All case studies

Natural language processingResearch portfolio project

Suggest several genres from one story description

A story can belong to more than one genre. This NLP experiment learns from summaries and genre labels so it can suggest several relevant categories.

Contribution
Mehrab Mashrafi
Scope
Preparation, multi label classification and inference
Project context
Original portfolio experiment

Let the label match how stories are organised

A single category would flatten stories that combine several genres. The project collected summaries and labels from Royal Road and treated prediction as a multi label task.

Mehrab prepared the inputs, trained a transformer based model and documented evaluation and inference in notebooks.

Keep suggestions open to editorial review

Predicted genres can help organise a catalogue, but they do not replace an editor’s judgment.

The project also collected cover colour information as a separate exploratory dataset. That is not evidence that the delivered genre model can design a cover or choose its colours.

How the work connects

  1. 01

    Story summaries

  2. 02

    Genre labels

  3. 03

    Text preparation

  4. 04

    Transformer training

  5. 05

    Genre suggestions

What this work establishes

The project documents a multi label classifier and an inference workflow using Hugging Face Transformers and Fastai.

The model reflects its training summaries and labels. Another catalogue, language or taxonomy needs a separate evaluation.

Inspect the work or discuss the approach

Have a similar workflow to improve?

Tell us who uses it, which records it depends on and where the handoff slows down. We can discuss a practical implementation scope.

Discuss your use case

Project notes from Mehrab Mashrafi and AI Byters. Updated .