feat: 独立词库、三向认词与当日快闪复习

## 独立词库

理解词原先只能跟着课程单元走,学完 A0 十课词汇量只增加约 22 个实词,
不足以解决"记不住单词"。新增一份独立词库 assets/words/wordbank.json
(2748 词,A1–B1),挂进 receptiveWordRegistry 的合成单元 bank-A1/A2/B1,
完全复用理解词已有的状态机,不依赖课程进度,第一天就能用。

数据来源、许可与合成规则记在 tool/words/DATA-NOTE.md:CEFR-J 定等级、
公开词书提供音标、AI 重写全部释义并生成例句、OpenSubtitles 提供口语词频。
词书部分为 CC BY-NC-SA 4.0 且上游权利不明,仅供个人非商用;
若要分发或上架,须替换音标那一列。

## 背单词机制

- 间隔阶梯 1/3/7/15/30/60/120 天,连续答对上一级,答错回第一级。
  原先首次答对后要等 7 天才复习,正是"第二天就忘"的成因。
- 每日新词上限(10 分钟 8 个 / 20 分钟 15 个 / 30 分钟 20 个)。
  阶梯第一级是次日,今天引入的新词就是明天的工作量。
- 新词按口语频率发放,不再按字母序 —— A1 从 a.m./ability 变成 no/not/know/just。
- 三个方向按层级轮转:看词(英→中)→ 听词(音→中)→ 想词(中→英)。
  想词题仍是选择题,不要求产出,理解词定位不变,不进升级分母。
- 单词页独立成 tab,首页今日任务卡下方给一张认词入口卡。

## 用法对照

课程 JSON 增加 usage 字段(when/reply/swap/confuse):一个句型用在什么场合、
对方通常怎么答、还能怎么说、跟哪个学过的句型容易混。
知道 How are you? 的意思,不等于知道它不是用来问名字的。

## 复习流

- 当日快闪(recap)独立成队列,不占复习预算,也不计入积压。
- 只发放当日预算内的量,其余保持到期状态等下次,不悄悄丢弃或改期。
- 答错的项隔几题后回来,而不是立刻重问。

测试 296 通过,flutter analyze 干净。

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
shenlei
2026-09-20 23:54:17 +09:00
co-authored by Claude Opus 5
parent 1a10ca88e0
commit 6b42b7abc3
52 changed files with 4044 additions and 82 deletions
@@ -8,6 +8,7 @@ import '../core/models.dart';
import '../core/courses/courses.dart';
import '../core/voice_service.dart';
import 'app_widgets.dart';
import 'usage_card.dart';
List<VocabularyItem>? _cachedEntries;
Map<String, VocabularyItem>? _cachedExactMap;
@@ -454,7 +455,7 @@ class _LexiconLookupSheetState extends State<_LexiconLookupSheet> {
try {
ScaffoldMessenger.of(context).showSnackBar(
SnackBar(
content: Text('将短语 "${phrase.phrase}" 加入复习计划'),
content: Text('加入单词,之后在「单词」里认它'),
duration: const Duration(seconds: 2),
),
);
@@ -469,7 +470,7 @@ class _LexiconLookupSheetState extends State<_LexiconLookupSheet> {
try {
ScaffoldMessenger.of(context).showSnackBar(
SnackBar(
content: Text('将 "${item.word}" 加入复习计划'),
content: Text('加入单词,之后在「单词」里认它'),
duration: const Duration(seconds: 2),
),
);
@@ -663,7 +664,7 @@ class _LexiconLookupSheetState extends State<_LexiconLookupSheet> {
_addedToReview.contains(phrase.phrase.toLowerCase())
? Chip(
label: Text(
'已在复习',
'已在单词',
style: TextStyle(fontSize: 12),
),
avatar: Icon(
@@ -688,7 +689,7 @@ class _LexiconLookupSheetState extends State<_LexiconLookupSheet> {
size: 14,
),
label: const Text(
'复习',
'到单词',
style: TextStyle(fontSize: 12),
),
),
@@ -843,14 +844,15 @@ class _LexiconLookupSheetState extends State<_LexiconLookupSheet> {
tint: AppColors.softGreen,
child: Text('${entry!.example}\n${entry!.exampleMeaning}'),
),
UsageCard(itemId: entry!.id, state: widget.state),
Row(
children: [
Expanded(
child: PrimaryButton(
label:
_addedToReview.contains(entry!.word.toLowerCase())
? '已在复习中'
: '入复习',
? '已在单词表'
: '到单词',
onPressed: () {
_addVocabItemToReview(entry!);
},
+107
View File
@@ -0,0 +1,107 @@
import 'package:flutter/material.dart';
import '../core/app_state.dart';
import '../core/app_theme.dart';
import '../core/courses/courses.dart';
import 'app_widgets.dart';
/// What a taught pattern is *for*, next to what it means. Knowing that
/// `How are you?` means 你怎么样 does not tell a learner that it is not the
/// way to ask someone's name, so the course states the situation, the reply
/// it invites, and the items it is easiest to mix up with.
///
/// The card renders nothing when the pack explains nothing about the item,
/// which is the normal case for a plain word.
class UsageCard extends StatelessWidget {
const UsageCard({
super.key,
required this.itemId,
required this.state,
this.compact = false,
});
final String itemId;
final AppState state;
/// Inside a task step, only the situation and the difference are shown —
/// the learner is mid-exercise, not reading a reference card.
final bool compact;
/// A pattern counts as taught once it has a mastery row, which
/// `AppState` creates when the lesson first presents it.
bool _isTaught(String id) => state.mastery.containsKey(id);
@override
Widget build(BuildContext context) {
final usage = coreUsage(itemId);
if (usage == null) return const SizedBox.shrink();
final lines = <({String label, String text})>[
if (usage.when.isNotEmpty) (label: '什么时候用', text: usage.when),
if (!compact && usage.reply.isNotEmpty) (label: '对方会怎么答', text: usage.reply),
if (!compact && usage.swap.isNotEmpty) (label: '换个说法', text: usage.swap),
// Mid-exercise only the nearest confusion is worth a line; the full
// list belongs on the preview and lookup cards.
for (final other in visibleConfusables(
itemId,
isTaught: _isTaught,
).take(compact ? 1 : 3))
if (other.note.isNotEmpty)
(label: '别和 ${coreItemSpoken(other.id)} 弄混', text: other.note),
];
if (lines.isEmpty) return const SizedBox.shrink();
return SectionCard(
tint: AppColors.warm,
child: SpacedColumn(
spacing: 8,
children: [
for (final line in lines)
RichText(
text: TextSpan(
style: DefaultTextStyle.of(context).style.copyWith(
color: AppColors.warmInk,
height: 1.45,
),
children: [
TextSpan(
text: '${line.label}',
style: const TextStyle(fontWeight: FontWeight.w600),
),
TextSpan(text: line.text),
],
),
),
],
),
);
}
}
/// Every usage note a teaching segment carries, in course order. Shown in the
/// speaking and writing steps, where a learner is about to produce the pattern
/// and the question is which one this situation calls for.
class SegmentUsage extends StatelessWidget {
const SegmentUsage({
super.key,
required this.segmentId,
required this.state,
this.compact = true,
});
final String segmentId;
final AppState state;
final bool compact;
@override
Widget build(BuildContext context) {
final ids = findLessonSegment(segmentId)?.targetItemIds ?? const <String>[];
final explained = [for (final id in ids) if (coreUsage(id) != null) id];
if (explained.isEmpty) return const SizedBox.shrink();
return SpacedColumn(
spacing: 8,
children: [
for (final id in explained)
UsageCard(itemId: id, state: state, compact: compact),
],
);
}
}