math_cms/app/Services/OCRStructureParser.php

<?php

namespace App\Services;

class OCRStructureParser
{
    /**
     * 解析阿里云OCR返回的碎片化blocks，重构为题目结构
     */
    public function parse(array $ocrData): array
    {
        // 递归解析data字段（防止嵌套字符串）
        $data = $this->parseNestedJson($ocrData);

        // 提取所有文本块
        $blocks = $this->extractAllTextBlocks($data);

        // 根据题号分组
        $questionGroups = $this->groupBlocksByQuestionNumber($blocks);

        // 组装每道题的结构
        $structuredQuestions = $this->assembleQuestions($questionGroups);

        return $structuredQuestions;
    }

    /**
     * 递归解析嵌套的JSON字符串
     */
    private function parseNestedJson($data)
    {
        if (is_string($data)) {
            $decoded = json_decode($data, true);
            if (json_last_error() === JSON_ERROR_NONE) {
                return $decoded;
            }
            return $data;
        }

        // 递归处理嵌套结构
        if (is_array($data)) {
            foreach ($data as $key => $value) {
                $data[$key] = $this->parseNestedJson($value);
            }
        }

        return $data;
    }

    /**
     * 提取所有文本块
     */
    private function extractAllTextBlocks(array $data): array
    {
        $blocks = [];

        if (!isset($data['data']['page_list'])) {
            return $blocks;
        }

        foreach ($data['data']['page_list'] as $page) {
            if (!isset($page['answer_list'])) {
                continue;
            }

            foreach ($page['answer_list'] as $item) {
                if (!isset($item['content_list_info'])) {
                    continue;
                }

                foreach ($item['content_list_info'] as $content) {
                    $text = $content['text'] ?? '';
                    $text = trim($text);

                    if ($text !== '') {
                        $blocks[] = [
                            'text' => $text,
                            'ids' => $item['ids'] ?? [],
                            'position' => $content['pos'] ?? null,
                            'confidence' => $content['confidence'] ?? null,
                            'doc_index' => $content['doc_index'] ?? 1,
                            'is_multipage' => $item['is_multipage'] ?? false
                        ];
                    }
                }
            }
        }

        return $blocks;
    }

    /**
     * 根据题号将文本块分组
     */
    private function groupBlocksByQuestionNumber(array $blocks): array
    {
        $questionNumbers = [];
        $groups = [];

        // 第一步：识别所有题号
        foreach ($blocks as $index => $block) {
            $text = $block['text'];

            // 匹配题号格式：1. 1、 1)、(1) ①等
            if (preg_match('/^\s*(\d+)\s*[\.\、\)\)]/', $text, $matches)) {
                $questionNum = (int)$matches[1];
                $y = $this->getBlockCenterY($block);

                $questionNumbers[] = [
                    'index' => $index,
                    'number' => $questionNum,
                    'text' => $text,
                    'y' => $y
                ];
            }
        }

        // 按题号排序
        usort($questionNumbers, function($a, $b) {
            return $a['number'] <=> $b['number'];
        });

        // 第二步：根据题号Y坐标分组
        foreach ($questionNumbers as $i => $currentQN) {
            $nextQN = $questionNumbers[$i + 1] ?? null;

            $yStart = $currentQN['y'];
            $yEnd = $nextQN ? $nextQN['y'] : PHP_INT_MAX;

            // 收集这个题号范围内的所有blocks
            $groupBlocks = [];
            foreach ($blocks as $block) {
                $blockY = $this->getBlockCenterY($block);
                if ($blockY >= $yStart && ($blockY < $yEnd)) {
                    $groupBlocks[] = $block;
                }
            }

            $groups[] = [
                'question_number' => $currentQN['number'],
                'question_text' => $currentQN['text'],
                'blocks' => $groupBlocks,
                'y_range' => ['start' => $yStart, 'end' => $yEnd]
            ];
        }

        return $groups;
    }

    /**
     * 组装每道题的结构
     */
    private function assembleQuestions(array $questionGroups): array
    {
        $questions = [];

        foreach ($questionGroups as $group) {
            $question = [
                'q' => $group['question_number'],
                'text' => '',
                'options' => [],
                'blocks' => $group['blocks']
            ];

            $questionText = [];
            $options = [];
            $questionNumbers = [];

            foreach ($group['blocks'] as $block) {
                $text = $block['text'];

                // 识别题号
                if (preg_match('/^\s*(\d+)\s*[\.\、\)\)]/', $text, $matches)) {
                    continue; // 跳过题号本身
                }

                // 识别选择题选项
                if (preg_match('/^([A-Da-d])[\.\、]?/', $text, $optionMatch)) {
                    $optionLetter = strtoupper($optionMatch[1]);
                    $options[$optionLetter] = substr($text, 2);
                } elseif (in_array(substr($text, 0, 1), ['A', 'B', 'C', 'D'])) {
                    // 单字母选项
                    $options[substr($text, 0, 1)] = substr($text, 1);
                } else {
                    // 题干或其他内容
                    $questionText[] = $text;
                }
            }

            // 合并题干文本
            $question['text'] = implode(' ', array_filter($questionText));

            // 处理选项：如果有多个选项连在一起，需要拆分
            if (count($options) === 0 && preg_match('/([A-Da-d])/', $question['text'])) {
                $options = $this->splitMergedOptions($question['text']);
            }

            $question['options'] = $options;
            $questions[] = $question;
        }

        return $questions;
    }

    /**
     * 拆分连在一起的选项
     */
    private function splitMergedOptions(string $text): array
    {
        $options = [];

        // 匹配选项模式
        if (preg_match_all('/([A-Da-d])[\.\、]?([^A-D]*)/', $text, $matches, PREG_SET_ORDER)) {
            for ($i = 0; $i < count($matches[1]); $i++) {
                $letter = strtoupper($matches[1][$i]);
                $content = trim($matches[2][$i]);
                if ($content) {
                    $options[$letter] = $content;
                }
            }
        }

        return $options;
    }

    /**
     * 获取block的Y坐标中心
     */
    private function getBlockCenterY(array $block): int
    {
        if (!isset($block['position']) || empty($block['position'])) {
            return 0;
        }

        $yValues = [];
        foreach ($block['position'] as $point) {
            if (isset($point['y'])) {
                $yValues[] = $point['y'];
            }
        }

        if (empty($yValues)) {
            return 0;
        }

        return (int)(array_sum($yValues) / count($yValues));
    }
}