Model Card for Cultural Heritage Metadata Accuracy Detection model

This model is trained to detect the quality of Italian cultural heritage metadata, assigning a score of high quality or low quality to input text. The model was trained on the Annotated dataset to assess the accuracy of the textual description of cultural heritage records dataset.

The dataset contains more than 100K textual descriptions of cultural items from Cultura Italia, the Italian National Cultural aggregator. Each of the description is labeled either HIGH or LOW quality, according its adherence to the standard cataloguing guidelines provided by Istituto Centrale per il Catalogo e la Documentazione (ICCD). More precisely, each description is labeled as HIGH quality if the object and subject of the item (for which the description is provided) are both described according to the ICCD guidelines, and as LOW quality in all other cases. Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections. The dataset was developed to support the training and testing of ML text classification approaches for automatically assessing the quality of textual descriptions in digital Cultural Heritage repositories.

Uses

This model could potentially be useful for performing validation on metadata quality. However, before using this model, it would be sensible to validate:

  • how it performs on your data
  • if you agree with the quality ratings assigned in the original dataset.

It will likely make more sense to use this model in the context of a 'human in the loop' pipeline whereby the model is used to surface metadata records which may benefit from additional human attention rather than using it to make automatic decisions.

Model Trained Using AutoTrain

  • Problem type: Multi-class Classification
  • Model ID: 48840118272
  • CO2 Emissions (in grams): 7.1714

Validation Metrics

  • Loss: 0.085
  • Accuracy: 0.972
  • Macro F1: 0.972
  • Micro F1: 0.972
  • Weighted F1: 0.972
  • Macro Precision: 0.972
  • Micro Precision: 0.972
  • Weighted Precision: 0.972
  • Macro Recall: 0.972
  • Micro Recall: 0.972
  • Weighted Recall: 0.972

Usage

You can use cURL to access this model:

$ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "Elemento di decorazione architettonica a rilievo"}' https://api-inference.huggingface.co/models/davanstrien/autotrain-cultural_heritage_metadata_accuracy-48840118272

You can also use the model locally be leveraging a Transformers pipeline

from transformers import pipeline

pipe = pipeline('text-classification', model='biglam/cultural_heritage_metadata_accuracy')
pipe("Elemento di decorazione architettonica a rilievo")
Downloads last month
29
Safetensors
Model size
110M params
Tensor type
I64
·
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train biglam/cultural_heritage_metadata_accuracy