Services

Coherence estimation web service

The web service (Docker image) is created in order to perform the following processing of Ukrainian texts:

  • Estimation of the coherence of a text (a number of text’s sentences should exceed 3).
  • Extraction of noun phrases.
  • Search for coreferent pairs (test mode).

The web service is implemented as REST API. Data processing is performed using HTTP POST requests to the following endpoints:

  • Estimation of the coherence of a text: /api/get_coherence
  • Extraction of noun phrases: /api/get_phrases
  • Search for coreferent pairs: /api/get_coreferent_clusters

The input format of a request body: {"text": "<:text>"}. The response from the server is returned in JSON format according to the endpoint.

The web service is created by Artem Kramov

Постійне посилання

Ukrainian LLM leaderboard

Leaderboard of large language models on Ukrainian benchmarks: quality metrics and model comparison.

Постійне посилання

Автори
Юрій Панів
Ліцензія
MIT
NER microservice

NER microservice allows the annotation of a tokenized text using models trained with the help of MITIE library for the Ukrainian, Russian, and English languages (depending on what Dockerfile you choose during the launch). This microservice was developed by Mykhailo Chalyi.

Example of an HTTP query:

$ curl -X POST -H "Content-Type: application/json" -d '{ "tokens": ["Несе","Галя","воду",",","Коромисло","гнеться" ]}' http://localhost:8080/ 

Постійне посилання

NLP_UK microservice

Microservice on the basis of NLP_UK library gives an opportunity to lemmatize the input text with the dict_uk dictionary, which also includes tokenization. It was developed by Andriy Rysin.

Example of an HTTP query:

$ curl -X POST -H "Content-Type: application/json" -d "{'text': 'Сьогодні у продажі. 12-те зібрання творів 1969 р. І. П. Котляревського.'}" http://localhost:8080/lemmatize/ 

Постійне посилання

Language identification microservice (WILD)

With the help of the wiki-lang-detect library, the WILD microservice allows identifying the language of the input text out of 156 languages that are used on the Internet.

Example of an HTTP query:

$ curl -X POST -H "Content-Type: application/json" -d "{'text': 'Несе Галя'}" http://localhost:8080/

Постійне посилання

Tokenization microservice

Microservice for Ukrainian text tokenization based on the tokenize-uk library.

Постійне посилання

Автори
Дмитро Чаплинський
Ліцензія
MIT
Посилання
GitHub