q792602257/DataMate
0
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

feat(annotation): 添加分段标注统计和进度跟踪功能

Browse Source 
- 新增 SegmentStats 类型定义用于分段统计
- 实现分段标注进度计算和缓存机制
- 添加标注任务状态判断逻辑支持分段模式
- 集成分段统计数据显示到任务列表界面
- 实现分段总数自动计算和验证功能
- 扩展标注状态枚举支持进行中标注状态
- 优化任务选择逻辑基于分段完成状态
- 添加分段统计数据预加载和同步机制
This commit is contained in:
Jerry Yan
2026-01-31 15:42:04 +08:00
parent 3e0a15ac8e
commit 33cf65c9f8
5 changed files with 265 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

173
frontend/src/pages/DataAnnotation/Annotate/LabelStudioTextEditor.tsx
View File

@@ -28,6 +28,7 @@ type EditorTaskListItem = {
hasAnnotation: boolean;
annotationUpdatedAt?: string | null;
annotationStatus?: AnnotationResultStatus | null;
segmentStats?: SegmentStats;
};
type LsfMessage = {
@@ -45,6 +46,11 @@ type SegmentInfo = {
chunkIndex: number;
};
type SegmentStats = {
done: number;
total: number;
};
type ApiResponse<T> = {
code?: number;
message?: string;
@@ -136,6 +142,16 @@ const isAnnotationResultEmpty = (annotation?: Record<string, unknown>) => {
};
const resolveTaskStatusMeta = (item: EditorTaskListItem) => {
const segmentSummary = resolveSegmentSummary(item);
if (segmentSummary) {
if (segmentSummary.done >= segmentSummary.total) {
return { text: "已标注", type: "success" as const };
}
if (segmentSummary.done > 0) {
return { text: "标注中", type: "warning" as const };
}
return { text: "未标注", type: "secondary" as const };
}
if (!item.hasAnnotation) {
return { text: "未标注", type: "secondary" as const };
}
@@ -145,6 +161,9 @@ const resolveTaskStatusMeta = (item: EditorTaskListItem) => {
if (item.annotationStatus === AnnotationResultStatus.NOT_APPLICABLE) {
return { text: NOT_APPLICABLE_LABEL, type: "warning" as const };
}
if (item.annotationStatus === AnnotationResultStatus.IN_PROGRESS) {
return { text: "标注中", type: "warning" as const };
}
return { text: "已标注", type: "success" as const };
};
@@ -184,6 +203,25 @@ const buildAnnotationSnapshot = (annotation?: Record<string, unknown>) => {
const buildSnapshotKey = (fileId: string, segmentIndex?: number) =>
`${fileId}::${segmentIndex ?? "full"}`;
const buildSegmentStats = (segmentList?: SegmentInfo[] | null): SegmentStats | null => {
if (!Array.isArray(segmentList) || segmentList.length === 0) return null;
const total = segmentList.length;
const done = segmentList.reduce((count, seg) => count + (seg.hasAnnotation ? 1 : 0), 0);
return { done, total };
};
const normalizeSegmentStats = (stats?: SegmentStats | null): SegmentStats | null => {
if (!stats) return null;
const total = Number(stats.total);
const done = Number(stats.done);
if (!Number.isFinite(total) || total <= 0) return null;
const safeDone = Math.min(Math.max(done, 0), total);
return { done: safeDone, total };
};
const resolveSegmentSummary = (item: EditorTaskListItem) =>
normalizeSegmentStats(item.segmentStats);
const mergeTaskItems = (base: EditorTaskListItem[], next: EditorTaskListItem[]) => {
if (next.length === 0) return base;
const seen = new Set(base.map((item) => item.fileId));
@@ -234,6 +272,9 @@ export default function LabelStudioTextEditor() {
const exportCheckSeqRef = useRef(0);
const savedSnapshotsRef = useRef<Record<string, string>>({});
const pendingAutoAdvanceRef = useRef(false);
const segmentStatsCacheRef = useRef<Record<string, SegmentStats>>({});
const segmentStatsSeqRef = useRef(0);
const segmentStatsLoadingRef = useRef<Set<string>>(new Set());
const [loadingProject, setLoadingProject] = useState(true);
const [loadingTasks, setLoadingTasks] = useState(false);
@@ -276,6 +317,70 @@ export default function LabelStudioTextEditor() {
win.postMessage({ type, payload }, origin);
}, [origin]);
const applySegmentStats = useCallback((fileId: string, stats: SegmentStats | null) => {
if (!fileId) return;
const normalized = normalizeSegmentStats(stats);
setTasks((prev) =>
prev.map((item) =>
item.fileId === fileId
? { ...item, segmentStats: normalized || undefined }
: item
)
);
}, []);
const updateSegmentStatsCache = useCallback((fileId: string, stats: SegmentStats | null) => {
if (!fileId) return;
const normalized = normalizeSegmentStats(stats);
if (normalized) {
segmentStatsCacheRef.current[fileId] = normalized;
} else {
delete segmentStatsCacheRef.current[fileId];
}
applySegmentStats(fileId, normalized);
}, [applySegmentStats]);
const fetchSegmentStatsForFile = useCallback(async (fileId: string, seq: number) => {
if (!projectId || !fileId) return;
if (segmentStatsCacheRef.current[fileId] || segmentStatsLoadingRef.current.has(fileId)) return;
segmentStatsLoadingRef.current.add(fileId);
try {
const resp = (await getEditorTaskUsingGet(projectId, fileId, {
segmentIndex: 0,
})) as ApiResponse<EditorTaskResponse>;
if (segmentStatsSeqRef.current !== seq) return;
const data = resp?.data;
if (!data?.segmented) return;
const stats = buildSegmentStats(data.segments);
if (!stats) return;
segmentStatsCacheRef.current[fileId] = stats;
applySegmentStats(fileId, stats);
} catch (e) {
console.error(e);
} finally {
segmentStatsLoadingRef.current.delete(fileId);
}
}, [applySegmentStats, projectId]);
const prefetchSegmentStats = useCallback((items: EditorTaskListItem[]) => {
if (!projectId) return;
const fileIds = items
.map((item) => item.fileId)
.filter((fileId) => fileId && !segmentStatsCacheRef.current[fileId]);
if (fileIds.length === 0) return;
const seq = segmentStatsSeqRef.current;
let cursor = 0;
const workerCount = Math.min(3, fileIds.length);
const runWorker = async () => {
while (cursor < fileIds.length && segmentStatsSeqRef.current === seq) {
const fileId = fileIds[cursor];
cursor += 1;
await fetchSegmentStatsForFile(fileId, seq);
}
};
void Promise.all(Array.from({ length: workerCount }, () => runWorker()));
}, [fetchSegmentStatsForFile, projectId]);
const confirmEmptyAnnotationStatus = useCallback(() => {
return new Promise<AnnotationResultStatus | null>((resolve) => {
let resolved = false;
@@ -327,8 +432,13 @@ export default function LabelStudioTextEditor() {
}, [message, projectId]);
const updateTaskSelection = useCallback((items: EditorTaskListItem[]) => {
const isCompleted = (item: EditorTaskListItem) => {
const summary = resolveSegmentSummary(item);
if (summary) return summary.done >= summary.total;
return item.hasAnnotation;
};
const defaultFileId =
items.find((item) => !item.hasAnnotation)?.fileId || items[0]?.fileId || "";
items.find((item) => !isCompleted(item))?.fileId || items[0]?.fileId || "";
setSelectedFileId((prev) => {
if (prev && items.some((item) => item.fileId === prev)) return prev;
return defaultFileId;
@@ -385,6 +495,9 @@ export default function LabelStudioTextEditor() {
if (mode === "reset") {
prefetchSeqRef.current += 1;
setPrefetching(false);
segmentStatsSeqRef.current += 1;
segmentStatsCacheRef.current = {};
segmentStatsLoadingRef.current = new Set();
}
if (mode === "append") {
setLoadingMore(true);
@@ -469,13 +582,16 @@ export default function LabelStudioTextEditor() {
? resolveSegmentIndex(data.currentSegmentIndex) ?? 0
: undefined;
if (data?.segmented) {
const stats = buildSegmentStats(data.segments);
setSegmented(true);
setSegments(data.segments || []);
setCurrentSegmentIndex(segmentIndex ?? 0);
updateSegmentStatsCache(fileId, stats);
} else {
setSegmented(false);
setSegments([]);
setCurrentSegmentIndex(0);
updateSegmentStatsCache(fileId, null);
}
const taskData = {
@@ -535,7 +651,7 @@ export default function LabelStudioTextEditor() {
} finally {
if (seq === initSeqRef.current) setLoadingTaskDetail(false);
}
}, [iframeReady, message, postToIframe, project, projectId]);
}, [iframeReady, message, postToIframe, project, projectId, updateSegmentStatsCache]);
const advanceAfterSave = useCallback(async (fileId: string, segmentIndex?: number) => {
if (!fileId) return;
@@ -643,13 +759,13 @@ export default function LabelStudioTextEditor() {
// 分段模式下更新当前段落的标注状态
if (segmented && segmentIndex !== undefined) {
setSegments((prev) =>
prev.map((seg) =>
seg.idx === segmentIndex
? { ...seg, hasAnnotation: true }
: seg
)
const nextSegments = segments.map((seg) =>
seg.idx === segmentIndex
? { ...seg, hasAnnotation: true }
: seg
);
setSegments(nextSegments);
updateSegmentStatsCache(String(fileId), buildSegmentStats(nextSegments));
}
if (options?.autoAdvance) {
await advanceAfterSave(String(fileId), segmentIndex);
@@ -669,8 +785,10 @@ export default function LabelStudioTextEditor() {
message,
projectId,
segmented,
segments,
selectedFileId,
tasks,
updateSegmentStatsCache,
]);
const requestExportForCheck = useCallback(() => {
@@ -834,6 +952,9 @@ export default function LabelStudioTextEditor() {
setSegments([]);
setCurrentSegmentIndex(0);
savedSnapshotsRef.current = {};
segmentStatsSeqRef.current += 1;
segmentStatsCacheRef.current = {};
segmentStatsLoadingRef.current = new Set();
if (exportCheckRef.current?.timer) {
window.clearTimeout(exportCheckRef.current.timer);
}
@@ -847,6 +968,12 @@ export default function LabelStudioTextEditor() {
loadTasks({ mode: "reset" });
}, [project?.supported, loadTasks]);
useEffect(() => {
if (!segmented) return;
if (tasks.length === 0) return;
prefetchSegmentStats(tasks);
}, [prefetchSegmentStats, segmented, tasks]);
useEffect(() => {
if (!selectedFileId) return;
initEditorForFile(selectedFileId);
@@ -1097,6 +1224,7 @@ export default function LabelStudioTextEditor() {
dataSource={tasks}
loadMore={loadMoreNode}
renderItem={(item) => {
const segmentSummary = resolveSegmentSummary(item);
const statusMeta = resolveTaskStatusMeta(item);
return (
<List.Item
@@ -1110,18 +1238,25 @@ export default function LabelStudioTextEditor() {
onClick={() => setSelectedFileId(item.fileId)}
>
<div className="flex flex-col w-full gap-1">
<Typography.Text ellipsis style={{ fontSize: 13 }}>
{item.fileName}
</Typography.Text>
<div className="flex items-center justify-between">
<Typography.Text type={statusMeta.type} style={{ fontSize: 11 }}>
{statusMeta.text}
<Typography.Text ellipsis style={{ fontSize: 13 }}>
{item.fileName}
</Typography.Text>
{item.annotationUpdatedAt && (
<Typography.Text type="secondary" style={{ fontSize: 10 }}>
{item.annotationUpdatedAt}
</Typography.Text>
)}
<div className="flex items-center justify-between">
<div className="flex items-center gap-2">
<Typography.Text type={statusMeta.type} style={{ fontSize: 11 }}>
{statusMeta.text}
</Typography.Text>
{segmentSummary && (
<Typography.Text type="secondary" style={{ fontSize: 10 }}>
已标注 {segmentSummary.done}/{segmentSummary.total}
</Typography.Text>
)}
</div>
{item.annotationUpdatedAt && (
<Typography.Text type="secondary" style={{ fontSize: 10 }}>
{item.annotationUpdatedAt}
</Typography.Text>
)}
</div>
</div>
</List.Item>

1
frontend/src/pages/DataAnnotation/annotation.model.ts
View File

@@ -10,6 +10,7 @@ export enum AnnotationTaskStatus {
export enum AnnotationResultStatus {
ANNOTATED = "ANNOTATED",
IN_PROGRESS = "IN_PROGRESS",
NO_ANNOTATION = "NO_ANNOTATION",
NOT_APPLICABLE = "NOT_APPLICABLE",
}

9
runtime/datamate-python/app/db/models/annotation_management.py
View File

@@ -9,10 +9,17 @@ from app.db.session import Base
ANNOTATION_STATUS_ANNOTATED = "ANNOTATED"
ANNOTATION_STATUS_NO_ANNOTATION = "NO_ANNOTATION"
ANNOTATION_STATUS_NOT_APPLICABLE = "NOT_APPLICABLE"
ANNOTATION_STATUS_IN_PROGRESS = "IN_PROGRESS"
ANNOTATION_STATUS_VALUES = {
ANNOTATION_STATUS_ANNOTATED,
ANNOTATION_STATUS_NO_ANNOTATION,
ANNOTATION_STATUS_NOT_APPLICABLE,
ANNOTATION_STATUS_IN_PROGRESS,
}
ANNOTATION_STATUS_CLIENT_VALUES = {
ANNOTATION_STATUS_ANNOTATED,
ANNOTATION_STATUS_NO_ANNOTATION,
ANNOTATION_STATUS_NOT_APPLICABLE,
}
class AnnotationTemplate(Base):
@@ -101,7 +108,7 @@ class AnnotationResult(Base):
String(32),
nullable=False,
default=ANNOTATION_STATUS_ANNOTATED,
comment="标注状态: ANNOTATED/NO_ANNOTATION/NOT_APPLICABLE",
comment="标注状态: ANNOTATED/NO_ANNOTATION/NOT_APPLICABLE/IN_PROGRESS",
)
created_at = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="创建时间")
updated_at = Column(TIMESTAMP, server_default=func.current_timestamp(), onupdate=func.current_timestamp(), comment="更新时间")

4
runtime/datamate-python/app/module/annotation/schema/editor.py
View File

@@ -16,6 +16,7 @@ from pydantic import BaseModel, Field, ConfigDict
from app.db.models.annotation_management import (
ANNOTATION_STATUS_ANNOTATED,
ANNOTATION_STATUS_IN_PROGRESS,
ANNOTATION_STATUS_NO_ANNOTATION,
ANNOTATION_STATUS_NOT_APPLICABLE,
)
@@ -25,6 +26,7 @@ class AnnotationStatus(str, Enum):
"""标注状态枚举"""
ANNOTATED = ANNOTATION_STATUS_ANNOTATED
IN_PROGRESS = ANNOTATION_STATUS_IN_PROGRESS
NO_ANNOTATION = ANNOTATION_STATUS_NO_ANNOTATION
NOT_APPLICABLE = ANNOTATION_STATUS_NOT_APPLICABLE
@@ -112,7 +114,7 @@ class UpsertAnnotationRequest(BaseModel):
annotation_status: Optional[AnnotationStatus] = Field(
None,
alias="annotationStatus",
description="标注状态（无标注传 NO_ANNOTATION，不适用传 NOT_APPLICABLE）",
description="标注状态（无标注传 NO_ANNOTATION，不适用传 NOT_APPLICABLE，IN_PROGRESS 由后端维护）",
)
expected_updated_at: Optional[datetime] = Field(
None,

109
runtime/datamate-python/app/module/annotation/service/editor.py
View File

@@ -26,9 +26,10 @@ from app.core.logging import get_logger
from app.db.models import AnnotationResult, Dataset, DatasetFiles, LabelingProject, LabelingProjectFile
from app.db.models.annotation_management import (
ANNOTATION_STATUS_ANNOTATED,
ANNOTATION_STATUS_IN_PROGRESS,
ANNOTATION_STATUS_CLIENT_VALUES,
ANNOTATION_STATUS_NO_ANNOTATION,
ANNOTATION_STATUS_NOT_APPLICABLE,
ANNOTATION_STATUS_VALUES,
)
from app.module.annotation.config import LabelStudioTagConfig
from app.module.annotation.schema.editor import (
@@ -61,6 +62,7 @@ SEGMENT_INDEX_KEY = "segment_index"
SEGMENT_INDEX_CAMEL_KEY = "segmentIndex"
SEGMENTED_KEY = "segmented"
SEGMENTS_KEY = "segments"
SEGMENT_TOTAL_KEY = "total_segments"
SEGMENT_RESULT_KEY = "result"
SEGMENT_CREATED_AT_KEY = "created_at"
SEGMENT_UPDATED_AT_KEY = "updated_at"
@@ -416,6 +418,76 @@ class AnnotationEditorService:
result = payload.get(SEGMENT_RESULT_KEY)
return isinstance(result, list) and len(result) > 0
@staticmethod
def _resolve_segment_total(payload: Optional[Dict[str, Any]]) -> Optional[int]:
if not payload or not isinstance(payload, dict):
return None
value = payload.get(SEGMENT_TOTAL_KEY)
if isinstance(value, int):
return value if value > 0 else None
if isinstance(value, float) and value.is_integer():
return int(value) if value > 0 else None
if isinstance(value, str) and value.isdigit():
parsed = int(value)
return parsed if parsed > 0 else None
return None
async def _compute_segment_total(
self,
project: LabelingProject,
file_record: DatasetFiles,
file_id: str,
) -> Optional[int]:
dataset_type = self._normalize_dataset_type(await self._get_dataset_type(project.dataset_id))
if dataset_type != DATASET_TYPE_TEXT:
return None
if not self._resolve_segmentation_enabled(project):
return None
text_content = await self._fetch_text_content_via_download_api(project.dataset_id, file_id)
if not isinstance(text_content, str):
return None
label_config = await self._resolve_project_label_config(project)
primary_text_key = self._resolve_primary_text_key(label_config)
file_name = str(getattr(file_record, "file_name", "")).lower()
records: List[Tuple[Optional[Dict[str, Any]], str]] = []
if file_name.endswith(JSONL_EXTENSION):
records = self._parse_jsonl_records(text_content)
else:
parsed_payload = self._try_parse_json_payload(text_content)
if parsed_payload:
records = [(parsed_payload, text_content)]
if not records:
records = [(None, text_content)]
record_texts = [
self._resolve_primary_text_value(payload, raw_text, primary_text_key)
for payload, raw_text in records
]
if not record_texts:
record_texts = [text_content]
needs_segmentation = len(records) > 1 or any(
len(text or "") > self.SEGMENT_THRESHOLD for text in record_texts
)
if not needs_segmentation:
return None
splitter = AnnotationTextSplitter(max_chars=self.SEGMENT_THRESHOLD)
total_segments = 0
for record_text in record_texts:
normalized_text = record_text or ""
if len(normalized_text) > self.SEGMENT_THRESHOLD:
raw_segments = splitter.split(normalized_text)
total_segments += len(raw_segments) if raw_segments else 1
else:
total_segments += 1
return total_segments if total_segments > 0 else 1
@classmethod
def _build_source_document_filter(cls):
file_type_lower = func.lower(DatasetFiles.file_type)
@@ -946,19 +1018,36 @@ class AnnotationEditorService:
final_payload = annotation_payload
requested_status = request.annotation_status
if requested_status is not None and requested_status not in ANNOTATION_STATUS_VALUES:
if requested_status is not None and requested_status not in ANNOTATION_STATUS_CLIENT_VALUES:
raise HTTPException(status_code=400, detail="annotationStatus 不合法")
has_result = self._has_annotation_result(final_payload)
if has_result:
final_status = ANNOTATION_STATUS_ANNOTATED
segment_total = None
segment_done = None
if request.segment_index is not None:
segment_total = self._resolve_segment_total(final_payload)
if segment_total is None:
segment_total = await self._compute_segment_total(project, file_record, file_id)
if segment_total and segment_total > 0:
final_payload[SEGMENT_TOTAL_KEY] = segment_total
segment_done = len(self._extract_segment_annotations(final_payload))
if (
segment_total is not None
and segment_done is not None
and segment_done < segment_total
):
final_status = ANNOTATION_STATUS_IN_PROGRESS
else:
if requested_status == ANNOTATION_STATUS_NO_ANNOTATION:
final_status = ANNOTATION_STATUS_NO_ANNOTATION
elif requested_status == ANNOTATION_STATUS_NOT_APPLICABLE:
final_status = ANNOTATION_STATUS_NOT_APPLICABLE
has_result = self._has_annotation_result(final_payload)
if has_result:
final_status = ANNOTATION_STATUS_ANNOTATED
else:
raise HTTPException(status_code=400, detail="未发现标注内容，请确认无标注/不适用后再保存")
if requested_status == ANNOTATION_STATUS_NO_ANNOTATION:
final_status = ANNOTATION_STATUS_NO_ANNOTATION
elif requested_status == ANNOTATION_STATUS_NOT_APPLICABLE:
final_status = ANNOTATION_STATUS_NOT_APPLICABLE
else:
raise HTTPException(status_code=400, detail="未发现标注内容，请确认无标注/不适用后再保存")
if existing:
if request.expected_updated_at and existing.updated_at: