Sadedegel

**ADD**: `/doc/statistics` service to calculate various document metrics to be used by [sadedeGel Chrome Extension](https://github.com/GlobalMaksimum/sadedegel-chrome-extension.git)

* **ADD**: `sadedegel.server` is now hosted on [Heroku](https://sadedegel.herokuapp.com/docs)
* **UPDATE**: We have significantly improved our `sadedegel.server` services.
* **ADD**: `wpm` (word per minute) based duration calculation to calculate total number of sentences to be filtered out by summarizer services.
* **ADD**: Several service routes
* `/api/info`: Returning `sadedegel` metadata information
* `/`: Redirection request to [sadedegel.ai](http://sadedegel.ai)
* **UPDATED**: `/api/summarizer/random`: New route for `RandomSummarizer`
* **UPDATED**: `/api/summarizer/firstk`: New route for firstK (`PositionSummarizer`)
* `/api/summarizer/rouge1`: First non-baseline summarizer. Rouge1Summarizer is an unsupervised summarizer using rouge1 score of sentences to obtain a score for each sentence in a document.
* **UPDATE**: `sadedegel` now supports Python 3.6+ because of [`fastapi`](https://fastapi.tiangolo.com/) dependency

**ADD**: Github Action for master branch tests all supported Python versions.
**ADD**: [codeconv](https://codecov.io/gh/globalmaksimum/sadedegel) badge is added.
**ADD**: Features are documented.
**ADD**: Extended dataset (`sadedegel.dataset.extended`) now has more 35K documents from various resources.
**UPDATE**: Extraction based summarizers now share a common prototype returning a score for each sentences in a given document.

* CORS added to sadedegel server
* Test cases for sadedegel server

Integrating sadedegel with Github Actions revealed several issues which we haven't detected due to lack of a CI flow

**ADD:** Github Actions integration.
**ADD:** sadedeGel HTTP server is introduced.
**ADD:** ROUGE1 score is added into `sadedegel.metrics`.
**ADD:** `CONTRIBUTING.md`, borrowed from SpaCy project, shows our guidelines for contributing sadedeGel.
**FIX:** Missing development and production dependencies are added.
**FIX:** NLTK ML based model dowload before unit testing.

**ADD:** ML based Sentence Boundary Detection (SBD) achieves an IoU score of `0.8946` (previously used Regular Expression rule based detector achieves `0.7224`)
**ADD:** `-m sadedegel.tokenize evaluate` to evaluate a sbd.
**ADD:** `-m sadedegel.tokenize diff` to analyze tokenization errors between model annotated (sents) dataset.
**ADD:** `-m sadedegel.tokenize train` to train a new ML based sbd.
**FIX:** Performance improvement in loading the summarizer pipeline caused by `AutoTokenizer`
**ADD:** raw corpus cleaner.
**ADD:** optional `base_path` parameter for raw and sent corpus loader.
**FIX:** Lot's of dataset issues on raw and sent corpus
**ADD:** `-m sadedegel.dataset validate` for corpus sanity check.