Flask Syntax Tree Generator

A Flask-based web application for generating syntax trees from sentences using generative grammar rules, spaCy POS tagging, NLTK parsing, and SVG rendering.

The application allows users to enter a sentence, select a grammar mode, optionally add custom grammar rules, choose a spaCy language model, and generate visual syntax trees together with bracket notation and the full generated grammar.

---

Project Overview

This project combines:

• Flask for the web application
• Gunicorn for production serving
• spaCy for POS tagging and linguistic annotation
• NLTK for context-free grammar parsing
• svgling for SVG syntax tree rendering
• Docker for containerized deployment (Dockerfile included)

The project is useful for demonstrating basic concepts in syntax, generative grammar, POS tagging, ambiguity, relative clauses, conjunctions, and grammar-based parsing.

---

Main Features

• Generate syntax trees from user-provided sentences
• Use a default generative grammar
• Extend the default grammar with additional user-defined rules
• Use a completely custom grammar
• Select English or German spaCy POS tagging
• Display generated syntax trees as SVG graphics
• Display bracket notation for generated trees
• Display the complete grammar used for parsing
• Includes demo sentences for testing
• Docker-ready deployment
• Gunicorn production setup
• Optimized for small server deployments by using one Gunicorn worker

---

Technology Stack

Component	Purpose
Flask	Web framework
Gunicorn	Production WSGI server
spaCy	POS tagging and linguistic annotation
NLTK	Context-free grammar parsing
svgling	Syntax tree SVG rendering
Docker	Containerization

---

Project Structure

flask_syntax_tree_generator/
│
├── app.py
├── requirements.txt
├── Dockerfile
├── docker-compose.yml
│
├── syntax/
│   ├── syntax.py
│   └── syntax_tree_generator.ipynb
│
├── templates/
│   ├── index.html
│   └── syntax.html
│
├── static/
│   ├── favicon.ico
│   └── styles/
│       └── style.css
│
├── .dockerignore
├── .gitignore
└── README.md

---

Core Files

app.py

Main Flask application file.

It defines the web routes:

Route	Purpose
/	Landing page
/syntax	Syntax tree output page
/parse-sentence/<inputData>	Parses input data and generates syntax trees
/favicon.ico	Serves the favicon

The route /parse-sentence/<inputData> extracts:

• sentence
• grammar rules
• grammar option
• language option

Then it calls:

generate_syntaxtree(sentence, grammar_rules, option_grammar, option_lang)

from:

syntax/syntax.py

---

syntax/syntax.py

Contains the core parsing logic.

Main responsibilities:

• Load spaCy language models
• Format sentence input
• POS-tag the input sentence
• Generate POS-based grammar rules
• Combine default grammar and custom rules
• Parse the sentence with NLTK ChartParser
• Render syntax trees as SVG
• Return:
  • syntax tree SVGs
  • bracket notation
  • grammar used for parsing

---

Grammar Modes

The application supports three grammar modes.

1. Default Grammar

Uses the predefined grammar:

S -> NP VP
S -> NP RC
S -> 'If' S 'then' S

VP -> VERB
VP -> VERB NP
VP -> VERB ADJ
VP -> VP PP
VP -> AUX VP
VP -> AUX 'not' VP
VP -> VP CCONJ VP

NP -> NOUN
NP -> DET NOUN
NP -> DET PROPN
NP -> PROPN
NP -> PRON
NP -> NP PP
NP -> NP RC
NP -> NP CCONJ NP

PP -> P NP

RC -> RPRON S

The sentence is POS-tagged by spaCy, and terminal rules are generated automatically.

Example:

He sees a piano

may generate terminal rules such as:

PRON -> 'He'
VERB -> 'sees'
DET -> 'a'
NOUN -> 'piano'

---

2. Default Grammar + Additional Rules

Uses the default grammar and extends it with user-defined rules.

This is useful when the default grammar does not cover a specific construction or terminal category.

Example:

VERB -> 'is'

---

3. Own Grammar

Uses only the grammar rules provided by the user.

Example:

S -> NP VP
NP -> D N
VP -> V NP
D -> 'a'
N -> 'fish'
N -> 'He'
V -> 'eats'

---

Language Options

The application supports POS tagging with spaCy.

Option	spaCy Model
English	en_core_web_sm or en_core_web_lg
German	de_core_news_sm or de_core_news_lg

For production on small servers, the smaller models are recommended:

en_core_web_sm
de_core_news_sm

The large models require significantly more RAM:

en_core_web_lg
de_core_news_lg

---

Installation

Create a virtual environment & Install Depencies:

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

Run the app with Gunicorn:

gunicorn app:app --bind 0.0.0.0:5000 --workers 1 --timeout 120

Open the application:

http://localhost:5000

---

Author

Development by Deniz Inan.

---