topic_dictionary.py

DB_VENUES = {"SIGMOD", "VLDB", "PVLDB", "PACMMOD"}
NEURIPS_VENUES = {"NEURIPS", "NIPS"}


DB_TOPIC_DICT = {
    "Query Optimization": {
        "category": "Query Processing",
        "keywords": [
            "query optimization",
            "query optimizer",
            "cardinality estimation",
            "join order",
            "cost model",
            "learned optimizer",
            "query processing",
            "sql",
            "join",
            "joins",
            "querying",
            "execution",
            "execution engine",
            "approximate processing",
            "analytical processing",
        ],
    },
    "Indexing": {
        "category": "Storage & Access",
        "keywords": [
            "index",
            "indexing",
            "b-tree",
            "btree",
            "lsm",
            "inverted index",
            "bitmap index",
            "storage",
            "main memory",
            "persistent memory",
            "flash memory",
            "in-memory",
        ],
    },
    "Transactions": {
        "category": "Systems",
        "keywords": [
            "transaction",
            "concurrency control",
            "serializability",
            "commit protocol",
            "recovery",
            "acid",
        ],
    },
    "Distributed DB": {
        "category": "Systems",
        "keywords": [
            "distributed database",
            "distributed db",
            "distributed query",
            "shared-nothing",
            "replication",
            "partitioning",
            "distributed transaction",
            "distributed",
            "parallel",
            "massively parallel",
            "mapreduce",
            "spark",
            "key-value",
        ],
    },
    "Cloud DB": {
        "category": "Systems",
        "keywords": [
            "cloud",
            "serverless",
            "elastic",
            "multi-tenant",
            "storage disaggregation",
            "warehouse-as-a-service",
        ],
    },
    "XML": {
        "category": "Data Models",
        "keywords": [
            "xml",
            "xpath",
            "xquery",
        ],
    },
    "RDF / Semantic Web": {
        "category": "Data Models",
        "keywords": [
            "rdf",
            "semantic web",
            "sparql",
            "ontology",
            "knowledge graph",
        ],
    },
    "Graph Data": {
        "category": "Data Models",
        "keywords": [
            "graph",
            "graphs",
            "graph database",
            "graph query",
            "subgraph",
            "property graph",
            "graph pattern",
            "knowledge base",
            "knowledge bases",
            "information network",
            "information networks",
            "road network",
            "road networks",
            "social network",
            "social networks",
            "bipartite graph",
            "community detection",
            "community search",
            "shortest path",
            "influence maximization",
        ],
    },
    "Streaming": {
        "category": "Data Processing",
        "keywords": [
            "stream",
            "streaming",
            "data stream",
            "event processing",
            "window aggregation",
            "continuous query",
        ],
    },
    "Data Integration": {
        "category": "Data Processing",
        "keywords": [
            "data integration",
            "schema matching",
            "entity resolution",
            "data cleaning",
            "data fusion",
            "etl",
            "entity matching",
            "information extraction",
            "heterogeneous information",
            "functional dependency",
            "functional dependencies",
        ],
    },
    "Data Analytics": {
        "category": "Applications",
        "keywords": [
            "analytics",
            "big analytics",
            "interactive exploration",
            "exploratory analysis",
            "analysis",
            "evaluation",
            "benchmark",
            "benchmarking",
            "workload",
            "workloads",
            "performance",
            "interactive",
            "large-scale",
            "big data",
            "data science",
        ],
    },
    "Information Retrieval": {
        "category": "Applications",
        "keywords": [
            "information retrieval",
            "search engine",
            "ranking",
            "top-k",
            "retrieval",
            "search",
            "keyword search",
            "natural language",
            "question answering",
            "social media",
            "web",
        ],
    },
    "ML for DB": {
        "category": "AI for Data Systems",
        "keywords": [
            "machine learning",
            "deep learning",
            "neural",
            "learned",
            "ai for database",
            "ml for database",
            "learning",
            "federated learning",
            "reinforcement learning",
            "probabilistic",
            "clustering",
            "sampling",
            "estimation",
            "detection",
        ],
    },
    "Text-to-SQL / NLIDB": {
        "category": "AI for Data Systems",
        "keywords": [
            "text-to-sql",
            "nlidb",
            "natural language interface",
            "question answering over databases",
            "text to sql",
        ],
    },
    "Vector Search": {
        "category": "AI for Data Systems",
        "keywords": [
            "vector search",
            "approximate nearest neighbor",
            "ann search",
            "embedding search",
            "similarity search",
        ],
    },
    "Privacy / Security": {
        "category": "Trust",
        "keywords": [
            "privacy",
            "security",
            "access control",
            "differential privacy",
            "encrypted",
            "secure query",
        ],
    },
    "Spatial / Temporal": {
        "category": "Specialized Data",
        "keywords": [
            "spatial",
            "trajectory",
            "geospatial",
            "temporal",
            "time series",
            "spatio-temporal",
        ],
    },
    "Crowdsourcing": {
        "category": "Applications",
        "keywords": [
            "crowdsourcing",
            "human-in-the-loop",
            "crowd",
        ],
    },
}


NEURIPS_TOPIC_DICT = {
    "Classical Statistical ML": {
        "category": "Core ML",
        "keywords": [
            "classification",
            "regression",
            "clustering",
            "kernel",
            "sparse",
            "feature selection",
            "matrix factorization",
            "matrix completion",
            "low-rank",
            "tensor decomposition",
            "principal component",
            "principal component analysis",
            "random forest",
            "random forests",
            "gradient boosting",
            "recommender systems",
            "topic modeling",
            "survival analysis",
            "semi-supervised",
            "unsupervised",
            "metric learning",
            "structured prediction",
            "dimensionality reduction",
        ],
    },
    "Probabilistic Modeling": {
        "category": "Core ML",
        "keywords": [
            "bayesian",
            "probabilistic",
            "variational inference",
            "variational",
            "gaussian process",
            "gaussian processes",
            "graphical model",
            "graphical models",
            "markov",
            "monte carlo",
            "posterior",
            "latent variable",
            "estimation",
            "inference",
            "sampling",
            "density estimation",
            "distribution estimation",
            "maximum likelihood",
            "point process",
            "point processes",
        ],
    },
    "Optimization & Learning Theory": {
        "category": "Core ML",
        "keywords": [
            "optimization",
            "optimal",
            "gradient descent",
            "stochastic gradient",
            "policy gradient",
            "generalization",
            "sample complexity",
            "regret bounds",
            "bounds",
            "convergence",
            "regularization",
            "online learning",
            "theoretical",
            "learning curves",
            "bilevel",
            "risk minimization",
            "variance reduction",
            "mirror descent",
            "approximation",
            "minimization",
            "lower bound",
            "lower bounds",
        ],
    },
    "Reinforcement Learning & Bandits": {
        "category": "Decision Making",
        "keywords": [
            "reinforcement learning",
            "offline reinforcement learning",
            "policy optimization",
            "policy evaluation",
            "bandit",
            "bandits",
            "markov decision process",
            "pomdp",
            "control",
            "planning",
            "reward shaping",
            "multi-agent reinforcement learning",
            "policy",
            "decision making",
            "thompson sampling",
            "best arm",
            "best-arm identification",
            "best arm identification",
        ],
    },
    "Games & Mechanism Design": {
        "category": "Decision Making",
        "keywords": [
            "game",
            "games",
            "zero-sum",
            "zero-sum games",
            "extensive-form games",
            "stackelberg games",
            "equilibrium computation",
            "influence maximization",
        ],
    },
    "Representation Learning": {
        "category": "Foundation Learning",
        "keywords": [
            "representation learning",
            "self-supervised learning",
            "contrastive learning",
            "embedding",
            "embeddings",
            "metric learning",
            "feature learning",
            "pretraining",
            "pre-training",
            "continual learning",
            "transfer learning",
            "domain adaptation",
            "distillation",
            "representations",
            "feature",
            "features",
            "latent",
            "attention",
            "predictive coding",
            "positional encoding",
        ],
    },
    "Generative Models & Diffusion": {
        "category": "Generative AI",
        "keywords": [
            "generative model",
            "generative models",
            "generative",
            "diffusion model",
            "diffusion models",
            "diffusion",
            "diffusion training",
            "generative adversarial",
            "gan",
            "gans",
            "variational autoencoder",
            "vae",
            "flow matching",
            "image generation",
        ],
    },
    "Language Models & NLP": {
        "category": "Generative AI",
        "keywords": [
            "language model",
            "language models",
            "large language model",
            "large language models",
            "llm",
            "llms",
            "transformer",
            "transformers",
            "machine translation",
            "semantic parsing",
            "in-context learning",
            "fine-tuning",
            "natural language",
            "question answering",
            "text generation",
            "recurrent",
            "speech recognition",
            "communication",
        ],
    },
    "Computer Vision & 3D Perception": {
        "category": "Perception",
        "keywords": [
            "image",
            "image classification",
            "object detection",
            "semantic segmentation",
            "image segmentation",
            "video",
            "visual",
            "visual recognition",
            "vision transformer",
            "vision transformers",
            "point cloud",
            "pose estimation",
            "3d detector",
            "stereo matching",
            "tracking",
            "reconstruction",
        ],
    },
    "Multimodal & Vision-Language": {
        "category": "Perception",
        "keywords": [
            "multimodal",
            "vision-language",
            "visual question answering",
            "language-guided",
            "text-to-image",
            "text-to-video",
            "text to video",
            "event-to-video",
            "table semantic parsing",
            "video generation",
        ],
    },
    "Graph Machine Learning": {
        "category": "Structured Learning",
        "keywords": [
            "graph",
            "graphs",
            "graph neural",
            "graph neural network",
            "graph neural networks",
            "graph representation learning",
            "graph learning",
            "graph transformer",
            "link prediction",
            "out-of-graph",
            "node classification",
            "message passing",
            "molecular graph",
            "learning on graphs",
        ],
    },
    "Robustness, Privacy & Safety": {
        "category": "Trustworthy AI",
        "keywords": [
            "adversarial",
            "robustness",
            "robust",
            "certified defense",
            "certifiable defense",
            "out-of-distribution",
            "ood",
            "differential privacy",
            "privacy",
            "poisoned samples",
            "safety",
        ],
    },
    "Federated & Distributed Learning": {
        "category": "Systems for ML",
        "keywords": [
            "federated learning",
            "distributed",
            "decentralized",
            "byzantine",
            "communication-efficient",
            "parallel algorithms",
            "device unavailability",
            "distributed newton",
        ],
    },
    "Causality, Dynamics & Scientific ML": {
        "category": "Scientific ML",
        "keywords": [
            "causal",
            "dynamical systems",
            "neural ode",
            "state space",
            "time series",
            "spatiotemporal",
            "spatio-temporal",
            "physics",
            "scientific",
            "trajectory",
            "uncertainty",
            "control with multiple frequencies",
        ],
    },
    "Biology, Molecules & Health": {
        "category": "AI for Science",
        "keywords": [
            "protein",
            "molecular",
            "molecule",
            "structural biology",
            "cellular",
            "genomics",
            "drug",
            "patient",
            "medical",
            "neuroimaging",
            "biological",
        ],
    },
}


GENERIC_TERMS = {"data", "database", "system", "efficient", "scalable", "query", "model", "models"}


def normalize_venue_name(venue: str) -> str:
    normalized = str(venue).strip().upper()
    if normalized in {"NIPS", "NEURIPS"}:
        return "NEURIPS"
    return normalized


def topic_dict_for_venue(venue: str) -> dict[str, dict[str, list[str] | str]]:
    normalized = normalize_venue_name(venue)
    if normalized in NEURIPS_VENUES:
        return NEURIPS_TOPIC_DICT
    return DB_TOPIC_DICT