grobidOrg/grobid

Features

• PDF Semantic Parsers - Provides a machine learning system designed to extract semantic entities and technical metadata from PDF files.
• PDF to XML Converters - Transforms complete academic PDF documents, including headers, bodies, and bibliographies, into structured XML using semantic segmentation.
• Bibliographic Metadata Extraction - Identifies and normalizes authors, affiliations, and titles from scientific documents into structured XML or BibTeX.
• Academic Paper Metadata Extractors - Identifies and normalizes key scholarly metadata including titles, authors, and affiliations from research papers.
• Affiliation Parsers - Converts raw organizational address strings and academic affiliations into normalized XML formats.
• Citation Parsers - Transforms individual raw bibliographical reference strings into structured XML or BibTeX records.
• Citation Reference Extraction - Identifies and extracts bibliographic citation fragments from the reference sections of academic documents.
• Document Structure Analysis - Uses sequence labeling models to segment PDF documents into functional regions like headers, footers, and body text.
• Scientific Document Segmentation - Uses deep learning to segment academic PDFs into functional regions like abstracts and bodies.
• Document Parsing Model Training - Provides a training pipeline for neural networks designed to recognize academic document layouts and text.
• Document Layout Analysis - Analyzes font styles and spatial positioning to identify structural hierarchies, text, and tables within complex documents.
• Neural Bibliographic Extractors - Uses deep learning models to parse citations, affiliations, and publication headers with high accuracy.
• Scholarly Document Digitization - Digitizes academic papers by transforming research PDFs into machine-readable formats with semantic labeling and mapped coordinates.
• Bibliographic Reference Parsing - Deconstructs raw bibliographic reference strings from PDFs into structured metadata and resolves them via APIs.
• Citation Analysis Workflows - Links inline citation callouts to full bibliographic references and validates them against external registries and DOIs.
• Citation Context Resolution - Associates inline citation callouts within the text with their corresponding entries in the bibliography.
• Document Reference Mapping - Links inline callouts and citations to their corresponding bibliographic references, figures, and tables.
• Bibliographic Registry Resolution - Resolves extracted bibliographic data against external academic registries to normalize records and append DOIs.
• Publication Context Extraction - Identifies journal titles, publishers, and document types to categorize the source of a publication.
• Bibliographic Identifier Extractions - Detects and classifies strong academic identifiers such as DOI, PII, ISSN, and ISBN.
• Document Header Extraction - Extracts titles, authors, and abstracts from the initial pages of academic PDF documents into structured formats.
• Scholarly Entity Parsing - Recognizes and normalizes scholarly metadata strings, including physical quantities and funding information.
• Scholarly Document Structure Parsing - Converts academic PDF publications into structured XML by extracting headers, body text, and bibliographic sections.
• Semantic Entity Coordinate Mapping - Maps extracted semantic entities to their original PDF bounding boxes for precise visual highlighting.
• Document Layout Bounding Box Extractors - Extracts bounding box coordinates and font styles to improve the accuracy of structural document recognition.
• Academic Author Name Parsing - Normalizes raw strings of author and editor names from scientific headers and references into XML.
• Body Text Segmentation - Segments the PDF body into structured elements including paragraphs, section titles, footnotes, and figures.
• Bibliographic Metadata Enrichment - Enriches bibliographic records by matching extracted data against external services to append missing identifiers.
• Document Pre-annotation Generators - Produces XML files from raw PDFs to serve as starting points for manual correction in the training pipeline.
• Document Language Detection - Automatically identifies the natural language used within a document to facilitate linguistic analysis.
• Domain-Specific Parsing Adaptations - Applies specialized machine learning processing logic tailored for specific document types like patents or medical publications.
• Machine Learning Training - Fine-tunes document processing models using custom datasets and integrated evaluation pipelines.
• GPU Training Accelerators - Utilizes GPU hardware to accelerate the deep learning process and reduce model training time.
• Annotation Validation Tools - Provides interfaces for correcting and validating pre-annotated XML data side-by-side with original source PDFs.
• Training Corpus Generators - Generates pre-annotated XML files from raw PDFs to create ground-truth training corpora for parsing models.
• Sequence Labeling Architectures - Implements sequence labeling architectures to segment structural regions of academic documents.
• Training Data Generators - Creates structured training data files from PDF documents to bootstrap machine learning parsing models.
• Funding - Extracts grant and funder details from research documents and matches them against established registries.
• Table Bounding Box Detections - Detects the overall position and boundaries of tables within document layouts using bounding box coordinates.
• Technical Element Isolation - Isolates technical components such as mathematical formulas, figures, and list items from the main body text.
• Publication License Identification - Detects copyright owners and specific licensing terms associated with scholarly publications.
• Scientific Domain - Identifies specialized mentions such as software, datasets, and astronomical entities within scientific literature.
• Administrative Metadata Extraction - Captures administrative and legal metadata including funding disclosures, copyright statements, and submission dates.
• Patent Data Extraction - Extracts and parses structured bibliographic and reference data specifically from patent publications.
• Parsing Model Trainers - Provides a dedicated environment for training, evaluating, and exporting custom sequence labeling models for document parsing.
• Model Architecture Configurations - Provides a configuration system to select specific deep learning or CRF architectures for different parsing tasks.
• Batch Document Processing - Processes large volumes of PDF files simultaneously using multi-threaded clients to increase throughput.
• Multi-Threaded Batch Processing - Implements multi-threaded batch processing to maximize hardware throughput when parsing large document collections.
• Hybrid Parsing Engines - Uses a hybrid parsing engine that switches between deep learning and feature-based models to balance speed and accuracy.
• Extraction Accuracy Evaluators - Measures the precision, recall, and F1-scores of parsed data against ground-truth documents to evaluate extraction accuracy.
• Parsing Pipeline Evaluators - Measures the end-to-end accuracy of the PDF-to-XML processing pipeline using independent holdout datasets.

Star 历史