/**
* Topic Extractor - Detect topic boundaries and page ranges
*/
import type { TopicPageRange } from '../types/ngss.js';
import { PDFReader, parsePageContent } from './pdf-reader.js';
import { PatternExtractor } from './pattern-extractor.js';
export class TopicExtractor {
private pdfReader: PDFReader;
private patternExtractor: PatternExtractor;
constructor() {
this.pdfReader = new PDFReader();
this.patternExtractor = new PatternExtractor();
}
async extractTopicPages(
pdfPath: string,
topicPattern: string
): Promise<TopicPageRange | null> {
// Get all content to scan for topics
const content = await this.pdfReader.extractAll(pdfPath);
const pages = parsePageContent(content, 'all');
// Find pages matching topic pattern
const matchingPages: number[] = [];
for (const page of pages) {
if (this.matchesTopic(page.content, topicPattern)) {
matchingPages.push(page.pageNumber);
}
}
if (matchingPages.length === 0) {
return null;
}
// Determine page range
const start_page = Math.min(...matchingPages);
const end_page = Math.max(...matchingPages);
// Extract standard codes from this range
const codes = await this.patternExtractor.extractStandardCodes(
pdfPath,
`${start_page}-${end_page}`
);
return {
topic: topicPattern,
start_page,
end_page,
standard_codes: codes.map((c) => c.code)
};
}
private matchesTopic(content: string, pattern: string): boolean {
// Case-insensitive fuzzy matching
const normalizedContent = content.toLowerCase();
const normalizedPattern = pattern.toLowerCase();
// Direct match
if (normalizedContent.includes(normalizedPattern)) {
return true;
}
// Header pattern: "MS.Topic"
const headerRegex = new RegExp(`MS\\.${pattern}`, 'i');
if (headerRegex.test(content)) {
return true;
}
// Word-based similarity (simple)
const patternWords = normalizedPattern.split(/\s+/);
const matchCount = patternWords.filter((word) =>
normalizedContent.includes(word)
).length;
// At least 70% of pattern words must match
return matchCount / patternWords.length >= 0.7;
}
async listAllTopics(pdfPath: string): Promise<TopicPageRange[]> {
const content = await this.pdfReader.extractAll(pdfPath);
const pages = parsePageContent(content, 'all');
const topics = new Map<string, number[]>();
for (const page of pages) {
// Look for topic headers like "MS.Chemical Reactions"
// Fixed regex to capture full topic name including capital letters
const headerMatch = page.content.match(/MS\.([A-Z][A-Za-z\s&-]+)/);
if (headerMatch && headerMatch[1]) {
const topic = headerMatch[1].trim();
const existing = topics.get(topic) || [];
existing.push(page.pageNumber);
topics.set(topic, existing);
}
}
// Convert to TopicPageRange array
const results: TopicPageRange[] = [];
for (const [topic, pages] of topics.entries()) {
const start_page = Math.min(...pages);
const end_page = Math.max(...pages);
const codes = await this.patternExtractor.extractStandardCodes(
pdfPath,
`${start_page}-${end_page}`
);
results.push({
topic,
start_page,
end_page,
standard_codes: codes.map((c) => c.code)
});
}
return results.sort((a, b) => a.start_page - b.start_page);
}
}
