RNC

API for Russian National Corpus

Installation

pip install rnc

Structure

Corpus object contains list of obtained examples. There are two types of examples:

• If out is normal, API uses normal example, which name is equal to the Corpus class name:

ru = rnc.MainCorpus(...)
ru.request_examples()

print(type(ru[0]))
>>> MainExample

• if out is kwic, API uses KwicExample.

Examples’ objects fields

Usage

import rnc

ru = rnc.MainCorpus(
    query='корпус', 
    p_count=5,
    file='filename.csv',
    marker=str.upper,
    **kwargs
)

ru.request_examples()

• query – one str or dict with tags. Words to find, you should give the vocabulary form of them.
• p_count – count of PAGES.
• file – path to local csv file, optional. Example: file='data\\filename.csv'.
• marker – function, with which found wordforms will be marked, optional.
• kwargs – additional params.

Corpora you can use.

Full query form

query = {
    'word1': {
        'gramm': 'acc', # grammar tags for lexgramm search
        'flags': 'bdot' # additional tags for lexgramm search
    },
    # you can get as a value one string or dict of params
    # params are: any name of dict key, name of tag (you can see them below)  
    'word2': {
        'gramm': { 
            # the NAMES of these keys might be any
            'pos (any name)': 'S' or ['S', 'A'], # one value or list of values,
            'case (any name)': 'acc' or ['acc', 'nom'],
        },
        'flags': {}, # all the same to here
        # distance between first and second words
        'min': 1,  
        'max': 3
    },  
}

corp = rnc.MainCorpus(
    query, 5, file='filename.csv', marker=str.upper, **kwargs)
corp.reques_examples()

Lexgramm search params

String as a query

Also you can pass as a query a string with the vocabulary forms of the words, divided by space: query = 'get down' or query = 'я получить'. Distance between them will be default.

Additional request params

These params are optional, you can ignore them. Here are the default values.

corp = rnc.ParallelCorpus(
    query=query, 
    p_count=5,
    file='filename.csv',
    marker=str.upper,
    
    dpp=5, # documents per page
    spd=10, # sentences per document (<= than spd)
    text='lexgramm' or 'lexform', # way to search
    out='normal' or 'kwic', # output format
    kwsz=5, # if out=kwic, count of words in context
    sort='i_grtagging', # way to sort the results, see HOWTO section below
    mycorp='', # see HOWTO section below
    lang=rnc.Languages.en,
    accent=0, # with accentology (1) or without (0), if it is available
)

Sort keys

API can work with a local file too

ru = rnc.SpokenCorpus(file='local_database.csv') # it must exist
print(ru)

If the file exists, API works with it. If the data list is not empty you cannot request new examples.

If you work with a file, it is not demanded to pass any argument to Corpus except for the file name (file=...).

Working with corpora

corp = rnc.corpus_name(...) 

• corp.request_examples() – request examples. There is an exception if:
  • Data still exist.
  • No results found.
  • A requested page does not exist (if there are 10 pages in the RNC, but you have requested > 10).
  • There is a mistake in the request.
  • You have no access to the Internet.
  • There is a problem while getting access to RNC.
  • another problems…
• corp.data – list of examples (only getter)
• corp.query – query (only getter).
• corp.forms_in_query – requested wordforms (only getter).
• corp.p_count – requested count of pages (only getter).
• corp.file – path to the local csv file (only getter).
• corp.marker – marker (only getter).
• corp.params – dict, HTTP tags (only getter).
• corp.found_wordforms – dict with found wordforms and their frequency (only getter).
• corp.ex_type – type of example (only getter).
• corp.amount_of_docs – amount of docs where the query was found.
• corp.amount_of_contexts – amount of contexts where the query was found.
• corp.graphic_link – link to the graphic of the distribution of query occurrences by years.
• corp.dump() – write two files: csv file with all data and json file with config.
• corp.copy() – create a copy.
• corp.shuffle() – shuffle data list.
• corp.sort_data(key=, reverse=) – sort the list of examples. Here HTTP keys do not work, key is applied to Example objects.
• corp.pop(index) – remove and return the example at the index.
• corp.clear() – empty the data list.
• corp.filter(key) – filter the data list, remove some examples using the key. Key is applied to the Example objects.
• corp.url – URL of the first RNC page (only getter).
• corp.findall(pattern, args) – get all examples where the pattern found and the match.
• corp.finditer(pattern, args) – get all examples where the pattern found and the match.
• async corp.request_examples_async() – make request in the running event loop.

Magic methods:

• corp.dpp or another request param (only getter).
• corp() – all the same to request_examples().

• str(corp) or print(corp) – str with info about Corpus, enumerated examples. By default, Corpus shows first 50 examples, but you can change it or turn the restriction off.

  Info about Corpus:

    Russian National Corpus (https://ruscorpora.ru)
    Class: CorpusName, len = amount of examples 
    Pages: n of 'words' requested
  

• len(corp) – count of examples.
• bool(corp) – whether data exist.
• corp[index or slice] – get element at the index or create a new object with sliced data:

  from_2_to_10 = corp[2:10:2]
  

• del corp[10] or del corp[:10] – remove some examples from the data list.

• Also you can use cycle for. For example we want to see only left context (out=kwic) and source: ```python corp = rnc.ParallelCorpus( ‘corpus’, 5, out=’kwic’, kwsz=7, lang=rnc.Languages.en ) corp.request_examples()

for r in corp: print(r.left) print(r.src)

Set default values to all objects you will create:
* `corpus_name.set_dpp(value)` – change default `document per page` value.
* `corpus_name.set_spd(value)` – change default `sentences per document` value.
* `corpus_name.set_text(value)` – change default search way.
* `corpus_name.set_sort(value)` – change default sort key.
* `corpus_name.set_min(value)` – change default min distance between words.
* `corpus_name.set_max(value)` – change default max distance between words.
* `corpus_name.set_restrict_show(value)` – change default amount of shown examples in print. 
If it is equal to `False`, the Corpus shows all examples. 


### Corpora features
#### ParallelCorpus
* The query might be both in the original language and in the language of 
  translation. 

#### MultilingualParaCorpus
* Working with files is removed.
* Param `mycorp` is not demanded by default, but it might be passed, see 
  **HOWTO** section below.

#### MultimodalCorpus
* `corp.download_all()` – download all media files. **It is recommended** to use 
this method instead of `expl.download_file()`.
* `async corp.download_all_async()` – download all media files using the running event loop.


## Logger
* See all log messages
```python
rnc.set_stream_handler_level('debug')

• See less than all messages

  rnc.set_stream_handler_level('info')
  

• Turn the logger off

  rnc.set_logger_level('critical')
  

• Turn off all messages in the stream, but dump logs to file

  rnc.set_stream_handler_level('critical')
  

• Turn off dumping logs to file

  rnc.set_file_handler_level('critical')
  

ATTENTION

• Do not forget to call this function

  corp.request_examples()
  

• If you have requested more than 10 pages, RNC returns 429 error (Too many requests). For example requesting 100 pages you should wait about 3 minutes:
• Do not call the marker you pass

RIGHT:

ru = rnc.MainCorpus(...,  marker=str.upper)

WRONG:

ru = rnc.MainCorpus(..., marker=str.upper())

• Pass an empty string as a param if you do not want to set them

  query = {
    'word1': '',
    'word2': {'min': 2, 'max': 5}
  }
  

• If accent=1, marker does not work.
• Do not run corp.request_examples() in the running event loop, instead use await corp.request_examples_async()

---

HOWTO

You can ask any question you want here.

How to set sort?

There are some sort keys:

1. i_grtagging – by default.
2. random – randomly.
3. i_grauthor – by author.
4. i_grcreated_inv – by creation date.
5. i_grcreated – by creation date in reversed order.
6. i_grbirthday_inv – by author’s birth date.
7. i_grbirthday – by author’s birth date in reversed order.

Some of HTTP params.

How to set language in ParallelCorpus?

en = rnc.ParallelCorpus('get', 5, lang=rnc.Languages.en)

Languages the corpus supports:

1. Armenian
2. Bashkir
3. Belarusian
4. Bulgarian
5. Buryatian
6. Chinese
7. Czech
8. English
9. Estonian
10. Finnish
11. French
12. German
13. Italian
14. Latvian
15. Lithuanian
16. Polish
17. Spanish
18. Swedish
19. Ukrainian

If you want to search something by several languages, choose and set the mycorp in the site, pass this param to Corpus.

How to set subcorpus?

Means specify the sample where you want to search the query.

There are default keys in rnc.mycorp (working checked in MainCorpus) – Russian writers and poets:

• Pushkin
• Dostoyevsky
• TolstoyLN
• Chekhov
• Gogol
• Turgenev

Example:

ru = rnc.MainCorpus('нету', 1, mycorp=rnc.mycorp['Pushkin'])

OR

ru = rnc.MainCorpus('нету', 1, mycorp=rnc.mycorp.Pushkin)

OR

Links

• Russian National Corpus
• Docs
• Examples’ objects fields
• Corpora you can use.
• Lexgramm search params

• Sort keys

Requirements

• Python >= 3.7

Licence

rnc is offered under MIT licence.

Source code

The project is hosted on Github

---

Please file an issue in the bug tracker if you have found a bug or have some suggestions to improve the library.

This site is open source. Improve this page.