RTC-NER

rtc-ner is a domain-specific Named Entity Recognition (NER) model developed for extracting road traffic crash (RTC)-related geographic, medical, and incident entities from unstructured textual road traffic crash reports.

The model is designed to transform narrative RTC reports into structured information that can subsequently support road traffic crash analysis, geospatial processing, and emergency-response research.

Model Description

rtc-ner is a transformer-based spaCy NER model trained for the identification of domain-specific entities occurring in road traffic crash narratives.

The model focuses particularly on extracting geographic locations and selected incident-related entities that are relevant to the analysis and geospatial structuring of RTC reports.

Entity Labels

The model recognizes the following entity types:

Entity	Description
ROAD	Road, highway, street, expressway, or other road-related location
LANDMARK	Named or identifiable landmark mentioned in an RTC report
SUBURB	Suburb or locality within a larger urban area
TOWN	Town or city/locality
LGA	Local Government Area
STATE	Nigerian state or other state-level administrative location
HOSPITAL	Hospital or healthcare facility mentioned in the report
CASUALTY	Person reported as a casualty in the crash
INJURED	Person reported as injured in the crash

Intended Use

rtc-ner is intended for research and information-extraction applications involving road traffic crash narratives, including:

• Extraction of geographic entities from RTC reports
• Structuring unstructured crash narratives
• Identification of locations for subsequent geocoding
• Road traffic crash database construction
• Road safety research
• Natural language processing of RTC reports
• Preparation of textual RTC data for geospatial analysis
• Research on emergency response and road traffic crash management

Model Architecture

The model was developed using the spaCy NLP framework with a transformer-based architecture.

The repository contains the trained spaCy pipeline and its associated configuration, vocabulary, tokenizer, and model components.

Model Files

The repository contains the complete trained spaCy model, including:

config.cfg
meta.json
tokenizer
vocab/
ner/
tok2vec/

These files should be retained together when loading the model.

Usage

Requirements

Install spaCy:

pip install spacy

Loading the Model

After downloading or cloning this repository, the model can be loaded using spaCy.

For example:

import spacy

nlp = spacy.load("path/to/rtc-ner")

text = "A road traffic crash occurred on Lokoja-Abuja Road near a major the Niger bridge in Koton-Karfe."

doc = nlp(text)

for ent in doc.ents:
    print(ent.text, ent.label_)

The output contains the entities identified by the model together with their corresponding entity labels.

Example

Given an RTC narrative containing a road, landmark, town, state, hospital, or injury-related information, the model attempts to identify and classify the relevant spans according to its nine domain-specific entity categories.

Limitations

rtc-ner is a domain-specific research model and should not be assumed to generalize perfectly to all forms of text.

Performance may be affected by:

• Differences in writing style between training and application data
• Spelling variations and typographical errors
• Ambiguous place names
• Unreported or implicitly mentioned locations
• Differences between news reports and other forms of RTC narratives
• Geographic names that have multiple interpretations
• Domain shifts outside Nigerian road traffic crash reporting

Extracted geographic entities should therefore be reviewed or validated before being used for high-precision geospatial applications.

Research Context

rtc-ner was developed as part of research investigating the transformation of unstructured road traffic crash narratives into structured, machine-readable information for road safety and emergency-response applications.

The model is intended to serve as an information-extraction component within a broader research workflow involving RTC data curation, geospatial processing, and artificial intelligence.

Citation

If you use rtc-ner in academic research, please cite the associated research publication and dataset from which the model was developed.

[1] P. O. Idakwo, O. Adekanmbi, A. Soronnadi, and A. David, “Geo-parsing and analysis of road traffic crash incidents for data-driven emergency response planning,” Heliyon, vol. 11, no. 4, p. e41067, 2025, doi: 10.1016/j.heliyon.2024.e41067.

[2] P. O. Idakwo, O. Adekanmbi, and A. David, “Nigerian Multi-modal Road Traffic Crash Data,” 2026, doi: 10.5281/ZENODO.15862127.

Disclaimer

This model is provided for research purposes. Predictions generated by the model should be treated as automated information-extraction results and should be independently validated before being used in operational, safety-critical, medical, or emergency-response decision-making.