座",
+            "副驾驶席",
+        ),
     ),
     DeviceTerm(
         "座椅坐垫",
@@ -65,14 +81,42 @@ DEVICE_TERMS: tuple[DeviceTerm, ...] = (
         "seat",
         "seat.ventilation",
         "seat",
-        ("座椅通风", "通风座椅"),
+        (
+            "座椅通风",
+            "通风座椅",
+            "驾驶座通风",
+            "驾驶席通风",
+            "司机座通风",
+            "司机座椅通风",
+            "主驾驶座通风",
+            "主驾驶座椅通风",
+            "乘客座通风",
+            "乘客座椅通风",
+            "副驾驶座通风",
+            "副驾驶座椅通风",
+            "副驾驶席通风",
+        ),
     ),
     DeviceTerm(
         "座椅加热",
         "seat",
         "seat.heating",
         "seat",
-        ("座椅加热", "加热座椅"),
+        (
+            "座椅加热",
+            "加热座椅",
+            "驾驶座加热",
+            "驾驶席加热",
+            "司机座加热",
+            "司机座椅加热",
+            "主驾驶座加热",
+            "主驾驶座椅加热",
+            "乘客座加热",
+            "乘客座椅加热",
+            "副驾驶座加热",
+            "副驾驶座椅加热",
+            "副驾驶席加热",
+        ),
     ),
     DeviceTerm(
         "车窗",
@@ -126,8 +170,26 @@ DEVICE_TERMS: tuple[DeviceTerm, ...] = (
 )
 
 SCOPE_ALIASES: dict[str, tuple[str, ...]] = {
-    "driver": ("主驾", "主驾驶", "驾驶位", "司机位", "左前"),
-    "passenger": ("副驾", "副驾驶", "乘客位", "右前"),
+    "driver": (
+        "主驾",
+        "主驾驶",
+        "主驾驶座",
+        "驾驶位",
+        "驾驶座",
+        "驾驶席",
+        "司机位",
+        "司机座",
+        "左前",
+    ),
+    "passenger": (
+        "副驾",
+        "副驾驶",
+        "副驾驶座",
+        "副驾驶席",
+        "乘客位",
+        "乘客座",
+        "右前",
+    ),
     "front_row": ("前排",),
     "rear_row": ("后排",),
     "rear_left": ("左后", "后排左侧"),
@@ -148,8 +210,12 @@ ACTION_ALIASES: dict[str, tuple[str, ...]] = {
         "设置为",
         "调到",
         "调成",
+        "调为",
         "变成",
         "设为",
+        "设成",
+        "设到",
+        "设置",
         "开到",
     ),
     "forward": ("向前", "往前", "朝前", "前移", "前倾"),
@@ -176,13 +242,23 @@ CORRECTION_MARKERS: tuple[str, ...] = (
 
 CANCELLATION_MARKERS: tuple[str, ...] = ("算了", "取消", "作废")
 POSITIVE_CONNECTORS: tuple[str, ...] = (
+    "接下来再",
+    "紧接着",
+    "再然后",
+    "接下来",
     "然后",
     "同时",
     "并且",
     "接着",
     "随后",
+    "之后",
+    "而后",
+    "顺便",
+    "另外",
+    "最后",
     "再",
     "并",
+    "先",
 )
 NEGATION_PREFIXES: tuple[str, ...] = ("不要", "别", "不想", "无需", "不必", "禁止")
 UNITS: tuple[str, ...] = ("摄氏度", "百分比", "毫米", "厘米", "毫秒", "分钟", "度", "%", "档", "秒")