content-extractor-rl 0.1.0

RL-based article extraction from HTML using Deep Q-Networks and heuristic fallback
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212

//! HTML parsing and DOM manipulation utilities

// ============================================================================
// FILE: crates/content-extractor-rl/src/html_parser.rs
// ============================================================================

use scraper::{Html, Selector, ElementRef, Element};
use crate::Result;


pub struct HtmlParser;

impl HtmlParser {

    /// Parse HTML string into document
    pub fn parse(html: &str) -> Result<Html> {
        Ok(Html::parse_document(html))
    }

    /// Extract text content from element
    pub fn extract_text(element: ElementRef) -> String {
        element.text().collect::<Vec<_>>().join(" ")
    }

    /// Get XPath-like selector for element
    pub fn get_element_path(element: ElementRef) -> String {
        let mut path = Vec::new();
        let mut current = Some(element);

        while let Some(elem) = current {
            let tag = elem.value().name();

            // Get position among siblings
            let position = elem.prev_siblings()
                .filter(|s| s.value().as_element().is_some_and(|e| e.name() == tag))
                .count() + 1;

            path.push(format!("{}[{}]", tag, position));
            current = elem.parent().and_then(ElementRef::wrap);
        }

        path.reverse();
        format!("/{}", path.join("/"))
    }


    /// Clean HTML by removing script, style, comments, etc.
    pub fn clean_html(html: &str) -> Result<Html> {
        let document = Html::parse_document(html);

        // Create cleaned HTML string (simplified - proper cleaning would modify DOM)
        let mut cleaned = html.to_string();

        // Remove script tags and their content (case-insensitive, multiline)
        let script_re = regex::Regex::new(r"(?is)<script\b[^>]*>.*?</script>").unwrap();
        cleaned = script_re.replace_all(&cleaned, "").to_string();

        // Remove style tags and their content
        let style_re = regex::Regex::new(r"(?is)<style\b[^>]*>.*?</style>").unwrap();
        cleaned = style_re.replace_all(&cleaned, "").to_string();

        // Remove HTML comments
        let comment_re = regex::Regex::new(r"(?s)<!--.*?-->").unwrap();
        cleaned = comment_re.replace_all(&cleaned, "").to_string();

        // Remove inline JavaScript event handlers
        let event_re = regex::Regex::new(r#"\son\w+\s*=\s*["'][^"']*["']"#).unwrap();
        cleaned = event_re.replace_all(&cleaned, "").to_string();

        // Remove JavaScript: protocol links
        let js_protocol_re = regex::Regex::new(r#"javascript:[^"'\s>]*"#).unwrap();
        cleaned = js_protocol_re.replace_all(&cleaned, "").to_string();

        // Remove script tags
        let script_selector = Selector::parse("script").unwrap();
        for element in document.select(&script_selector) {
            if let Some(html) = element.html().get(0..100) {
                cleaned = cleaned.replace(html, "");
            }
        }

        // Remove style tags
        let style_selector = Selector::parse("style").unwrap();
        for element in document.select(&style_selector) {
            if let Some(html) = element.html().get(0..100) {
                cleaned = cleaned.replace(html, "");
            }
        }

        Ok(Html::parse_document(&cleaned))
    }

    /// Get candidate article nodes from document
    pub fn get_candidate_nodes(document: &Html, top_k: usize) -> Vec<ElementRef<'_>> {
        let mut candidates = Vec::new();

        // Try article tags first
        let article_selector = Selector::parse("article").unwrap();
        for element in document.select(&article_selector) {
            candidates.push(element);
        }

        // Try divs
        let div_selector = Selector::parse("div").unwrap();
        for element in document.select(&div_selector) {
            candidates.push(element);
        }

        // Try sections
        let section_selector = Selector::parse("section").unwrap();
        for element in document.select(&section_selector) {
            candidates.push(element);
        }

        candidates.truncate(top_k);
        candidates
    }

    /// Extract paragraphs from element
    pub fn extract_paragraphs(element: ElementRef) -> Vec<String> {
        let p_selector = Selector::parse("p").unwrap();

        element.select(&p_selector)
            .map(|p| Self::extract_text(p).trim().to_string())
            .filter(|text| !text.is_empty())
            .collect()
    }

    /// Get parent element
    pub fn get_parent(element: ElementRef) -> Option<ElementRef> {
        element.parent().and_then(ElementRef::wrap)
    }

    /// Get previous sibling element
    pub fn get_prev_sibling(element: ElementRef) -> Option<ElementRef> {
        element.prev_sibling_element()
    }

    /// Get next sibling element
    pub fn get_next_sibling(element: ElementRef) -> Option<ElementRef> {
        element.next_sibling_element()
    }

    /// Count child elements
    pub fn count_children(element: ElementRef) -> usize {
        element.children().filter(|n| n.value().is_element()).count()
    }

    /// Get tree depth
    pub fn get_tree_depth(document: &Html) -> usize {
        fn depth_recursive(element: ElementRef) -> usize {
            let children: Vec<_> = element.children()
                .filter_map(ElementRef::wrap)
                .collect();

            if children.is_empty() {
                1
            } else {
                1 + children.into_iter()
                    .map(depth_recursive)
                    .max()
                    .unwrap_or(0)
            }
        }

        document.root_element()
            .children()
            .filter_map(ElementRef::wrap)
            .map(depth_recursive)
            .max()
            .unwrap_or(0)
    }

    /// Get node depth in tree
    pub fn get_node_depth(element: ElementRef) -> usize {
        let mut depth = 0;
        let mut current = Some(element);

        while let Some(elem) = current {
            depth += 1;
            current = elem.parent().and_then(ElementRef::wrap);
        }

        depth
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_parse_html() {
        let html = r#"<html><body><p>Hello World</p></body></html>"#;
        let doc = HtmlParser::parse(html).unwrap();
        assert!(doc.root_element().html().contains("Hello World"));
    }

    #[test]
    fn test_extract_paragraphs() {
        let html = r#"
            <article>
                <p>First paragraph.</p>
                <p>Second paragraph.</p>
            </article>
        "#;
        let doc = HtmlParser::parse(html).unwrap();
        let article = doc.select(&Selector::parse("article").unwrap()).next().unwrap();
        let paragraphs = HtmlParser::extract_paragraphs(article);
        assert_eq!(paragraphs.len(), 2);
    }
}