Method extract_first_author_institution

utils/logo_utils.py:340–503

Extract the first author's institution from paper content. Args: paper_content: Text content of the paper (markdown format) Returns: First author's institution if found and matched with available logos

(self, paper_content: str)

        }

    def extract_first_author_institution(self, paper_content: str) -> Optional[str]:
        """
        Extract the first author's institution from paper content.

        Args:
            paper_content: Text content of the paper (markdown format)

        Returns:
            First author's institution if found and matched with available logos
        """
        print("   📝 Looking for first author's institution...")

        # Look for authors section in the beginning of the paper
        lines = paper_content.split('\n')[:100]  # Focus on first 100 lines where authors usually appear

        # Common institution patterns
        institution_patterns = [
            r"(?:University of|University) [\w\s]+",
            r"[\w\s]+ University",
            r"[\w\s]+ Institute of Technology",
            r"[\w\s]+ Institute",
            r"MIT|CMU|UCLA|UCSD|NYU|ETH|EPFL|Stanford|Berkeley|Harvard|Princeton|Oxford|Cambridge",
            r"Google Research|DeepMind|Microsoft Research|Facebook AI Research|OpenAI|NVIDIA Research",
            r"Max Planck Institute",
            r"[\w\s]+ College",
            r"[\w\s]+ Research",
            r"[\w\s]+ Lab",
            r"[\w\s]+ Laboratory"
        ]

        all_pattern = '|'.join(f'({p})' for p in institution_patterns)

        # First pass: Look for the first line with superscript 1 (¹) which typically indicates first author affiliation
        first_institution = None
        for i, line in enumerate(lines):
            # Stop at abstract or introduction
            if 'abstract' in line.lower() or 'introduction' in line.lower():
                break

            # Look for lines with ¹ (first affiliation marker) at the beginning
            if '¹' in line:
                # Use finditer on each pattern individually to find all matches
                # This avoids the issue where re.findall() with alternation only finds non-overlapping matches
                all_matches = []
                for pattern_idx, pattern in enumerate(institution_patterns):
                    for match in re.finditer(pattern, line, re.IGNORECASE):
                        matched_text = match.group()
                        # Filter out very short matches (1-2 chars) but allow valid abbreviations (MIT, NYU, etc.)
                        if matched_text and len(matched_text.strip()) >= 3:
                            all_matches.append((match.start(), matched_text.strip(), pattern_idx))
                
                if all_matches:
                    # Prioritize pattern 0 (University of X) over pattern 1 (X University)
                    pattern0_matches = [m for m in all_matches if m[2] == 0]
                    if pattern0_matches:
                        first_institution = pattern0_matches[0][1]  # Take first pattern 0 match
                        print(f"   🎯 Found first author institution (from affiliation marker): {first_institution}")
                        break