evolvedinclusion / text-inflector

Scaffolding for TextBlob and Lemminflect behind FastAPI in a container

text-inflector

Goal: a simple, lightweight service for:

• Generating parts of speech tags for a body of text
• Obtaining the form of a word given a parts of speech tag
• Obtaining basic tokenization of a given body of text (sentences)

This service doesn't have any logic other than to wrap APIs from TextBlob and Lemminflect. This runs textblob and lemminflect as APIs behind FastAPI in a docker container. The supported actions from textblob are tags, and sentences, but more may be added. The supported action from lemminflect is getInflection. Contributions are welcome. Note this doesn't do anything with certs/ssl/tls/https. Setting up a cluster for ssl termination isn't in scope here.

Container

[temporary update] The container logic has been rewritten to craft together a working version on Python 3.9. The previous logic (referenced below) uses the official container image(s) from FastAPI but Python 3.9 hasn't been added yet. As a result, building the container per instructions below will create a Python 3.9-based container. Documentation and code will be returned to the official versions after support is released.

Builds on the FastAPI official container image from https://hub.docker.com/r/tiangolo/uvicorn-gunicorn-fastapi/ per the FastAPI deployment docs for Python 3.6.

Current image tag: python3.7-2020-12-19

Textblob

Textblob is an NLP library written in Python. See textblob docs. From Textblob: "TextBlob is a Python (2 and 3) library for processing textual data. It provides a simple API for diving into common natural language processing (NLP) tasks such as part-of-speech tagging, noun phrase extraction, sentiment analysis, classification, translation, and more."source

Current Textblob version: 0.15.3

Lemminflect

Lemminflect is an NLP library written in Python with a purpose of providing an English word inflection system. From Lemminflect: "LemmInflect uses a dictionary approach to lemmatize English words and inflect them into forms specified by a user supplied Universal Dependencies or Penn Treebank tag. The library works with out-of-vocabulary (OOV) words by applying neural network techniques to classify word forms and choose the appropriate morphing rules."source Lemminflect is one of a handful of libraries that will take a word and a Part of Speech tag and return a best effort at the form of that word corresponding to the supplied Part of Speech. See the lemminflect repo.

Current Lemminflect version: 0.2.2

Run tests

python -m venv env
. env/bin/activate
pip install -r requirements_dev.txt
python -m textblob.download_corpora
pytest

Build/run locally

WARNING DO NOT USE THIS TO BUILD FOR DEPLOY TO AWS SEE README IN EVOLVED REPO FOR INSTRUCTIONS

docker build -t text-inflector-image .
docker run -d --name text-inflector -p 1234:80 text-inflector-image // (replace 1234 with the port you want the container to expose)

Listen port

By default w/no intervention the service will listen on port 80 and expose port 80 on the container. If run locally, map 80 to a different external port to avoid conflicts. To set a different listen port on the service and expose a different port from the container, pass the API_LISTEN_PORT build argument with a port value to docker build. This is required when building for non-local deployment.

Stop/destroy

docker stop text-inflector
docker rm text-inflector

Endpoints

Tags

Get the Parts of Speech tags for a given body of text. See the Penn treebank parts of speech for a map of tags/meanings.

Endpoint: /tags
Action: POST
Example JSON body: '{"text": "I am a pear" }'
Expected Response:
  { "tags": [["I", "PRP"], ["am", "VBP"], ["a", "DT"], ["pear", "NN"]] }

Example:

curl -X POST -H "Content-Type: application/json" -d '{"text": "I am a pear"}' http://localhost:1234/tags

Inflections

Get the inflected form of a word corresponding to a Part of Speech tag. Note that the supplied JSON body has required key (object) names (word and pos) per FastAPIs Body module.

Endpoint: /inflections
Action: POST
Example JSON body: '{
  "word": {  "text": "pear" },
  "pos": { "tag": "NNS" }
}'
Expected Response:
  {'inflection': ['pears']}

Example:

curl -X POST -H "Content-Type: application/json" -d '{ "word": {  "text": "pear" }, "pos": { "tag": "NNS" } }' http://localhost:1234/inflections

Sentences

Get a list of sentences (strings) in a given body of text Endpoint: /sentences Action: POST Example JSON body: '{ "text": "I am a pear." }' Expected Response: '{ "sentences": ["I am a pear."] }'

Example:

curl -X POST -H "Content-Type: application/json" -d '{"text": "I am a pear."}' http://localhost:1234/tokenizations