从摘要中推断特定部分的内容 Args: abstract: 论文摘要 section_type: 要推断的部分类型 (background, method, results) Returns: 推断出的部分内容
(abstract: str, section_type: str)
| 3757 | |
| 3758 | def _infer_section_from_abstract(abstract: str, section_type: str) -> str: |
| 3759 | """ |
| 3760 | 从摘要中推断特定部分的内容 |
| 3761 | |
| 3762 | Args: |
| 3763 | abstract: 论文摘要 |
| 3764 | section_type: 要推断的部分类型 (background, method, results) |
| 3765 | |
| 3766 | Returns: |
| 3767 | 推断出的部分内容 |
| 3768 | """ |
| 3769 | if not abstract: |
| 3770 | return "" |
| 3771 | |
| 3772 | sentences = _split_sentences(abstract) |
| 3773 | if not sentences: |
| 3774 | return "" |
| 3775 | |
| 3776 | # 根据部分类型选择不同的线索词 |
| 3777 | section_cues = { |
| 3778 | "background": ("we address", "we tackle", "challenge", "problem", "limitation", "however", "despite", "existing", "prior work"), |
| 3779 | "method": ("we propose", "we present", "we introduce", "we develop", "our approach", "our method", "framework", "model", "architecture"), |
| 3780 | "results": ("outperform", "improve", "achieve", "demonstrate", "results show", "experimental", "evaluation", "gain", "better than", "superior"), |
| 3781 | } |
| 3782 | |
| 3783 | cues = section_cues.get(section_type, ()) |
| 3784 | |
| 3785 | # 优先选择包含线索词的句子 |
| 3786 | matched = [] |
| 3787 | for sentence in sentences: |
| 3788 | lowered = sentence.lower() |
| 3789 | if any(cue in lowered for cue in cues): |
| 3790 | matched.append(sentence) |
| 3791 | |
| 3792 | if matched: |
| 3793 | return " ".join(matched[:2]) |
| 3794 | |
| 3795 | # 如果没有匹配的，根据位置返回 |
| 3796 | if section_type == "background": |
| 3797 | return " ".join(sentences[:2]) if len(sentences) >= 2 else sentences[0] if sentences else "" |
| 3798 | elif section_type == "results": |
| 3799 | return " ".join(sentences[-2:]) if len(sentences) >= 2 else sentences[-1] if sentences else "" |
| 3800 | else: # method |
| 3801 | mid = len(sentences) // 2 |
| 3802 | return " ".join(sentences[mid:mid+2]) if len(sentences) > 2 else sentences[mid] if sentences else "" |
| 3803 | |
| 3804 | |
| 3805 | def _extract_heuristic_keywords( |
| 3806 | paper: Dict[str, Any], |
no test coverage detected