---
datasets:
- svilens/auto-tires-ner
language:
- en
- bg
base_model:
- knowledgator/gliner-multitask-large-v0.5
tags:
- ner
- automotive
- tires
---

# Overview
This is a named entity recognition model based on `knowledgator/gliner-multitask-large-v0.5`, specifically fine-tuned for automotive tire attributes extraction. The fine-tuning dataset consists of queries in Bulgarian.

# How to use

## Installation

```
pip install gliner
```

## Usage
The example below includes all labels that the model was fine-tuned on. Although it supports extracting unseen entities, the accuracy of any other labels is not guaranteed.

```python
from gliner import GLiNER

model = GLiNER.from_pretrained("svilens/gliner-multitask-large-v0.5-auto-tires")

text = "Търся зимни гуми 225/65 R17 91V runflat за Honda CR-V 2022"

labels = [
    "tyre_size", "tyre_type", "tyre_width",
    "tyre_aspect_ratio", "tyre_diameter", 
    "tyre_load_index", "tyre_speed_rating",
    "tyre_type", "tyre_season",
    "car_make", "car_model", "car_year"
]

entities = model.predict_entities(text, labels)

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

Output:
```
зимни => tyre_season
225 => tyre_width
65 => tyre_aspect_ratio
17 => tyre_diameter
runflat => tyre_type
Honda => car_make
CR-V => car_model
2022 => car_year
```

## Known issues
- As shown above, load index and speed rating are sometimes not properly recognized. More training examples with these two attributes are required.
- The training data contains `tyre_size` entity that captures the combination of width, ratio and diameter, but these three are additionally provided as standalone entities. Therefore, the model can output either, or both entities, and this inconsistency might cause some inconvenience.