misc-mirrors/article_scraper
1
0
Fork 0
mirror of https://gitlab.com/news-flash/article_scraper.git synced 2025-07-07 16:15:32 +02:00
Code Releases Activity

improve title extraction

Browse source
This commit is contained in:
Jan Lukas Gernert 2023-02-20 02:32:58 +01:00
parent cce912c354
commit 98c06e11f4
7 changed files with 107 additions and 54 deletions
Download patch file Download diff file Expand all files Collapse all files

10
src/full_text_parser/readability/constants.rs → src/constants.rs
View file

@ -11,7 +11,7 @@ pub static BYLINE: Lazy<Regex> = Lazy::new(|| {
});
pub static NORMALIZE: Lazy<Regex> =
Lazy::new(|| Regex::new(r#"/\s{2,}/g"#).expect("NORMALIZE regex"));
pub static TOKENIZE: Lazy<Regex> = Lazy::new(|| Regex::new(r#"/\W+/g"#).expect("TOKENIZE regex"));
pub static TOKENIZE: Lazy<Regex> = Lazy::new(|| Regex::new(r#"\W+"#).expect("TOKENIZE regex"));
pub static UNLIELY_CANDIDATES: Lazy<Regex> = Lazy::new(|| {
Regex::new(r#"/-ad-|ai2html|banner|breadcrumbs|combx|comment|community|cover-wrap|disqus|extra|footer|gdpr|header|legends|menu|related|remark|replies|rss|shoutbox|sidebar|skyscraper|social|sponsor|supplemental|ad-break|agegate|pagination|pager|popup|yom-remote/i"#).expect("UNLIELY_CANDIDATES regex")
});
@ -31,6 +31,14 @@ pub static POSITIVE: Lazy<Regex> = Lazy::new(|| {
pub static NEGATIVE: Lazy<Regex> =
Lazy::new(|| Regex::new(r#"/-ad-|hidden|^hid$| hid$| hid |^hid"#).expect("NEGATIVE regex"));
pub static TITLE_SEPARATOR: Lazy<Regex> =
Lazy::new(|| Regex::new(r#"[-|\\/>»]"#).expect("TITLE_SEPARATOR regex"));
pub static TITLE_CUT_END: Lazy<Regex> =
Lazy::new(|| Regex::new(r#"(.*)[-|\\/>»] .*"#).expect("TITLE_CUT_END regex"));
pub static WORD_COUNT: Lazy<Regex> = Lazy::new(|| Regex::new(r#"\s+"#).expect("WORD_COUNT regex"));
pub static TITLE_CUT_FRONT: Lazy<Regex> =
Lazy::new(|| Regex::new(r#"/[^-|\\/>»]*[-|\\/>»](.*)/gi"#).expect("TITLE_CUT_FRONT regex"));
pub const SCORE_ATTR: &str = "content_score";
pub const MINIMUM_TOPCANDIDATES: usize = 3;
pub const UNLIKELY_ROLES: &[&str] = &[

43
src/full_text_parser/metadata.rs
View file

@ -1,5 +1,5 @@
use super::config::ConfigEntry;
use crate::{article::Article, util::Util};
use crate::{article::Article, constants, util::Util};
use chrono::{DateTime, Utc};
use libxml::xpath::Context;
use log::{debug, warn};
@ -8,14 +8,27 @@ use std::str::FromStr;
pub fn extract(
context: &Context,
config: Option<&ConfigEntry>,
global_config: &ConfigEntry,
global_config: Option<&ConfigEntry>,
article: &mut Article,
) {
if article.title.is_none() {
article.title = extract_title(context, config, global_config).map(|title| {
match escaper::decode_html(&title) {
article.title = extract_title(context, config, global_config)
.map(|title| match escaper::decode_html(&title) {
Ok(escaped_title) => escaped_title,
Err(_error) => title,
})
.map(|title| {
// clean titles that contain separators
if constants::TITLE_SEPARATOR.is_match(&title) {
let new_title = constants::TITLE_CUT_END.replace(&title, "$1");
let word_count = constants::WORD_COUNT.split(&title).count();
if word_count < 3 {
constants::TITLE_CUT_FRONT.replace(&title, "$1").to_string()
} else {
new_title.to_string()
}
} else {
title
}
});
}
@ -38,7 +51,7 @@ pub fn extract(
fn extract_title(
context: &Context,
config: Option<&ConfigEntry>,
global_config: &ConfigEntry,
global_config: Option<&ConfigEntry>,
) -> Option<String> {
// check site specific config
if let Some(config) = config {
@ -51,27 +64,30 @@ fn extract_title(
}
// check global config
if let Some(global_config) = global_config {
for xpath_title in &global_config.xpath_title {
if let Ok(title) = Util::extract_value_merge(context, xpath_title) {
debug!("Article title: '{}'", title);
return Some(title);
}
}
}
// generic meta (readablity)
get_meta(context, "dc:title")
Util::extract_value(context, "//title")
.ok()
.or_else(|| get_meta(context, "dc:title"))
.or_else(|| get_meta(context, "dcterm:title"))
.or_else(|| get_meta(context, "og:title"))
.or_else(|| get_meta(context, "weibo:article:title"))
.or_else(|| get_meta(context, "weibo:webpage:title"))
.or_else(|| get_meta(context, "title"))
.or_else(|| get_meta(context, "twitter:title"))
}
fn extract_author(
context: &Context,
config: Option<&ConfigEntry>,
global_config: &ConfigEntry,
global_config: Option<&ConfigEntry>,
) -> Option<String> {
// check site specific config
if let Some(config) = config {
@ -84,23 +100,26 @@ fn extract_author(
}
// check global config
if let Some(global_config) = global_config {
for xpath_author in &global_config.xpath_author {
if let Ok(author) = Util::extract_value(context, xpath_author) {
debug!("Article author: '{}'", author);
return Some(author);
}
}
}
// generic meta (readablity)
get_meta(context, "dc:creator")
Util::extract_value(context, "//author")
.ok()
.or_else(|| get_meta(context, "dc:creator"))
.or_else(|| get_meta(context, "dcterm:creator"))
.or_else(|| get_meta(context, "author"))
}
fn extract_date(
context: &Context,
config: Option<&ConfigEntry>,
global_config: &ConfigEntry,
global_config: Option<&ConfigEntry>,
) -> Option<DateTime<Utc>> {
// check site specific config
if let Some(config) = config {
@ -117,6 +136,7 @@ fn extract_date(
}
// check global config
if let Some(global_config) = global_config {
for xpath_date in &global_config.xpath_date {
if let Ok(date_string) = Util::extract_value(context, xpath_date) {
debug!("Article date: '{}'", date_string);
@ -127,6 +147,7 @@ fn extract_date(
}
}
}
}
None
}

7
src/full_text_parser/mod.rs
View file

@ -174,7 +174,7 @@ impl FullTextParser {
}
}
metadata::extract(&xpath_ctx, config, global_config, article);
metadata::extract(&xpath_ctx, config, Some(global_config), article);
if article.thumbnail_url.is_none() {
Self::check_for_thumbnail(&xpath_ctx, article);
}
@ -182,7 +182,8 @@ impl FullTextParser {
let found_body = Self::extract_body(&xpath_ctx, root, config, global_config)?;
if !found_body {
if let Err(error) = Readability::extract_body(document, root) {
if let Err(error) = Readability::extract_body(document, root, article.title.as_deref())
{
log::error!("Both ftr and readability failed to find content: {}", error);
return Err(error);
}
@ -246,7 +247,7 @@ impl FullTextParser {
let html = Self::download(url, client, headers).await?;
let document = Self::parse_html(&html, config, global_config)?;
let xpath_ctx = Self::get_xpath_ctx(&document)?;
metadata::extract(&xpath_ctx, config, global_config, article);
metadata::extract(&xpath_ctx, config, Some(global_config), article);
Self::check_for_thumbnail(&xpath_ctx, article);
Self::strip_junk(&xpath_ctx, config, global_config, url);
Self::extract_body(&xpath_ctx, root, config, global_config)?;

34
src/full_text_parser/readability/mod.rs
View file

@ -1,4 +1,3 @@
mod constants;
mod state;
#[cfg(test)]
@ -10,11 +9,16 @@ use libxml::tree::{node, Document, Node, NodeType};
use self::state::State;
use super::error::FullTextParserError;
use crate::constants;
pub struct Readability;
impl Readability {
pub fn extract_body(document: Document, root: &mut Node) -> Result<bool, FullTextParserError> {
pub fn extract_body(
document: Document,
root: &mut Node,
title: Option<&str>,
) -> Result<bool, FullTextParserError> {
node::set_node_rc_guard(6);
let mut state = State::default();
@ -49,7 +53,9 @@ impl Readability {
continue;
}
if state.should_remove_title_header && Self::header_duplicates_title(node_ref) {
if state.should_remove_title_header
&& Self::header_duplicates_title(node_ref, title)
{
state.should_remove_title_header = false;
node = Self::remove_and_next(node_ref);
continue;
@ -278,7 +284,8 @@ impl Readability {
constants::MINIMUM_TOPCANDIDATES,
);
for ancestor in alternative_candidate_ancestors.iter().take(tmp) {
lists_containing_this_ancestor += if ancestor == parent { 1 } else { 0 };
lists_containing_this_ancestor +=
if ancestor == parent { 1 } else { 0 };
}
if lists_containing_this_ancestor >= constants::MINIMUM_TOPCANDIDATES {
@ -668,13 +675,18 @@ impl Readability {
// Check if this node is an H1 or H2 element whose content is mostly
// the same as the article title.
fn header_duplicates_title(node: &Node) -> bool {
fn header_duplicates_title(node: &Node, title: Option<&str>) -> bool {
let name = node.get_name().to_lowercase();
if name != "h1" && name != "h2" {
return false;
}
let heading = Self::get_inner_text(node, false);
Self::text_similarity(&heading, "Get your Frontend JavaScript Code Covered") > 0.75
if let Some(title) = title {
Self::text_similarity(&heading, title) > 0.75
} else {
false
}
}
fn get_inner_text(node: &Node, normalize_spaces: bool) -> String {
@ -695,18 +707,12 @@ impl Readability {
return 0.0;
}
let tokens_b_total: f64 = tokens_b
.iter()
.map(|t| t.len())
.fold(0.0, |a, b| a + b as f64);
let tokens_b_total = tokens_b.join(" ").len() as f64;
let uniq_tokens_b = tokens_b
.into_iter()
.filter(|token| !tokens_a.iter().any(|t| t == token))
.collect::<Vec<_>>();
let uniq_tokens_b_total: f64 = uniq_tokens_b
.iter()
.map(|t| t.len())
.fold(0.0, |a, b| a + b as f64);
let uniq_tokens_b_total = uniq_tokens_b.join(" ").len() as f64;
let distance_b = uniq_tokens_b_total / tokens_b_total;
1.0 - distance_b

29
src/full_text_parser/readability/tests.rs
View file

@ -1,17 +1,30 @@
use libxml::tree::{Document, Node};
use libxml::{
tree::{Document, Node},
xpath::Context,
};
use reqwest::Url;
use crate::full_text_parser::config::ConfigEntry;
use crate::{
article::Article,
full_text_parser::{config::ConfigEntry, metadata},
};
async fn prepare(html: &str, url: &Url) -> Document {
async fn prepare(html: &str, url: &Url) -> (Document, Context, Article) {
let empty_config = ConfigEntry::default();
let document = crate::FullTextParser::parse_html(html, None, &empty_config).unwrap();
let xpath_ctx = crate::FullTextParser::get_xpath_ctx(&document).unwrap();
crate::FullTextParser::strip_junk(&xpath_ctx, None, &empty_config, url);
document
let article = Article {
title: None,
author: None,
url: url.clone(),
date: None,
thumbnail_url: None,
document: None,
};
(document, xpath_ctx, article)
}
#[tokio::test]
async fn test_1() {
let _ = env_logger::builder().is_test(true).try_init();
@ -19,9 +32,11 @@ async fn test_1() {
let html = std::fs::read_to_string(r"./resources/tests/readability-test-1.html")
.expect("Failed to read HTML");
let url = Url::parse("http://google.com").unwrap();
let document = prepare(&html, &url).await;
let (document, xpath_ctx, mut article) = prepare(&html, &url).await;
let mut root = Node::new("article", None, &document).unwrap();
super::Readability::extract_body(document, &mut root).unwrap();
metadata::extract(&xpath_ctx, None, None, &mut article);
super::Readability::extract_body(document, &mut root, article.title.as_deref()).unwrap();
}

3
src/images/mod.rs
View file

@ -145,7 +145,8 @@ impl ImageDownloader {
}
let small_image_base64 = base64::engine::general_purpose::STANDARD.encode(&small_image);
let big_image_base64 = big_image.map(|img| base64::engine::general_purpose::STANDARD.encode(img));
let big_image_base64 =
big_image.map(|img| base64::engine::general_purpose::STANDARD.encode(img));
let small_image_string =
format!("data:{};base64,{}", content_type_small, small_image_base64);
let big_image_string = match big_image_base64 {

1
src/lib.rs
View file

@ -1,4 +1,5 @@
mod article;
mod constants;
mod error;
mod full_text_parser;
pub mod images;