MoodCat😼 (Sentence Mood Classifier for English)

This repository shows how to train a sentence mood classifier using spaCy's new SpanCategorizer component and the Georgetown University Multilayer (GUM) Corpus. The classifier uses a custom span suggester, which returns sentences for classification.

Please note that this repository is only for demonstration. The GUM corpus is too small for training a classifier from scratch and some labels are very rare. The classifier does a decent job with declaratives and interrogatives, but struggles with imperatives and rarer moods.

For information on classifier performance, see the file training/metrics.json.

📋 project.yml

The project.yml defines the data assets required by the project, as well as the available commands and workflows. For details, see the spaCy projects documentation.

⏯ Commands

The following commands are defined by the project. They can be executed using spacy project run [name]. Commands are only re-run if their inputs have changed.

Command	Description
convert	Convert the CoNLL-U data to spaCy's binary format
debug	Debug the data for insights on the corpus
train	Train the model for sentence mood classification
evaluate	Evaluate the model and export metrics
package	Package the trained model as a pip package

⏭ Workflows

The following workflows are defined by the project. They can be executed using spacy project run [name] and will run the specified commands in order. Commands are only re-run if their inputs have changed.

Workflow	Steps
all	convert → train → evaluate → package

🗂 Assets

The following assets are defined by the project. They can be fetched by running spacy project assets in the project directory.

File	Source	Description
assets/gum	Git	The Georgetown University Multilayer (GUM) Corpus

🤔 How to run the demo?

1. Run the command python setup.py install in the directory packages/en_moodcat-0.0.1 to install the pipeline
2. Run the file moodcat_demo.py