(html_text: str)
| 1098 | |
| 1099 | |
| 1100 | def _extract_homepage_sections(html_text: str) -> Dict[str, Any]: |
| 1101 | title = _strip_html(_extract_first_group(r"<title[^>]*>(.*?)</title>", html_text)) |
| 1102 | description = _strip_html( |
| 1103 | _extract_first_group( |
| 1104 | r'<meta[^>]+name=["\']description["\'][^>]+content=["\'](.*?)["\']', |
| 1105 | html_text, |
| 1106 | ) |
| 1107 | ) |
| 1108 | keywords_raw = _strip_html( |
| 1109 | _extract_first_group( |
| 1110 | r'<meta[^>]+name=["\']keywords["\'][^>]+content=["\'](.*?)["\']', |
| 1111 | html_text, |
| 1112 | ) |
| 1113 | ) |
| 1114 | keywords = _dedupe_preserve_order(re.split(r"\s*,\s*", keywords_raw)) if keywords_raw else [] |
| 1115 | |
| 1116 | sections = { |
| 1117 | "research_interests": [], |
| 1118 | "research": [], |
| 1119 | "about": [], |
| 1120 | "projects": [], |
| 1121 | } |
| 1122 | lines = _html_to_text_lines(html_text) |
| 1123 | active_section = "" |
| 1124 | remaining_lines = 0 |
| 1125 | |
| 1126 | for line in lines: |
| 1127 | section_name = _classify_homepage_section(line) |
| 1128 | if section_name: |
| 1129 | active_section = section_name |
| 1130 | remaining_lines = 6 if section_name in {"research_interests", "research"} else 4 |
| 1131 | inline_parts = re.split(r"[:：]\s*", line, maxsplit=1) |
| 1132 | if len(inline_parts) == 2 and inline_parts[1].strip(): |
| 1133 | sections[section_name].append(inline_parts[1].strip()) |
| 1134 | remaining_lines -= 1 |
| 1135 | continue |
| 1136 | |
| 1137 | if active_section and remaining_lines > 0: |
| 1138 | sections[active_section].append(line) |
| 1139 | remaining_lines -= 1 |
| 1140 | continue |
| 1141 | |
| 1142 | active_section = "" |
| 1143 | |
| 1144 | return { |
| 1145 | "title": title, |
| 1146 | "description": description, |
| 1147 | "keywords": keywords, |
| 1148 | "sections": {key: _dedupe_preserve_order(value) for key, value in sections.items()}, |
| 1149 | } |
| 1150 | |
| 1151 | |
| 1152 | def parse_research_homepage(homepage_url: str, use_llm: bool = True) -> Dict[str, Any]: |
no test coverage detected