(
paper: Dict[str, Any],
user_profile: Dict[str, Any],
parsed_pdf: Optional[Dict[str, Any]] = None,
pdf_error: Optional[str] = None,
response_language: str = "zh",
)
| 3876 | |
| 3877 | |
| 3878 | def build_heuristic_report_payload( |
| 3879 | paper: Dict[str, Any], |
| 3880 | user_profile: Dict[str, Any], |
| 3881 | parsed_pdf: Optional[Dict[str, Any]] = None, |
| 3882 | pdf_error: Optional[str] = None, |
| 3883 | response_language: str = "zh", |
| 3884 | ) -> Dict[str, Any]: |
| 3885 | language = _normalize_response_language(response_language) |
| 3886 | is_en = language == "en" |
| 3887 | sections = dict((parsed_pdf or {}).get("sections") or {}) |
| 3888 | parsed_source_kind = _clean_text((parsed_pdf or {}).get("source_kind")).lower() or ("pdf" if parsed_pdf else "abstract") |
| 3889 | if _prefer_candidate_abstract(paper.get("abstract"), (parsed_pdf or {}).get("abstract")): |
| 3890 | abstract = _clean_abstract_text((parsed_pdf or {}).get("abstract")) |
| 3891 | else: |
| 3892 | abstract = _clean_abstract_text(_first_non_empty(paper.get("abstract"), (parsed_pdf or {}).get("abstract"))) |
| 3893 | introduction = _truncate_text(sections.get("introduction"), MAX_SECTION_CHARS) |
| 3894 | method = _truncate_text(sections.get("method"), MAX_SECTION_CHARS) |
| 3895 | results = _truncate_text(sections.get("results"), MAX_SECTION_CHARS) |
| 3896 | discussion = _truncate_text(sections.get("discussion"), MAX_SECTION_CHARS) |
| 3897 | conclusion = _truncate_text(sections.get("conclusion"), MAX_SECTION_CHARS) |
| 3898 | retrieved_evidence = _retrieve_report_evidence(paper, user_profile, parsed_pdf) |
| 3899 | |
| 3900 | # 优先使用 PDF 切块后的语义检索证据，其次回退到章节摘要/abstract 启发式 |
| 3901 | bg_from_evidence = " ".join(_collect_evidence_sentences(retrieved_evidence, "background", limit=2)) |
| 3902 | bg_from_pdf = " ".join(_pick_sentences(introduction, limit=2)) if introduction else "" |
| 3903 | bg_from_abstract = _infer_section_from_abstract(abstract, "background") |
| 3904 | research_background = _first_non_empty(bg_from_evidence, bg_from_pdf, bg_from_abstract) |
| 3905 | if not research_background: |
| 3906 | research_background = ( |
| 3907 | "Start with the original abstract and introduction to understand the research background." |
| 3908 | if is_en |
| 3909 | else "建议先查看原文摘要和引言部分以了解研究背景。" |
| 3910 | ) |
| 3911 | |
| 3912 | method_from_evidence = " ".join( |
| 3913 | _collect_evidence_sentences( |
| 3914 | retrieved_evidence, |
| 3915 | "method", |
| 3916 | limit=2, |
| 3917 | cues=CONTRIBUTION_CUES, |
| 3918 | ) |
| 3919 | ) |
| 3920 | method_from_pdf = " ".join(_pick_sentences(method, limit=2, cues=CONTRIBUTION_CUES)) if method else "" |
| 3921 | method_from_abstract = _infer_section_from_abstract(abstract, "method") |
| 3922 | core_method = _first_non_empty(method_from_evidence, method_from_pdf, method_from_abstract) |
| 3923 | if not core_method: |
| 3924 | core_method = ( |
| 3925 | "Method details were not extracted reliably; focus on the Method / Approach section in the original paper." |
| 3926 | if is_en |
| 3927 | else "方法细节建议重点查看原文的 Method / Approach 部分。" |
| 3928 | ) |
| 3929 | |
| 3930 | results_from_evidence = " ".join( |
| 3931 | _collect_evidence_sentences( |
| 3932 | retrieved_evidence, |
| 3933 | "results", |
| 3934 | limit=2, |
| 3935 | cues=RESULT_CUES, |
no test coverage detected