镜像自地址
https://github.com/binary-husky/gpt_academic.git
已同步 2025-12-08 23:46:48 +00:00
stage academic conversation
这个提交包含在：
binary-husky
@@ -0,0 +1,76 @@
|
||||
# ADS query optimization prompt
|
||||
ADSABS_QUERY_PROMPT = """Analyze and optimize the following query for NASA ADS search.
|
||||
If the query is not related to astronomy, astrophysics, or physics, return <query>none</query>.
|
||||
If the query contains non-English terms, translate them to English first.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized ADS search query.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on the core research topic for the search query.
|
||||
|
||||
Relevant research areas for ADS:
|
||||
- Astronomy and astrophysics
|
||||
- Physics (theoretical and experimental)
|
||||
- Space science and exploration
|
||||
- Planetary science
|
||||
- Cosmology
|
||||
- Astrobiology
|
||||
- Related instrumentation and methods
|
||||
|
||||
Available search fields and filters:
|
||||
1. Basic fields:
|
||||
- title: Search in title (title:"term")
|
||||
- abstract: Search in abstract (abstract:"term")
|
||||
- author: Search for author names (author:"lastname, firstname")
|
||||
- year: Filter by year (year:2020-2023)
|
||||
- bibstem: Search by journal abbreviation (bibstem:ApJ)
|
||||
|
||||
2. Boolean operators:
|
||||
- AND
|
||||
- OR
|
||||
- NOT
|
||||
- (): Group terms
|
||||
- "": Exact phrase match
|
||||
|
||||
3. Special filters:
|
||||
- citations(identifier:paper): Papers citing a specific paper
|
||||
- references(identifier:paper): References of a specific paper
|
||||
- citation_count: Filter by citation count
|
||||
- database: Filter by database (database:astronomy)
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Black holes in galaxy centers after 2020"
|
||||
<query>title:"black hole" AND abstract:"galaxy center" AND year:2020-</query>
|
||||
|
||||
2. Query: "Papers by Neil deGrasse Tyson about exoplanets"
|
||||
<query>author:"Tyson, Neil deGrasse" AND title:exoplanet</query>
|
||||
|
||||
3. Query: "Most cited papers about dark matter in ApJ"
|
||||
<query>title:"dark matter" AND bibstem:ApJ AND citation_count:[100 TO *]</query>
|
||||
|
||||
4. Query: "Latest research on diabetes treatment"
|
||||
<query>none</query>
|
||||
|
||||
5. Query: "Machine learning for galaxy classification"
|
||||
<query>title:("machine learning" OR "deep learning") AND (title:galaxy OR abstract:galaxy) AND abstract:classification</query>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<query>Provide the optimized ADS search query using appropriate fields and operators, or "none" if not relevant</query>"""
|
||||
|
||||
# System prompt
|
||||
ADSABS_QUERY_SYSTEM_PROMPT = """You are an expert at crafting NASA ADS search queries.
|
||||
Your task is to:
|
||||
1. First determine if the query is relevant to astronomy, astrophysics, or physics research
|
||||
2. If relevant, optimize the natural language query for the ADS API
|
||||
3. If not relevant, return "none" to indicate the query should be handled by other databases
|
||||
|
||||
Focus on creating precise queries that will return relevant astronomical and physics literature.
|
||||
Always generate English search terms regardless of the input language.
|
||||
Consider using field-specific search terms and appropriate filters to improve search accuracy.
|
||||
|
||||
Remember: ADS is specifically for astronomy, astrophysics, and physics research.
|
||||
Medical, biological, or general research queries should return "none"."""
|
||||
@@ -0,0 +1,341 @@
|
||||
# Basic type analysis prompt
|
||||
ARXIV_TYPE_PROMPT = """Analyze the research query and determine if arXiv search is needed and its type.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task 1: Determine if this query requires arXiv search
|
||||
- arXiv is suitable for:
|
||||
* Computer science and AI/ML
|
||||
* Physics and mathematics
|
||||
* Quantitative biology and finance
|
||||
* Electrical engineering
|
||||
* Recent preprints in these fields
|
||||
- arXiv is NOT needed for:
|
||||
* Medical research (unless ML/AI applications)
|
||||
* Social sciences
|
||||
* Business studies
|
||||
* Humanities
|
||||
* Industry reports
|
||||
|
||||
Task 2: If arXiv search is needed, determine the most appropriate search type
|
||||
Available types:
|
||||
1. basic: Keyword-based search across all fields
|
||||
- For specific technical queries
|
||||
- When looking for particular methods or applications
|
||||
2. category: Category-based search within specific fields
|
||||
- For broad topic exploration
|
||||
- When surveying a research area
|
||||
3. none: arXiv search not needed for this query
|
||||
- When topic is outside arXiv's scope
|
||||
- For non-technical or clinical research
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "BERT transformer architecture"
|
||||
<search_type>basic</search_type>
|
||||
|
||||
2. Query: "latest developments in machine learning"
|
||||
<search_type>category</search_type>
|
||||
|
||||
3. Query: "COVID-19 clinical trials"
|
||||
<search_type>none</search_type>
|
||||
|
||||
4. Query: "psychological effects of social media"
|
||||
<search_type>none</search_type>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<search_type>Choose either 'basic', 'category', or 'none'</search_type>"""
|
||||
|
||||
# Query optimization prompt
|
||||
ARXIV_QUERY_PROMPT = """Optimize the following query for arXiv search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized arXiv search query using boolean operators and field tags.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on the core research topic for the search query.
|
||||
|
||||
Available field tags:
|
||||
- ti: Search in title
|
||||
- abs: Search in abstract
|
||||
- au: Search for author
|
||||
- all: Search in all fields (default)
|
||||
|
||||
Boolean operators:
|
||||
- AND: Both terms must appear
|
||||
- OR: Either term can appear
|
||||
- NOT: Exclude terms
|
||||
- (): Group terms
|
||||
- "": Exact phrase match
|
||||
|
||||
Examples:
|
||||
|
||||
1. Natural query: "Recent papers about transformer models by Vaswani"
|
||||
<query>ti:"transformer model" AND au:Vaswani AND year:[2017 TO 2024]</query>
|
||||
|
||||
2. Natural query: "Deep learning for computer vision, excluding surveys"
|
||||
<query>ti:(deep learning AND "computer vision") NOT (ti:survey OR ti:review)</query>
|
||||
|
||||
3. Natural query: "Attention mechanism in language models"
|
||||
<query>ti:(attention OR "attention mechanism") AND abs:"language model"</query>
|
||||
|
||||
4. Natural query: "GANs or generative adversarial networks for image generation"
|
||||
<query>(ti:GAN OR ti:"generative adversarial network") AND abs:"image generation"</query>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<query>Provide the optimized search query using appropriate operators and tags</query>
|
||||
|
||||
Note:
|
||||
- Use quotes for exact phrases
|
||||
- Combine multiple conditions with boolean operators
|
||||
- Consider both title and abstract for important concepts
|
||||
- Include author names when relevant
|
||||
- Use parentheses for complex logical groupings"""
|
||||
|
||||
# Sort parameters prompt
|
||||
ARXIV_SORT_PROMPT = """Determine optimal sorting parameters for the research query.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Select the most appropriate sorting parameters to help users find the most relevant papers.
|
||||
|
||||
Available sorting options:
|
||||
|
||||
1. Sort by:
|
||||
- relevance: Best match to query terms (default)
|
||||
- lastUpdatedDate: Most recently updated papers
|
||||
- submittedDate: Most recently submitted papers
|
||||
|
||||
2. Sort order:
|
||||
- descending: Newest/Most relevant first (default)
|
||||
- ascending: Oldest/Least relevant first
|
||||
|
||||
3. Result limit:
|
||||
- Minimum: 10 papers
|
||||
- Maximum: 50 papers
|
||||
- Recommended: 20-30 papers for most queries
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Latest developments in transformer models"
|
||||
<sort_by>submittedDate</sort_by>
|
||||
<sort_order>descending</sort_order>
|
||||
<limit>30</limit>
|
||||
|
||||
2. Query: "Foundational papers about neural networks"
|
||||
<sort_by>relevance</sort_by>
|
||||
<sort_order>descending</sort_order>
|
||||
<limit>20</limit>
|
||||
|
||||
3. Query: "Evolution of deep learning since 2012"
|
||||
<sort_by>submittedDate</sort_by>
|
||||
<sort_order>ascending</sort_order>
|
||||
<limit>50</limit>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<sort_by>Choose: relevance, lastUpdatedDate, or submittedDate</sort_by>
|
||||
<sort_order>Choose: ascending or descending</sort_order>
|
||||
<limit>Suggest number between 10-50</limit>
|
||||
|
||||
Note:
|
||||
- Choose relevance for specific technical queries
|
||||
- Use lastUpdatedDate for tracking paper revisions
|
||||
- Use submittedDate for following recent developments
|
||||
- Consider query context when setting the limit"""
|
||||
|
||||
# System prompts for each task
|
||||
ARXIV_TYPE_SYSTEM_PROMPT = """You are an expert at analyzing academic queries.
|
||||
Your task is to determine whether the query is better suited for keyword search or category-based search.
|
||||
Consider the query's specificity, scope, and intended search area when making your decision.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
ARXIV_QUERY_SYSTEM_PROMPT = """You are an expert at crafting arXiv search queries.
|
||||
Your task is to optimize natural language queries using boolean operators and field tags.
|
||||
Focus on creating precise, targeted queries that will return the most relevant results.
|
||||
Always generate English search terms regardless of the input language."""
|
||||
|
||||
ARXIV_CATEGORIES_SYSTEM_PROMPT = """You are an expert at arXiv category classification.
|
||||
Your task is to select the most relevant categories for the given research query.
|
||||
Consider both primary and related interdisciplinary categories, while maintaining focus on the main research area.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
ARXIV_SORT_SYSTEM_PROMPT = """You are an expert at optimizing search results.
|
||||
Your task is to determine the best sorting parameters based on the query context.
|
||||
Consider the user's likely intent and temporal aspects of the research topic.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
# 添加新的搜索提示词
|
||||
ARXIV_SEARCH_PROMPT = """Analyze and optimize the research query for arXiv search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized arXiv search query.
|
||||
|
||||
Available search options:
|
||||
1. Basic search with field tags:
|
||||
- ti: Search in title
|
||||
- abs: Search in abstract
|
||||
- au: Search for author
|
||||
Example: "ti:transformer AND abs:attention"
|
||||
|
||||
2. Category-based search:
|
||||
- Use specific arXiv categories
|
||||
Example: "cat:cs.AI AND neural networks"
|
||||
|
||||
3. Date range:
|
||||
- Specify date range using submittedDate
|
||||
Example: "deep learning AND submittedDate:[20200101 TO 20231231]"
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Recent papers about transformer models by Vaswani"
|
||||
<search_criteria>
|
||||
<query>ti:"transformer model" AND au:Vaswani AND submittedDate:[20170101 TO 99991231]</query>
|
||||
<categories>cs.CL, cs.AI, cs.LG</categories>
|
||||
<sort_by>submittedDate</sort_by>
|
||||
<sort_order>descending</sort_order>
|
||||
<limit>30</limit>
|
||||
</search_criteria>
|
||||
|
||||
2. Query: "Latest developments in computer vision"
|
||||
<search_criteria>
|
||||
<query>cat:cs.CV AND submittedDate:[20220101 TO 99991231]</query>
|
||||
<categories>cs.CV, cs.AI, cs.LG</categories>
|
||||
<sort_by>submittedDate</sort_by>
|
||||
<sort_order>descending</sort_order>
|
||||
<limit>25</limit>
|
||||
</search_criteria>
|
||||
|
||||
Please analyze the query and respond with XML tags containing search criteria."""
|
||||
|
||||
ARXIV_SEARCH_SYSTEM_PROMPT = """You are an expert at crafting arXiv search queries.
|
||||
Your task is to analyze research queries and transform them into optimized arXiv search criteria.
|
||||
Consider query intent, relevant categories, and temporal aspects when creating the search parameters.
|
||||
Always generate English search terms and respond in English regardless of the input language."""
|
||||
|
||||
# Categories selection prompt
|
||||
ARXIV_CATEGORIES_PROMPT = """Select the most relevant arXiv categories for the research query.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Choose 2-4 most relevant categories that best match the research topic.
|
||||
|
||||
Available Categories:
|
||||
|
||||
Computer Science (cs):
|
||||
- cs.AI: Artificial Intelligence (neural networks, machine learning, NLP)
|
||||
- cs.CL: Computation and Language (NLP, machine translation)
|
||||
- cs.CV: Computer Vision and Pattern Recognition
|
||||
- cs.LG: Machine Learning (deep learning, reinforcement learning)
|
||||
- cs.NE: Neural and Evolutionary Computing
|
||||
- cs.RO: Robotics
|
||||
- cs.IR: Information Retrieval
|
||||
- cs.SE: Software Engineering
|
||||
- cs.DB: Databases
|
||||
- cs.DC: Distributed Computing
|
||||
- cs.CY: Computers and Society
|
||||
- cs.HC: Human-Computer Interaction
|
||||
|
||||
Mathematics (math):
|
||||
- math.OC: Optimization and Control
|
||||
- math.PR: Probability
|
||||
- math.ST: Statistics
|
||||
- math.NA: Numerical Analysis
|
||||
- math.DS: Dynamical Systems
|
||||
|
||||
Statistics (stat):
|
||||
- stat.ML: Machine Learning
|
||||
- stat.ME: Methodology
|
||||
- stat.TH: Theory
|
||||
- stat.AP: Applications
|
||||
|
||||
Physics (physics):
|
||||
- physics.comp-ph: Computational Physics
|
||||
- physics.data-an: Data Analysis
|
||||
- physics.soc-ph: Physics and Society
|
||||
|
||||
Electrical Engineering (eess):
|
||||
- eess.SP: Signal Processing
|
||||
- eess.AS: Audio and Speech Processing
|
||||
- eess.IV: Image and Video Processing
|
||||
- eess.SY: Systems and Control
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Deep learning for computer vision"
|
||||
<categories>cs.CV, cs.LG, stat.ML</categories>
|
||||
|
||||
2. Query: "Natural language processing with transformers"
|
||||
<categories>cs.CL, cs.AI, cs.LG</categories>
|
||||
|
||||
3. Query: "Reinforcement learning for robotics"
|
||||
<categories>cs.RO, cs.AI, cs.LG</categories>
|
||||
|
||||
4. Query: "Statistical methods in machine learning"
|
||||
<categories>stat.ML, cs.LG, math.ST</categories>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<categories>List 2-4 most relevant categories, comma-separated</categories>
|
||||
|
||||
Note:
|
||||
- Choose primary categories first, then add related ones
|
||||
- Limit to 2-4 most relevant categories
|
||||
- Order by relevance (most relevant first)
|
||||
- Use comma and space between categories (e.g., "cs.AI, cs.LG")"""
|
||||
|
||||
# 在文件末尾添加新的 prompt
|
||||
ARXIV_LATEST_PROMPT = """Determine if the query is requesting latest papers from arXiv.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Analyze if the query is specifically asking for recent/latest papers from arXiv.
|
||||
|
||||
IMPORTANT RULE:
|
||||
- The query MUST explicitly mention "arXiv" or "arxiv" to be considered a latest arXiv papers request
|
||||
- Queries only asking for recent/latest papers WITHOUT mentioning arXiv should return false
|
||||
|
||||
Indicators for latest papers request:
|
||||
1. MUST HAVE keywords about arXiv:
|
||||
- "arxiv"
|
||||
- "arXiv"
|
||||
AND
|
||||
|
||||
2. Keywords about recency:
|
||||
- "latest"
|
||||
- "recent"
|
||||
- "new"
|
||||
- "newest"
|
||||
- "just published"
|
||||
- "this week/month"
|
||||
|
||||
Examples:
|
||||
|
||||
1. Latest papers request (Valid):
|
||||
Query: "Show me the latest AI papers on arXiv"
|
||||
<is_latest_request>true</is_latest_request>
|
||||
|
||||
2. Latest papers request (Valid):
|
||||
Query: "What are the recent papers about transformers on arxiv"
|
||||
<is_latest_request>true</is_latest_request>
|
||||
|
||||
3. Not a latest papers request (Invalid - no mention of arXiv):
|
||||
Query: "Show me the latest papers about BERT"
|
||||
<is_latest_request>false</is_latest_request>
|
||||
|
||||
4. Not a latest papers request (Invalid - no recency):
|
||||
Query: "Find papers on arxiv about transformers"
|
||||
<is_latest_request>false</is_latest_request>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<is_latest_request>true/false</is_latest_request>
|
||||
|
||||
Note: The response should be true ONLY if both conditions are met:
|
||||
1. Query explicitly mentions arXiv/arxiv
|
||||
2. Query asks for recent/latest papers"""
|
||||
|
||||
ARXIV_LATEST_SYSTEM_PROMPT = """You are an expert at analyzing academic queries.
|
||||
Your task is to determine if the query is specifically requesting latest/recent papers from arXiv.
|
||||
Remember: The query MUST explicitly mention arXiv to be considered valid, even if it asks for recent papers.
|
||||
Always respond in English regardless of the input language."""
|
||||
@@ -0,0 +1,55 @@
|
||||
# Crossref query optimization prompt
|
||||
CROSSREF_QUERY_PROMPT = """Analyze and optimize the query for Crossref search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized Crossref search query.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on the core research topic for the search query.
|
||||
|
||||
Available search fields and filters:
|
||||
1. Basic fields:
|
||||
- title: Search in title
|
||||
- abstract: Search in abstract
|
||||
- author: Search for author names
|
||||
- container-title: Search in journal/conference name
|
||||
- publisher: Search by publisher name
|
||||
- type: Filter by work type (journal-article, book-chapter, etc.)
|
||||
- year: Filter by publication year
|
||||
|
||||
2. Boolean operators:
|
||||
- AND: Both terms must appear
|
||||
- OR: Either term can appear
|
||||
- NOT: Exclude terms
|
||||
- "": Exact phrase match
|
||||
|
||||
3. Special filters:
|
||||
- is-referenced-by-count: Filter by citation count
|
||||
- from-pub-date: Filter by publication date
|
||||
- has-abstract: Filter papers with abstracts
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Machine learning in healthcare after 2020"
|
||||
<query>title:"machine learning" AND title:healthcare AND from-pub-date:2020</query>
|
||||
|
||||
2. Query: "Papers by Geoffrey Hinton about deep learning"
|
||||
<query>author:"Hinton, Geoffrey" AND (title:"deep learning" OR abstract:"deep learning")</query>
|
||||
|
||||
3. Query: "Most cited papers about transformers in Nature"
|
||||
<query>title:transformer AND container-title:Nature AND is-referenced-by-count:[100 TO *]</query>
|
||||
|
||||
4. Query: "Recent BERT applications in medical domain"
|
||||
<query>title:BERT AND abstract:medical AND from-pub-date:2020 AND type:journal-article</query>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<query>Provide the optimized Crossref search query using appropriate fields and operators</query>"""
|
||||
|
||||
# System prompt
|
||||
CROSSREF_QUERY_SYSTEM_PROMPT = """You are an expert at crafting Crossref search queries.
|
||||
Your task is to optimize natural language queries for Crossref's API.
|
||||
Focus on creating precise queries that will return relevant results.
|
||||
Always generate English search terms regardless of the input language.
|
||||
Consider using field-specific search terms and appropriate filters to improve search accuracy."""
|
||||
@@ -0,0 +1,47 @@
|
||||
# 新建文件，添加论文识别提示
|
||||
PAPER_IDENTIFY_PROMPT = """Analyze the query to identify paper details.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Extract paper identification information from the query.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on identifying paper details.
|
||||
|
||||
Possible paper identifiers:
|
||||
1. arXiv ID (e.g., 2103.14030, arXiv:2103.14030)
|
||||
2. DOI (e.g., 10.1234/xxx.xxx)
|
||||
3. Paper title (e.g., "Attention is All You Need")
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query with arXiv ID:
|
||||
Query: "Analyze paper 2103.14030"
|
||||
<paper_info>
|
||||
<paper_source>arxiv</paper_source>
|
||||
<paper_id>2103.14030</paper_id>
|
||||
<paper_title></paper_title>
|
||||
</paper_info>
|
||||
|
||||
2. Query with DOI:
|
||||
Query: "Review the paper with DOI 10.1234/xxx.xxx"
|
||||
<paper_info>
|
||||
<paper_source>doi</paper_source>
|
||||
<paper_id>10.1234/xxx.xxx</paper_id>
|
||||
<paper_title></paper_title>
|
||||
</paper_info>
|
||||
|
||||
3. Query with paper title:
|
||||
Query: "Analyze 'Attention is All You Need' paper"
|
||||
<paper_info>
|
||||
<paper_source>title</paper_source>
|
||||
<paper_id></paper_id>
|
||||
<paper_title>Attention is All You Need</paper_title>
|
||||
</paper_info>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags containing paper information."""
|
||||
|
||||
PAPER_IDENTIFY_SYSTEM_PROMPT = """You are an expert at identifying academic paper references.
|
||||
Your task is to extract paper identification information from queries.
|
||||
Look for arXiv IDs, DOIs, and paper titles."""
|
||||
@@ -0,0 +1,108 @@
|
||||
# PubMed search type prompt
|
||||
PUBMED_TYPE_PROMPT = """Analyze the research query and determine the appropriate PubMed search type.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Available search types:
|
||||
1. basic: General keyword search for medical/biomedical topics
|
||||
2. author: Search by author name
|
||||
3. journal: Search within specific journals
|
||||
4. none: Query not related to medical/biomedical research
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "COVID-19 treatment outcomes"
|
||||
<search_type>basic</search_type>
|
||||
|
||||
2. Query: "Papers by Anthony Fauci"
|
||||
<search_type>author</search_type>
|
||||
|
||||
3. Query: "Recent papers in Nature about CRISPR"
|
||||
<search_type>journal</search_type>
|
||||
|
||||
4. Query: "Deep learning for computer vision"
|
||||
<search_type>none</search_type>
|
||||
|
||||
5. Query: "Transformer architecture for NLP"
|
||||
<search_type>none</search_type>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<search_type>Choose: basic, author, journal, or none</search_type>"""
|
||||
|
||||
# PubMed query optimization prompt
|
||||
PUBMED_QUERY_PROMPT = """Optimize the following query for PubMed search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized PubMed search query.
|
||||
Requirements:
|
||||
- Always generate English search terms regardless of input language
|
||||
- Translate any non-English terms to English before creating the query
|
||||
- Never include non-English characters in the final query
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on the core medical/biomedical topic for the search query.
|
||||
|
||||
Available field tags:
|
||||
- [Title] - Search in title
|
||||
- [Author] - Search for author
|
||||
- [Journal] - Search in journal name
|
||||
- [MeSH Terms] - Search using MeSH terms
|
||||
|
||||
Boolean operators:
|
||||
- AND
|
||||
- OR
|
||||
- NOT
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "COVID-19 treatment in elderly patients"
|
||||
<query>COVID-19[Title] AND treatment[Title/Abstract] AND elderly[Title/Abstract]</query>
|
||||
|
||||
2. Query: "Cancer immunotherapy review articles"
|
||||
<query>cancer immunotherapy[Title/Abstract] AND review[Publication Type]</query>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<query>Provide the optimized PubMed search query</query>"""
|
||||
|
||||
# PubMed sort parameters prompt
|
||||
PUBMED_SORT_PROMPT = """Determine optimal sorting parameters for PubMed results.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Select the most appropriate sorting method and result limit.
|
||||
|
||||
Available sort options:
|
||||
- relevance: Best match to query
|
||||
- date: Most recent first
|
||||
- journal: Sort by journal name
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Latest developments in gene therapy"
|
||||
<sort_by>date</sort_by>
|
||||
<limit>30</limit>
|
||||
|
||||
2. Query: "Classic papers about DNA structure"
|
||||
<sort_by>relevance</sort_by>
|
||||
<limit>20</limit>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<sort_by>Choose: relevance, date, or journal</sort_by>
|
||||
<limit>Suggest number between 10-50</limit>"""
|
||||
|
||||
# System prompts
|
||||
PUBMED_TYPE_SYSTEM_PROMPT = """You are an expert at analyzing medical and scientific queries.
|
||||
Your task is to determine the most appropriate PubMed search type.
|
||||
Consider the query's focus and intended search scope.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
PUBMED_QUERY_SYSTEM_PROMPT = """You are an expert at crafting PubMed search queries.
|
||||
Your task is to optimize natural language queries using PubMed's search syntax.
|
||||
Focus on creating precise, targeted queries that will return relevant medical literature.
|
||||
Always generate English search terms regardless of the input language."""
|
||||
|
||||
PUBMED_SORT_SYSTEM_PROMPT = """You are an expert at optimizing PubMed search results.
|
||||
Your task is to determine the best sorting parameters based on the query context.
|
||||
Consider the balance between relevance and recency.
|
||||
Always respond in English regardless of the input language."""
|
||||
@@ -0,0 +1,276 @@
|
||||
# Search type prompt
|
||||
SEMANTIC_TYPE_PROMPT = """Determine the most appropriate search type for Semantic Scholar.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Analyze the research query and select the most appropriate search type for Semantic Scholar API.
|
||||
|
||||
Available search types:
|
||||
|
||||
1. paper: General paper search
|
||||
- Use for broad topic searches
|
||||
- Looking for specific papers
|
||||
- Keyword-based searches
|
||||
Example: "transformer models in NLP"
|
||||
|
||||
2. author: Author-based search
|
||||
- Finding works by specific researchers
|
||||
- Author profile analysis
|
||||
Example: "papers by Yoshua Bengio"
|
||||
|
||||
3. paper_details: Specific paper lookup
|
||||
- Getting details about a known paper
|
||||
- Finding specific versions or citations
|
||||
Example: "Attention is All You Need paper details"
|
||||
|
||||
4. citations: Citation analysis
|
||||
- Finding papers that cite a specific work
|
||||
- Impact analysis
|
||||
Example: "papers citing BERT"
|
||||
|
||||
5. references: Reference analysis
|
||||
- Finding papers cited by a specific work
|
||||
- Background research
|
||||
Example: "references in GPT-3 paper"
|
||||
|
||||
6. recommendations: Paper recommendations
|
||||
- Finding similar papers
|
||||
- Research direction exploration
|
||||
Example: "papers similar to Transformer"
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Latest papers about deep learning"
|
||||
<search_type>paper</search_type>
|
||||
|
||||
2. Query: "Works by Geoffrey Hinton since 2020"
|
||||
<search_type>author</search_type>
|
||||
|
||||
3. Query: "Papers citing the original Transformer paper"
|
||||
<search_type>citations</search_type>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<search_type>Choose the most appropriate search type from the list above</search_type>"""
|
||||
|
||||
# Query optimization prompt
|
||||
SEMANTIC_QUERY_PROMPT = """Optimize the following query for Semantic Scholar search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into an optimized search query for maximum relevance.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements. Focus only on the core research topic for the search query.
|
||||
|
||||
Query optimization guidelines:
|
||||
|
||||
1. Use quotes for exact phrases
|
||||
- Ensures exact matching
|
||||
- Reduces irrelevant results
|
||||
Example: "\"attention mechanism\"" vs attention mechanism
|
||||
|
||||
2. Include key technical terms
|
||||
- Use specific technical terminology
|
||||
- Include common variations
|
||||
Example: "transformer architecture" neural networks
|
||||
|
||||
3. Author names (if relevant)
|
||||
- Include full names when known
|
||||
- Consider common name variations
|
||||
Example: "Geoffrey Hinton" OR "G. E. Hinton"
|
||||
|
||||
Examples:
|
||||
|
||||
1. Natural query: "Recent advances in transformer models"
|
||||
<query>"transformer model" "neural architecture" deep learning</query>
|
||||
|
||||
2. Natural query: "BERT applications in text classification"
|
||||
<query>"BERT" "text classification" "language model" application</query>
|
||||
|
||||
3. Natural query: "Deep learning for computer vision by Kaiming He"
|
||||
<query>"deep learning" "computer vision" author:"Kaiming He"</query>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<query>Provide the optimized search query</query>
|
||||
|
||||
Note:
|
||||
- Balance between specificity and coverage
|
||||
- Include important technical terms
|
||||
- Use quotes for key phrases
|
||||
- Consider synonyms and related terms"""
|
||||
|
||||
# Fields selection prompt
|
||||
SEMANTIC_FIELDS_PROMPT = """Select relevant fields to retrieve from Semantic Scholar.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Determine which paper fields should be retrieved based on the research needs.
|
||||
|
||||
Available fields:
|
||||
|
||||
Core fields:
|
||||
- title: Paper title (always included)
|
||||
- abstract: Full paper abstract
|
||||
- authors: Author information
|
||||
- year: Publication year
|
||||
- venue: Publication venue
|
||||
|
||||
Citation fields:
|
||||
- citations: Papers citing this work
|
||||
- references: Papers cited by this work
|
||||
|
||||
Additional fields:
|
||||
- embedding: Paper vector embedding
|
||||
- tldr: AI-generated summary
|
||||
- venue: Publication venue/journal
|
||||
- url: Paper URL
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Latest developments in NLP"
|
||||
<fields>title, abstract, authors, year, venue, citations</fields>
|
||||
|
||||
2. Query: "Most influential papers in deep learning"
|
||||
<fields>title, abstract, authors, year, citations, references</fields>
|
||||
|
||||
3. Query: "Survey of transformer architectures"
|
||||
<fields>title, abstract, authors, year, tldr, references</fields>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<fields>List relevant fields, comma-separated</fields>
|
||||
|
||||
Note:
|
||||
- Choose fields based on the query's purpose
|
||||
- Include citation data for impact analysis
|
||||
- Consider tldr for quick paper screening
|
||||
- Balance completeness with API efficiency"""
|
||||
|
||||
# Sort parameters prompt
|
||||
SEMANTIC_SORT_PROMPT = """Determine optimal sorting parameters for the query.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Select the most appropriate sorting method and result limit for the search.
|
||||
Always generate English search terms regardless of the input language.
|
||||
|
||||
Sorting options:
|
||||
|
||||
1. relevance (default)
|
||||
- Best match to query terms
|
||||
- Recommended for specific technical searches
|
||||
Example: "specific algorithm implementations"
|
||||
|
||||
2. citations
|
||||
- Sort by citation count
|
||||
- Best for finding influential papers
|
||||
Example: "most important papers in deep learning"
|
||||
|
||||
3. year
|
||||
- Sort by publication date
|
||||
- Best for following recent developments
|
||||
Example: "latest advances in NLP"
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Recent breakthroughs in AI"
|
||||
<sort_by>year</sort_by>
|
||||
<limit>30</limit>
|
||||
|
||||
2. Query: "Most influential papers about GANs"
|
||||
<sort_by>citations</sort_by>
|
||||
<limit>20</limit>
|
||||
|
||||
3. Query: "Specific papers about BERT fine-tuning"
|
||||
<sort_by>relevance</sort_by>
|
||||
<limit>25</limit>
|
||||
|
||||
Please analyze the query and respond ONLY with XML tags:
|
||||
<sort_by>Choose: relevance, citations, or year</sort_by>
|
||||
<limit>Suggest number between 10-50</limit>
|
||||
|
||||
Note:
|
||||
- Consider the query's temporal aspects
|
||||
- Balance between comprehensive coverage and information overload
|
||||
- Use citation sorting for impact analysis
|
||||
- Use year sorting for tracking developments"""
|
||||
|
||||
# System prompts for each task
|
||||
SEMANTIC_TYPE_SYSTEM_PROMPT = """You are an expert at analyzing academic queries.
|
||||
Your task is to determine the most appropriate type of search on Semantic Scholar.
|
||||
Consider the query's intent, scope, and specific research needs.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
SEMANTIC_QUERY_SYSTEM_PROMPT = """You are an expert at crafting Semantic Scholar search queries.
|
||||
Your task is to optimize natural language queries for maximum relevance.
|
||||
Focus on creating precise queries that leverage the platform's search capabilities.
|
||||
Always generate English search terms regardless of the input language."""
|
||||
|
||||
SEMANTIC_FIELDS_SYSTEM_PROMPT = """You are an expert at Semantic Scholar data fields.
|
||||
Your task is to select the most relevant fields based on the research context.
|
||||
Consider both essential and supplementary information needs.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
SEMANTIC_SORT_SYSTEM_PROMPT = """You are an expert at optimizing search results.
|
||||
Your task is to determine the best sorting parameters based on the query context.
|
||||
Consider the balance between relevance, impact, and recency.
|
||||
Always respond in English regardless of the input language."""
|
||||
|
||||
# 添加新的综合搜索提示词
|
||||
SEMANTIC_SEARCH_PROMPT = """Analyze and optimize the research query for Semantic Scholar search.
|
||||
|
||||
Query: {query}
|
||||
|
||||
Task: Transform the natural language query into optimized search criteria for Semantic Scholar.
|
||||
|
||||
IMPORTANT: Ignore any requirements about journal ranking (CAS, JCR, IF index),
|
||||
or output format requirements when generating the search terms. These requirements
|
||||
should be considered only for post-search filtering, not as part of the core query.
|
||||
|
||||
Available search options:
|
||||
1. Paper search:
|
||||
- Title and abstract search
|
||||
- Author search
|
||||
- Field-specific search
|
||||
Example: "transformer architecture neural networks"
|
||||
|
||||
2. Field tags:
|
||||
- title: Search in title
|
||||
- abstract: Search in abstract
|
||||
- authors: Search by author names
|
||||
- venue: Search by publication venue
|
||||
Example: "title:transformer authors:\"Vaswani\""
|
||||
|
||||
3. Advanced options:
|
||||
- Year range filtering
|
||||
- Citation count filtering
|
||||
- Venue filtering
|
||||
Example: "deep learning year>2020 venue:\"NeurIPS\""
|
||||
|
||||
Examples:
|
||||
|
||||
1. Query: "Recent transformer papers by Vaswani with high impact"
|
||||
<search_criteria>
|
||||
<query>title:transformer authors:"Vaswani" year>2017</query>
|
||||
<search_type>paper</search_type>
|
||||
<fields>title,abstract,authors,year,citations,venue</fields>
|
||||
<sort_by>citations</sort_by>
|
||||
<limit>30</limit>
|
||||
</search_criteria>
|
||||
|
||||
2. Query: "Most cited papers about BERT in top conferences"
|
||||
<search_criteria>
|
||||
<query>title:BERT venue:"ACL|EMNLP|NAACL"</query>
|
||||
<search_type>paper</search_type>
|
||||
<fields>title,abstract,authors,year,citations,venue,references</fields>
|
||||
<sort_by>citations</sort_by>
|
||||
<limit>25</limit>
|
||||
</search_criteria>
|
||||
|
||||
Please analyze the query and respond with XML tags containing complete search criteria."""
|
||||
|
||||
SEMANTIC_SEARCH_SYSTEM_PROMPT = """You are an expert at crafting Semantic Scholar search queries.
|
||||
Your task is to analyze research queries and transform them into optimized search criteria.
|
||||
Consider query intent, field relevance, and citation impact when creating the search parameters.
|
||||
Focus on producing precise and comprehensive search criteria that will yield the most relevant results.
|
||||
Always generate English search terms and respond in English regardless of the input language."""
|
||||
在新工单中引用
屏蔽一个用户