Skip to content

Latest commit

 

History

History
63 lines (46 loc) · 4.39 KB

CHANGELOG.md

File metadata and controls

63 lines (46 loc) · 4.39 KB

Changelog


Version 1.1.0 (first public beta release)

New Features

  • Properties: Added colors (including roof and building colors), e.g., a “brown (bench).”
  • OSM Expert Editor: Added the first version of an editor to update and add tags to existing OSM queries in the GUI.
  • Feedback Form: Added a form to the GUI to allow users to give direct feedback.

Fixes

  • New Training Data: Better entity and property detection via improvements to data generation and training parameters.

Work in Progress

  • Laid Groundwork for Entity Clusters: Ability to generate training data for clusters of entities, allowing queries like "3 Italian restaurants next to each other" or "at least 5 wind generators in a radius of 200 m." This feature will be available in a future model iteration.

Version 1.0.0

New Features

  1. Entity Detection

    • Semantic Entities: Spot identifies general categories like "restaurant" and "train station," allowing recognition of places based on type.
    • Named Entities (Brands): Detection of specific brand names, including "McDonald's," "KFC," "Tchibo," and compound names like "Thalia bookstore."
  2. Entity Properties

    • Named Properties: Spot identifies properties like "vegan (food shop)" or "Italian (restaurant)" for refined queries.
    • Numerical Properties: Ability to interpret quantitative descriptors, including height, levels, and house numbers.
  3. Area Recognition

    • Named and Administrative Areas: Support for cities, districts, and regions, including multi-word areas (e.g., "New York") and regions like "Nordrhein-Westfalen."
    • Bounding Box for Undefined Area Queries: Introduces bounding box support for identifying entities within a broader, undefined area.
  4. Distance Relations

    • Numerical and Written Distances: Spot interprets both numeric distances (e.g., "100 meters") and written forms (e.g., "one hundred meters").
    • Relative Distance Terms: Supports terms like "next to," "opposite from," and "beside" to improve natural understanding of spatial relationships.
    • Distance Chain and Radius Support: Multiple distance-based relations are supported, including radius constraints (e.g., "A to B and C") and entity chains (e.g., "A to B and B to C").
  5. Contains Relations

    • Basic Containment: Recognizes relationships such as "a fountain within a park" and "a shop inside a mall."
    • With Relations: Expanded containment to support "with" relationships, such as "a park with a fountain" or "hotel with a parking lot."
  6. Spatial Terms and Descriptors

    • Relative Distance Phrasing: Enhanced natural language understanding for relative spatial terms like "close to," "next to," and "behind."
    • Descriptor Matching: Improved matching of descriptors with slight variations, such as plurals ("bookshops" vs. "bookshop") and minor differences ("bookstore" vs. "book shop").
  7. Prompt and Linguistic Features

    • Typo Handling: Improved error tolerance to manage typos in names and common words (e.g., "MacDonalds" for "McDonald's").
    • Language Style Variability: Added support for both formal and casual query styles.
    • Multilingual Area and Brand Names: Recognizes area names and locations in multiple languages and alphabets, including non-Roman alphabets like Cyrillic and Greek.
    • Multiple Sentence Queries: Supports both single and multi-sentence structures in user queries.
  8. User Interface and Frontend Enhancements

    • Rendering Results on the Map: Results from natural language queries are displayed directly on the map, allowing users to visually locate and assess relevant areas.
    • Exploring Candidate Locations: The interface integrates third-party map services, enabling users to investigate specific coordinates using tools like Google Maps or Google Street View.
    • Refining Search Queries Visually: Users can modify search parameters interactively, including adjusting distance relations between objects for more targeted results.
    • Session Management: Added functionality for saving search sessions for future use, as well as loading previously saved sessions to continue work seamlessly.
    • Exporting Map Data: The system supports exporting map data in multiple formats, allowing users to use the data in external applications or workflows.

Fixes

  • No fixes in version 1.0.0.