---
license: mit
---

This is the Placing the Holocaust's finetuned GliNER small model. GLiNER is a Named Entity Recognition (NER) model capable of identifying any entity type using a bidirectional transformer encoder (BERT-like). It provides a practical alternative to traditional NER models, which are limited to predefined entities, and Large Language Models (LLMs) that, despite their flexibility, are costly and large for resource-constrained scenarios.

## Links

* Original GliNER model: https://huggingface.co/urchade/gliner_small-v2.1
* Finetuned with this data: https://huggingface.co/datasets/placingholocaust/spacy-project
* GliNER paper: https://arxiv.org/abs/2311.08526
* GliNER Repository: https://github.com/urchade/GLiNER

## Labels

| Category                  | Definition                                                                                                                                                                                                                                 | Examples                                                                                         |
|---------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------|
| **building**              | Includes references to physical structures and places of labor or employment like factories. Institutions such as the "Judenrat" or "Red Cross" are also included.                                                                         | school, home, house, hospital, factory, station, office, store, synagogue, barracks              |
| **country** | Mostly country names, also includes "earth," "country," and "world." Distinguished from Region and Environmental feature based on context.                                                                                                  | germany, poland, states, israel, united, country, america, england, france, russia                |
| **dlf (distinct landscape feature)**     | Places not large enough to be a geographic or populated region but too large to be an Object, includes parts of buildings like "roof" or "chimney."                                                                                        | street, door, border, line, farm, window, streets, road, wall, field                            |
| **env feature (environmental feature)** | Any named or unnamed environmental feature, including bodies of water and landforms. General references like "nature" and "water" are included.                                                                                             | woods, forest, river, mountains, ground, trees, water, tree, mountain, sea                       |
| **interior space**        | References to distinct rooms within a building, or large place features of a building like a "factory floor."                                                                                                                             | room, apartment, floor, kitchen, rooms, gas, basement, bathroom, chambers, bunker                |
| **imaginary**    | Difficult terms that are context-dependent like "inside," "outside," or "side." Also includes unspecified locations like "community," and conceptual places like "hell" or "heaven."                                                      | place, outside, places, side, inside, hiding, hell, heaven, part, spot                          |
| **populated place**       | Includes cities, towns, villages, and hamlets or crossroads settlements. Names of places can be the same as a ghetto, camp, city, or district.                                                                                             | camp, ghetto, town, city, auschwitz, camps, new, york, concentration, village                   |
| **region**                | Sub-national regions, states, provinces, or islands. Includes references to sides of a geopolitical border or military zone.                                                                                                               | area, side, land, siberia, new, zone, jersey, california, russian, eastern                      |
| **spatial object**               | Objects of conveyance and movable objects like furniture. In specific contexts, refers to transportation vehicles or items like "ovens," where the common use case of the term prevails.                                                   | train, car, ship, boat, bed, truck, trains, cars, trucks                                        |


## Installation
To use this model, you must install the GLiNER Python library:

```bash
!pip install gliner
```
         
## Usage
Once you've downloaded the GLiNER library, you can import the GLiNER class. You can then load this model using `GLiNER.from_pretrained` and predict entities with `predict_entities`.

```python
from gliner import GLiNER

model = GLiNER.from_pretrained("placingholocaust/gliner_small-v2.1-holocaust")

text = """
Okay. So now it's spring of '44? A: ‘4, And she says, You're going to go to Brzezinka. I said, What is Brzezinka? She said, It's a crematorium and the gas chamber. They have a half a million Hungarian Jews are coming in. That's when the time they -- and they need people to select. We do not select the people to -- who die or not. The women fold the clothes and look for jewelry and make packages to send it to Germany.
"""

labels = ["dlf", "populated place", "country", "region", "interior space", "env feature", "building", "spatial object"]

entities = model.predict_entities(text, labels)

for entity in entities:
    print(entity["text"], "=>", entity["label"])
```