e-magyar text processing system (emtsv)

• inter-module communication via the xtsv framework
  • processing can be started or stopped at any module
  • module dependency checks before processing
  • easy to add new modules
  • multiple alternative modules for some tasks
• easy to use command-line interface
• convenient REST API with simple web frontend
• Python library API
• Docker image and runnable docker form

If a bug is found please leave feedback with the exact details.

Citing and License

See docs/cite.bib for BibTeX entries.

If you use the emtsv system, please cite:

Simon Eszter, Indig Balázs, Kalivoda Ágnes, Mittelholcz Iván, Sass Bálint, Vadász Noémi. Újabb fejlemények az e-magyar háza táján. In: Berend Gábor, Gosztolya Gábor, Vincze Veronika (szerk.): MSZNY 2020, XVI. Magyar Számítógépes Nyelvészeti Konferencia (MSZNY 2020). Szeged: Szegedi Tudományegyetem Informatikai Tanszékcsoport, 29-42.

Balázs Indig, Bálint Sass, Eszter Simon, Iván Mittelholcz, Noémi Vadász, and Márton Makrai: One format to rule them all – The emtsv pipeline for Hungarian. In: Proceedings of the 13th Linguistic Annotation Workshop. Association for Computational Linguistics, 2019, 155-165.

Indig Balázs, Sass Bálint, Simon Eszter, Mittelholcz Iván, Kundráth Péter, Vadász Noémi. emtsv – Egy formátum mind felett. In: Berend Gábor, Gosztolya Gábor, Vincze Veronika (szerk.): MSZNY 2019, XV. Magyar Számítógépes Nyelvészeti Konferencia (MSZNY 2019). Szeged: Szegedi Tudományegyetem Informatikai Tanszékcsoport, 235-247.

Váradi Tamás, Simon Eszter, Sass Bálint, Mittelholcz Iván, Novák Attila, Indig Balázs: e-magyar – A Digital Language Processing System. In: Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), 1307-1312.

Váradi Tamás, Simon Eszter, Sass Bálint, Gerőcs Mátyás, Mittelholcz Iván, Novák Attila, Indig Balázs, Prószéky Gábor, Farkas Richárd, Vincze Veronika: Az e-magyar digitális nyelvfeldolgozó rendszer. In: MSZNY 2017, XIII. Magyar Számítógépes Nyelvészeti Konferencia, Szeged: Szegedi Tudományegyetem Informatikai Tanszékcsoport, 49-60.

If you use emMorph module, please cite:

Attila Novák, Borbála Siklósi, Csaba Oravecz: A New Integrated Open-source Morphological Analyzer for Hungarian. In: Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), Paris: European Language Resources Association (ELRA).

Attila Novák: A New Form of Humor -- Mapping Constraint-Based Computational Morphologies to a Finite-State Representation. In: Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC 2014), Paris: European Language Resources Association (ELRA).

If you use emTag (PurePos) module, please cite:

Orosz György, Novák Attila: PurePos 2.0: a hybrid tool for morphological disambiguation. In: Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2013), 539-545.

If you use emBERT module, please cite:

Nemeskey Dávid Márk: Egy emBERT próbáló feladat. In: MSZNY 2020, XVI. Magyar Számítógépes Nyelvészeti Konferencia. Szeged: Szegedi Tudományegyetem Informatikai Tanszékcsoport, 409-418.

If you use emPreverb or emCompound module, please cite:

Pethő Gergely, Sass Bálint, Kalivoda Ágnes, Simon László, Lipp Veronika: Igekötő-kapcsolás. In: MSZNY 2022, XVIII. Magyar Számítógépes Nyelvészeti Konferencia. Szeged: Szegedi Tudományegyetem Informatikai Intézet, 77-91.

The emtsv system is a replacement for the original https://github.com/nytud/hunlp-GATE system.

emtsv is licensed under LGPL 3.0. The submodules have their own licenses.

Requirements

• GIT LFS (see notes)
• Python 3.5 <=
• HFST 3.13 <=
• Hunspell (libhunspell-dev, hunspell-hu) 1.6.2 <=
• OpenJDK 11 JDK (not JRE)
• An UTF-8 locale must be set.

Installation

• From the git repository: see detailed instructions in the documentation
• From prebuilt docker image (https://hub.docker.com/r/mtaril/emtsv):

  docker pull mtaril/emtsv:latest
  

  • See detailed instructions for building the docker image

Usage

Here we present the usage scenarios. The individual modules are documented in detail below. To extend the toolchain with new modules just add new modules to config.py.

#1 – by REST API

Start the emtsv server

• (recommended way) Run the docker image to provide a REST API service for emtsv

  docker run --rm -p5000:5000 -it mtaril/emtsv  # REST API starts listening at http://127.0.0.1:5000
  

  The container starts two emtsv processes by default. Should the throughput be insufficient (or conversely, the memory requirements too great even with two processes), this number can be configured via the EMTSV_NUM_PROCESSES variable:

  docker run --rm -p5000:5000 -it -e "EMTSV_NUM_PROCESSES=4" mtaril/emtsv
  

• (only for development) Flask debug server (single threaded, one request at a time):

  # Without parameters!
  python3 ./main.py
  

  When the server outputs a message like * Running on then it is ready to accept requests at http://127.0.0.1:5000 . We do not recommend using this method in production as it is built atop of Flask debug server! Please consider using the Docker image for REST API in production!

• Any wsgi server (uwsgi, gunicorn, waitress, etc.) can be configured to run with docker/emtsvREST.wsgi .

Use the emtsv server

Given that the server is running:

• by the simple web frontend: point your browser simply to http://127.0.0.1:5000 (without any URL parameters!).
• by curl:

  echo "A kutya elment sétálni." | curl -F "file=@-" http://127.0.0.1:5000/tok/morph/pos
  

• from Python:

  >>> import requests
  >>> # With input file
  >>> # The URL contains the tools to be run separated by `/` instead of ',' used in the CLI
  >>> r = requests.post('http://127.0.0.1:5000/tok/morph/pos', files={'file': open('tests/test_input/input.test', encoding='UTF-8')})
  >>> print(r.text)
  ...
  >>> # With input text
  >>> r = requests.post('http://127.0.0.1:5000/tok/morph/pos', data={'text': 'A kutya elment sétálni.'})
  >>> print(r.text)
  ...
  >>> # CoNLL style comments can be enabled per request (disabled by default):
  >>> r = requests.post('http://127.0.0.1:5000/tok/morph/pos', files={'file':open('tests/test_input/input.test', encoding='UTF-8')}, data={'conll_comments': True})
  >>> print(r.text)
  ...
  

  The server checks whether the module order with the provided input data is feasible, and returns an error message if there are any problems.

#2 – directly by CLI

• Multiple modules at once (not necessarily starting with raw text):

  echo "A kutya elment sétálni." | python3 ./main.py tok,spell,morph,pos,conv-morph,dep,chunk,ner
  

• Modules glued together one by one with the standard *nix pipelines where users can interact with the data between the modules:

  echo "A kutya elment sétálni." | \
     python3 main.py tok | \
     python3 main.py spell | \
     python3 main.py morph | \
     python3 main.py pos | \
     python3 main.py conv-morph | \
     python3 main.py dep | \
     python3 main.py chunk | \
     python3 main.py ner
  

• Independently from the other options, emtsv can also be used with input or output streams redirected or with string input (this applies to the runnable docker form as well):

  python3 ./main.py tok,spell,morph,pos,conv-morph,dep,chunk,ner -i input.txt -o output.txt
  python3 ./main.py tok,spell,morph,pos,conv-morph,dep,chunk,ner --text "A kutya elment sétálni."
  

• Using the docker image in CLI:

  cat input.txt | docker run -i mtaril/emtsv tok,morph,pos > output.txt
  

#3 – as Python Library

1. Install emtsv in emtsv directory or make sure the emtsv installation is in the PYTHONPATH environment variable
2. import emtsv
3. Now the full API is accessible (see the example above):

   import sys
   from emtsv import build_pipeline, jnius_config, tools, presets, process, pipeline_rest_api, singleton_store_factory
   
   jnius_config.classpath_show_warning = False  # To suppress warning
   
   # Imports end here. Must do only once per Python session
   
   # Set input from any stream or iterable and output stream...
   input_data = sys.stdin
   output_iterator = sys.stdout
   # Raw, or processed TSV input list and output file...
   # input_data = iter(['A kutya', 'elment sétálni.'])  # Raw text line by line
   # Processed data: header and the token POS-tag pairs line by line
   # input_data = iter([['form', 'xpostag'], ['A', '[/Det|Art.Def]'], ['kutya', '[/N][Nom]'], ['elment', '[/V][Pst.NDef.3Sg]'], ['sétálni', '[/V][Inf]'], ['.', '.']])
   # output_iterator = open('output.txt', 'w', encoding='UTF-8')  # File
   # input_data = 'A kutya elment sétálni.'  # Or raw string in any acceptable format.
   
   # Select a predefined task to do or provide your own list of pipeline elements
   used_tools = ['tok', 'morph', 'pos']
   
   conll_comments = True  # Enable the usage of CoNLL comments
   
   # Run the pipeline on input and write result to the output...
   output_iterator.writelines(build_pipeline(input_data, used_tools, tools, presets, conll_comments))
   
   # Alternative: Run specific tool for input streams (still in emtsv format).
   # Useful for training a module (see Huntag3 for details):
   output_iterator.writelines(process(sys.stdin, an_inited_tool))
   
   # Or process individual tokens further... WARNING: The header will be the first item in the iterator!
   for tok in build_pipeline(input_data, used_tools, tools, presets, conll_comments):
       if len(tok) > 1:  # Empty line (='\n') means end of sentence
           form, xpostag, *rest = tok.strip().split('\t')  # Split to the expected columns
   
   # Alternative2: Flask application (REST API)
   singleton_store = singleton_store_factory()
   app = application = pipeline_rest_api(name='e-magyar-tsv', available_tools=tools, presets=presets,
                               conll_comments=conll_comments, singleton_store=singleton_store,
                               form_title='e-magyar text processing system',
                               doc_link='https://github.com/nytud/emtsv')
   # And run the Flask debug server separately
   app.run()
   

The public API is equivalent to the xtsv API

Data format

Please see the specification in detail in the xtsv documentation.

For examples see files in tests/test_input and tests/test_output directories.

Modules

Modules are defined in config.py. The current toolchain consists of the following modules which can be called by their names (or using their shorthand names in brackets):

• emToken (tok): Tokenizer
• emMorph (morph): Morphological analyser together with emLem lemmatiser
• hunspell (spell): Spellchecker, stemmer and morphological analyser
• emTag (pos): POS-tagger
• emChunk (chunk): Maximal NP-chunker
• emNER (ner): Named-entity recogniser
• emmorph2ud (conv-morph): Converter from emMorph code to UDv1 upos and feats format
• emmorph2ud2 (conv-morph2): Converter from emMorph code to UDv2 upos and feats format
• emDep-ud (dep): Dependency parser
• emDep-ud50 (dep50): Dependency parser