调问更新 8.8 ~ 8.22：新增审核与多方评审，数据导出及问卷能力全面优化

{"title_zh":"调问 8.8–8.22 更新：从问卷收集走向可审计的协作调研","body_zh":"本次调问更新的重点，不只是增加题型，而是把问卷从“创建—填写—导出”的单线流程，扩展为包含人工审核、举报处理和多方评审的协作系统。与此同时，文本智能识别、模板管理、题库、编辑填写体验以及数据导出都得到补强。\n\n对于企业内部调研、客户满意度调查和合规场景，这类变化很实际：问卷发布之后，谁能审核、谁能复核、哪些内容需要处理，以及最终数据如何留痕，都会直接影响系统能否进入日常业务。\n\n## 审核与评审让问卷进入协作流程\n\n新增的人工审核和举报问卷功能，解决的是内容治理问题。企业问卷往往由多个部门共同维护，公开收集后还可能出现不合适内容、重复提交或需要人工确认的答案。将审核动作纳入系统，比依赖线下表格或聊天记录更容易形成稳定流程。\n\n多方评审则适合需要多个角色共同判断的场景，例如产品需求评估、供应商打分、培训效果评价和项目验收。这里的关键不是“多一个按钮”，而是把评审角色、处理状态和最终结果组织起来，让调研结果具备更清晰的责任边界。\n\n可以这样设计一个最小审核状态流，实际接入时再映射到调问现有的权限和接口：\n\npython\nfrom enum import StrEnum\n\n\nclass ReviewStatus(StrEnum):\n PENDING = "pending"\n APPROVED = "approved"\n REJECTED = "rejected"\n REPORTED = "reported"\n\n\nALLOWED_TRANSITIONS = {\n ReviewStatus.PENDING: {ReviewStatus.APPROVED, ReviewStatus.REJECTED, ReviewStatus.REPORTED},\n ReviewStatus.REPORTED: {ReviewStatus.APPROVED, ReviewStatus.REJECTED},\n ReviewStatus.APPROVED: set(),\n ReviewStatus.REJECTED: set(),\n}\n\n\ndef change_status(current: ReviewStatus, target: ReviewStatus) -> ReviewStatus:\n if target not in ALLOWED_TRANSITIONS[current]:\n raise ValueError(f"invalid transition: {current} -> {target}")\n return target\n\n\nprint(change_status(ReviewStatus.PENDING, ReviewStatus.REPORTED))\n\n\n运行环境为 Python 3.11 或更高版本。接入真实系统时，还应在状态变更记录中保存操作者、时间、原因和关联问卷 ID；对于高风险业务，建议禁止直接覆盖审核记录。\n\n## 判断题与文本识别扩大问卷表达能力\n\n判断题适合知识检查、资格确认、流程分流和事实核验等场景。它的价值在于让问卷设计者不必用单选题绕出“是/否”结构，也让后续统计和导出拥有更明确的数据类型。\n\n文本智能识别的优化，则可能减少录入题目和整理内容时的重复操作。对使用者来说，识别能力只有在边界清楚时才可靠：识别结果应当允许人工修改，题干、选项和格式不能因为一次误识别而悄悄改变。涉及敏感内容时，还需要考虑文本是否会被发送到外部服务，以及日志中是否会保留原文。\n\n一个实用的落地检查顺序是：\n\n1. 用短题干、长题干、带编号选项和混合标点分别测试。\n2. 对识别结果执行人工确认，而不是直接发布。\n3. 保存原始输入和修改后的版本，便于定位错误。\n4. 对姓名、电话、地址等字段做脱敏和访问控制。\n\n## 模板、题库和导出决定长期维护成本\n\n模板管理和题库能力，解决的是“下一份问卷如何快速开始”。成熟的企业调研系统通常不会每次从空白页面搭建，而是复用经过验证的题目、说明文字和评分规则。\n\n复用时要特别注意版本问题。题库中的题目一旦被问卷引用，后续修改不应无条件改变历史问卷，否则历史数据的含义会发生漂移。可以采用“引用时复制快照”或“题目版本号”的方式，确保问卷发布后使用固定内容。\n\n数据导出优化同样影响业务闭环。导出的文件至少需要明确列名、题目顺序、答案类型和缺失值规则；多方评审场景还应区分原始回答、评审意见、评分和汇总结果。导出前最好先用小样本核对：填写人数是否一致，判断题是否保持布尔语义，中文列名和特殊字符是否能被目标表格软件正确打开。\n\n## 可以怎样把更新接入企业流程\n\n不必一次启用全部能力。可以按风险和频率分阶段推进：\n\n- 内部问卷先启用模板和题库，统一常用题目。\n- 公开问卷增加人工审核和举报处理，定义处理时限。\n- 需要集体决策的问卷配置多方评审，并明确最终责任人。\n- 数据归档前固定导出字段和文件命名规则。\n- 对智能识别设置人工确认节点，并检查数据出境与隐私要求。\n\n调问坚持前后端代码 100% 开源，这对希望自主部署和持续定制的企业具有直接意义。但开源并不自动等于完成治理：部署方仍需负责权限设计、备份、日志留存、敏感数据保护和升级验证。更稳妥的做法是先选一个低风险问卷试运行，记录审核耗时、导出准确性和参与者反馈，再逐步扩大使用范围。\n\n## 结语：把功能更新转化为可执行制度\n\n8.8–8.22 这批更新覆盖了问卷全生命周期：设计阶段有题型、模板和题库支持，填写阶段改善识别与交互，发布后增加审核、举报和多方评审，结束后继续完善数据导出。\n\n真正落地时，建议同时写下三件事：谁能发布，谁能审核，谁对最终数据负责。软件功能负责提供路径，企业流程负责定义边界；两者结合，问卷系统才会从一次性收集工具变成可持续使用的调研基础设施。","title_en":"Diaowen’s Aug. 8–22 Update: Building an Auditable, Collaborative Survey Workflow","body_en":"Diaowen’s latest update goes beyond adding another question type. It expands the survey lifecycle from a simple “create, collect, export” flow into a collaborative workflow with manual review, report handling, multi-party evaluation, and stronger data management.\n\nThe release also improves intelligent text recognition, templates, question banks, editing and answering experiences, and data export. For internal research, customer feedback, and compliance-sensitive surveys, these changes address a practical question: who can review content, who can make a decision, and how can the final result be traced later?\n\n## From Survey Collection to Content Governance\n\nManual review and reported-survey handling are useful when surveys are maintained by several teams or opened to a broad audience. Responses and submissions may need confirmation, moderation, or follow-up. Keeping those actions in the system is easier to audit than relying on spreadsheets and chat messages.\n\nMulti-party evaluation is a natural fit for product reviews, supplier scoring, training assessments, and project acceptance. The important design issue is not merely adding several reviewers. It is defining roles, statuses, comments, and ownership of the final decision.\n\nA minimal review state machine can look like this:\n\npython\nfrom enum import StrEnum\n\n\nclass ReviewStatus(StrEnum):\n PENDING = "pending"\n APPROVED = "approved"\n REJECTED = "rejected"\n REPORTED = "reported"\n\n\nALLOWED_TRANSITIONS = {\n ReviewStatus.PENDING: {ReviewStatus.APPROVED, ReviewStatus.REJECTED, ReviewStatus.REPORTED},\n ReviewStatus.REPORTED: {ReviewStatus.APPROVED, ReviewStatus.REJECTED},\n ReviewStatus.APPROVED: set(),\n ReviewStatus.REJECTED: set(),\n}\n\n\ndef change_status(current: ReviewStatus, target: ReviewStatus) -> ReviewStatus:\n if target not in ALLOWED_TRANSITIONS[current]:\n raise ValueError(f"invalid transition: {current} -> {target}")\n return target\n\n\nprint(change_status(ReviewStatus.PENDING, ReviewStatus.REPORTED))\n\n\nThis requires Python 3.11 or later. In a production integration, store the operator, timestamp, reason, and survey ID for every transition. For higher-risk workflows, preserve review history instead of overwriting it.\n\n## Better Question Modeling and Safer Recognition\n\nTrue/false questions are useful for knowledge checks, eligibility confirmation, routing, and fact validation. Treating them as a first-class type makes both authoring and downstream analysis clearer than emulating them with a generic single-choice question.\n\nImprovements to intelligent text recognition can reduce repetitive authoring work. Recognition should remain an assistive step, however. Users need to verify and edit the result before publication, especially when the input contains long prompts, numbered options, mixed punctuation, or sensitive information. Teams should also check whether text is sent to an external service and whether raw content is retained in logs.\n\nA practical verification routine is: test representative inputs, require human confirmation, retain the original and edited versions, and restrict access to personal data such as names, phone numbers, and addresses.\n\n## Templates, Question Banks, and Export Quality\n\nTemplates and question banks lower the cost of creating the next survey. Organizations can reuse tested questions, instructions, and scoring rules instead of rebuilding every form from scratch.\n\nVersioning matters when content is reused. Once a question is part of a published survey, changing the source question should not silently change the meaning of historical data. A snapshot copied at reference time or an explicit question version can prevent that drift.\n\nExport improvements close the operational loop. Exported data should have stable column names, question order, answer types, and missing-value rules. Multi-party reviews may require separate fields for the original response, reviewer comments, scores, and aggregated results. Before archiving, validate a small export against the response count and verify that the target spreadsheet application preserves non-ASCII text and special characters.\n\n## A Practical Adoption Path\n\nOrganizations do not need to enable every capability at once. A staged rollout is easier to measure: start with templates and question banks for internal surveys; add manual review and reporting for public surveys; configure multi-party evaluation where decisions require several roles; standardize export fields and file names; and put a human approval step after intelligent recognition.\n\nDiaowen’s commitment to keeping both frontend and backend code fully open source is relevant to teams that need self-hosting and customization. Open source still leaves operational responsibilities with the deploying organization: permissions, backups, audit logs, sensitive-data protection, and upgrade validation all need explicit ownership.\n\n## Closing Checklist\n\nThe Aug. 8–22 update covers the full survey lifecycle: authoring gains question, template, and question-bank support; answering benefits from recognition and interaction improvements; publication gains review, reporting, and multi-party evaluation; and completion gains better export handling.\n\nFor adoption, document three decisions before rollout: who may publish, who may review, and who owns the final dataset. Product features provide the workflow, while organizational policy defines its boundaries. Together they turn a survey tool into maintainable research infrastructure.","seo_description_en":"Diaowen’s Aug. 8–22 update adds review, reporting, multi-party evaluation, question types, templates, recognition, and better exports for open-source surveys."}