Summarizing MMIF

Code to create a summary of an MMIF file, only keeping those annotations that are useful metadata (and reducing the size of the file by one or two orders of magnitude), and making some implicit relations between annotations and anchors in the source explicit.

See code/run.py for some of the limitations of the code.

Usage

Run the application without using the server by running the summarizer in test mode (using the -t option) on one of the sample input documents: examples/input-v7.mmif and examples/input-v9.mmif. The first has seven manually created views iews from the following applications: bars-and-tone, slates, audio-segmenter, kaldi, east, tesseract and slate-parser. The second has two views added by the app-spacy-nlp application.

$ cd code
$ python run.py -t examples/input-v7.mmif

The output for this should be something like

<?xml version="1.0" encoding="utf-8"?>
<Summary>
  <Documents>
    <Document type="VideoDocument" location="file:///var/archive/video-002.mp4"/>
  </Documents>
  <TimeFrames>
    <TimeFrame start="0" end="2600" frameType="bars-and-tone"/>
    <TimeFrame start="2700" end="5300" frameType="slate"/>
    <TimeFrame start="0" end="5500" frameType="non-speech"/>
    <TimeFrame start="5500" end="22000" frameType="speech"/>
  </Documents>
  <Entities>
  </Entities>
</Summary>

The example outputs for both input examples are in the examples folder.

To start a Flask server:

$ python run.py

To use the server (do this from a different terminal than the run.py script, but from the directory with this readme file):

$ curl -X GET http://0.0.0.0:5000/
$ curl -H "Accept: application/json" -X POST -d@examples/input-v7.mmif http://0.0.0.0:5000/

The first gets you the metadata, the second sends off the file examples/input-v7.mmif to get summarized.

Docker

Building the image (use any name of your liking for the tag):

$ cd code
$ docker build --tag mmif-summarizer .

Entering the container and testing the converter:

$ docker run --rm -it mmif-summarizer bash
$ python run.py -t examples/input-v7.mmif

In real life, the second prompt will be the container prompt, something like root@f9deb9a6115f:/app#.

To run the Flask server in a Docker container:

$ docker run --rm -d -p 5000:5000 mmif-summarizer
$ curl -H "Accept: application/json" -X POST -d@examples/input-v7.mmif http://0.0.0.0:5000/