Spaces:
Running
Running
File size: 3,707 Bytes
b904262 a689cd6 b904262 a689cd6 b904262 a689cd6 b904262 a689cd6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 |
---
title: Normalized ConceptNet Explorer
emoji: ⚡
colorFrom: green
colorTo: blue
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: true
license: cc-by-sa-4.0
tags:
- conceptnet
- knowledge-graph
- sqlite
- normalized
- gradio
- fast-queries
---
# ⚡ Normalized ConceptNet Explorer (V7)
This application is a high-performance explorer for a normalized, filtered, and optimized version of the ConceptNet 5.5 knowledge graph.
It is designed to be **extremely fast**, returning queries in milliseconds instead of minutes. It queries a 1.78 GB optimized SQLite database with integer-based joins, not the 23.6 GB un-normalized file.
## Features
This app provides a full suite of tools to explore the normalized database:
- **⚡ Semantic Profile**: Explore relations for any word in real-time. This now runs in ~4 fast SQL queries instead of 24+ slow ones.
- **⚡ Query Builder**: Build custom queries (start node, relation, end node) that are executed with fast, integer-based joins.
- **⚡ Raw SQL**: Execute SQL queries directly against the new, normalized database schema (see schema below).
- **⚡ Schema**: Browse the new, efficient database schema, including all tables, indexes, and row counts.
## How It Works: The Normalized Database
This app's speed and correctness come from the new database it queries: [cstr/conceptnet-normalized-multi](https://huggingface.co/datasets/cstr/conceptnet-normalized-multi).
This database was created by a V7 normalization script that fixed critical issues found in the original data:
1. **Normalization (Speed & Size)**: The original 23.6 GB `edge` table (34M rows) was bloated with text URLs. The new 1.78 GB `edge_norm` table replaces these with tiny integer foreign keys.
2. **Data Correctness (V7 Fix)**: The original `node` table (28M rows) was used as the source of truth. We migrated all 28M nodes and their authoritative `language` columns.
3. **Preserves Cross-Language Links**: The 34M edges were filtered to keep any edge where at least one node (start or end) was in our 11 target languages (`en`, `de`, `fr`, `it`, `es`, `ar`, `fa`, `grc`, `he`, `la`, `hbo`). This is critical, as it correctly preserves cross-language connections (e.g., `犬 (ja) -> hund (de)`), which were broken in previous attempts.
The result is a clean, fast, and data-correct database that contains all relevant connections for our target languages.
## Supported Languages
This normalized version includes edges for 11 languages:
- English (en)
- German (de)
- French (fr)
- Italian (it)
- Spanish (es)
- Arabic (ar)
- Persian (fa)
- Ancient Greek (grc)
- Hebrew (he)
- Latin (la)
- Biblical Hebrew (hbo)
Cross-language connections from other languages to these target languages are preserved.
## Original Dataset Information
This work includes data from ConceptNet 5, which was compiled by the Commonsense Computing Initiative. ConceptNet 5 is freely available under the Creative Commons Attribution-ShareAlike license (CC BY SA 4.0) from http://conceptnet.io.
For a full list of licenses and attributions for included resources such as WordNet, Open Multilingual WordNet, and Wikimedia projects, please see the original dataset card.
## Citation Information
If you use this data in your work, please cite the original ConceptNet 5.5 paper:
```bibtex
@inproceedings{speer2017conceptnet,
author = {Robyn Speer and Joshua Chin and Catherine Havasi},
title = {ConceptNet 5.5: An Open Multilingual Graph of General Knowledge},
booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2017},
pages = {4444--4451},
url = {http://aaai.org/ocs/index.php/AAAI/AAAI17/paper/view/14972}
}
``` |