nlp/README.md

# NLP - Natural Language Processing

This software implements a heavily parallelized pipeline for Natural Language Processing of text files. It is used for nopaque's NLP service but you can also use it standalone, for that purpose a convenient wrapper script is provided.

## Software used in this pipeline implementation
- Official Debian Docker image (buster-slim) and programs from its free repositories: https://hub.docker.com/_/debian
- pyFlow (1.1.20): https://github.com/Illumina/pyflow/releases/tag/v1.1.20
- spaCy (3.0.3): https://github.com/tesseract-ocr/tesseract/releases/tag/4.1.1
- spaCy medium sized models (3.0.0):
  - https://github.com/explosion/spacy-models/releases/tag/da_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/de_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/el_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/en_core_web_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/es_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/fr_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/it_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/nl_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/pt_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/ru_core_news_md-3.0.0
  - https://github.com/explosion/spacy-models/releases/tag/zh_core_web_md-3.0.0


## Use this image

1. Create input and output directories for the pipeline.
``` bash
mkdir -p /<my_data_location>/input /<my_data_location>/output
```

2. Place your text files inside `/<my_data_location>/input`. Files should all contain text of the same language.

3. Start the pipeline process. Check the [Pipeline arguments](#pipeline-arguments) section for more details.
```
# Option one: Use the wrapper script
## Install the wrapper script (only on first run). Get it from https://gitlab.ub.uni-bielefeld.de/sfb1288inf/nlp/-/raw/1.0.0/wrapper/nlp, make it executeable and add it to your ${PATH}
cd /<my_data_location>
nlp -i input -l <language_code> -o output <optional_pipeline_arguments>

# Option two: Classic Docker style
docker run \
    --rm \
    -it \
    -u $(id -u $USER):$(id -g $USER) \
    -v /<my_data_location>/input:/input \
    -v /<my_data_location>/output:/output \
    gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/nlp:1.0.0 \
        -i /input \
        -l <language_code>
        -o /output \
        <optional_pipeline_arguments>
```

4. Check your results in the `/<my_data_location>/output` directory.
```

### Pipeline arguments

`--check-encoding`
* If set, the pipeline tries to automatically determine the right encoding for
your texts. Only use it if you are not sure that your input is provided in UTF-8.
* default = False
* required = False

`-l languagecode`
* Tells spaCy which language will be used.
* options = da (Danish), de (German), el (Greek), en (English), es (Spanish), fr (French), it (Italian), nl (Dutch), pt (Portuguese), ru (Russian), zh (Chinese)
* required = True

`--nCores corenumber`
* Sets the number of CPU cores being used during the NLP process.
* default = min(4, multiprocessing.cpu_count())
* required = False

``` bash
# Example with all arguments used
docker run \
    --rm \
    -it \
    -u $(id -u $USER):$(id -g $USER) \
    -v "$HOME"/ocr/input:/input \
    -v "$HOME"/ocr/output:/output \
    gitlab.ub.uni-bielefeld.de:4567/sfb1288inf/nlp:1.0.0 \
        -i /input \
        -l en \
        -o /output \
        --check-encoding \
        --nCores 8 \
```