Skip to content

Latest commit

 

History

History
383 lines (309 loc) · 14.2 KB

CHANGELOG.md

File metadata and controls

383 lines (309 loc) · 14.2 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

[Unreleased]

Added

Changed

Deprecated

Removed

Fixed

Security

[0.19.1] - 2024-03-29

Fixed

  • Handled the duration mark (e.g., [# 0.4]) in utterance cleaning.

[0.19.0] - 2023-12-13

Added

  • Added support for Python 3.12.
  • Handled pre-clitics and post-clitics from %mor tiers and honored their distinction in the parsed utterance.

[0.18.0] - 2023-03-11

Added

  • Added support for Python 3.11.

Changed

  • Updated the test data from Brown's Eve from the upstream CHILDES.

Removed

  • Dropped support for Python 3.7.

[0.17.0] - 2022-06-09

Added

  • Added the exclude_switch option for MLU (mlu(), mlum(), and mluw()), so that words with @s for switching language may be excluded.

Fixed

  • Fixed MLU computation (mlu(), mlum(), and mluw()):
    • If xxx, yyy, or www appears in an utterance, the whole utterance is ignored.
    • If there are no MLU-relevant words/morphemes in an utterances, the whole utterance is ignored.

[0.16.2] - 2022-03-17

Fixed

  • Moved download_and_extract_brown test function to under the pylangacq package namespace, as tests from BaseTestCHATReader require downloaded CHAT data files.

[0.16.1] - 2022-03-17

Changed

  • Restructured the repository to use top-level src/ and tests/ directories.

Removed

  • Removed setup.py.

Fixed

  • Moved BaseTestCHATReader back under the pylangacq package namespace so that downstream packages can import BaseTestCHATReader for testing.

[0.16.0] - 2021-12-27

Added

  • Reader objects can now be concatenated by the addition operator +.
  • Implemented the head, tail, and info methods at Reader.
  • Added support for Python 3.10.
  • Turned on Windows testing on CircleCI.
  • Added pyproject.toml. Related to prioritizing setup.cfg for specifying build metadata and options.

Changed

  • The to_strs and to_chat methods of a Reader object return tabulated outputs by default.
  • Prioritized to_chat for the single file output use case.
  • Unzipping CHAT data now uses less memory.
  • Switched to setup.cfg to fully specify build metadata and options, while keeping a minimal setup.py for backward compatibility. Related to the new pyproject.toml.
  • Switched the Sphinx docs theme from sphinx-rtd-theme to furo.

Removed

  • Dropped support for Python 3.6.

Security

  • Turned on safety and bandit checks at CircleCI builds.

[0.15.0] - 2021-06-06

Added

  • Reader.from_zip (also read_chat) now keeps the downloaded ZIP archive in a non-temporary directory for possible re-use.
    • Added the kwarg use_cached in Reader.from_zip, so that we use the cached data by default for the same input URL, and that we can force re-downloading by setting use_cached to False.
    • Added the kwarg session in Reader.from_zip, in case using a customized requests.Session instance is desired. session also makes it possible to write tests for the new kwarg use_cached.
    • Added the helper functions cached_data_info and remove_cached_data.
  • Reader has the new to_strs method that yields CHAT data strings.
  • Reader has the new to_chat method that exports data to local files.

Changed

  • CHAT parsing for the header information is now more robust for varying whitespace characters between the head and its associated value.

Removed

  • Dropped kwarg allow_remote in Reader.from_zip. This kwarg wouldn't make any sense anymore, or at least would be confusing with the introduction of use_cached.

[0.14.1] - 2021-05-16

Fixed

  • The header/metadata has a more reasonable representation for emptiness when input data is empty.

[0.14.0] - 2021-05-12

Added

  • Added the parallel optional argument to the Reader methods {from_zip, from_dir, from_files, from_strs} so that parallelization can be turned off if desired.
  • Added the filter method to Reader for filtering data by file paths.

[0.13.3] - 2021-05-07

Fixed

  • The methods append, append_left, extend, and extend_left now work with a subclass of Reader, not just Reader itself.

[0.13.2] - 2021-05-02

Fixed

  • Fixed utterance cleaning so that it is now compatible with all CHILDES datasets.

[0.13.1] - 2021-03-23

Fixed

  • Fixed a CHAT parsing issue when correction and repetition are combined.

[0.13.0] - 2021-03-15

API-breaking changes: The Reader class has been completely rewritten. A couple methods have been removed, while others have been renamed. For methods that remain (renamed or not), their behavior for output data structure and arguments allowed has been changed. The details are in the following.

Added

  • New classmethods of Reader for reader instantiation:
    • from_zip
    • from_dir
  • New classes to better structure CHAT data:
    • Utterance
    • Token
    • Gra
  • New Reader methods:
    • append_left, extend, extend_left, pop, pop_left
    • tokens (which gives Token objects, essentially the "tagged words" from before)
  • In the header dictionary, each participant's info has the new key "dob" for date of birth (if the info is available in the CHAT header). The corresponding value is a datetime.date object. (The same info was previously exposed as the Reader method date_of_birth, now removed.)
  • The test suite now covers code snippets in both the docstrings and .rst doc files.

Changed

  • CHAT parsing in Reader instantiation has been completely rewritten. The previous private class _SingleReader has been removed. This private class duplicated a lot of the Reader code, which made it hard to make changes.
  • The Reader rewrite has also greatly sped up the reading and parsing of CHAT data.
  • The by_files argument, which many Reader methods has, now gives you a simpler list of results for each data file, no longer the previous output of a dict that mapped a file path to the file's result.
  • The participant argument, which many Reader methods has for specifying which participants' data to include in the output, has been renamed as participants to avoid confusion. There is no change to its behavior of handling either a single string (e.g., "CHI") or a collection of strings (e.g., {"CHI", "MOT"}) .
  • The following Reader methods have been renamed as indicated, some for stylistic or Pythonic reasons, others for reasons as given:
    • age -> ages
    • number_of_utterances -> n_utterances
    • number_of_files -> n_files
    • filenames -> file_paths
    • MLU -> mlu
    • MLUm -> mlum
    • MLUw -> mluw
    • TTR -> ttr
    • IPSyn -> ipsyn
    • word_frequency -> word_frequencies
    • from_chat_str -> from_strs
    • from_chat_files -> from_files
    • add -> append. Since the data files in a Reader have a natural ordering (by time of recording sessions, and therefore commonly by file paths as well), a reader is list-like rather than an unordered set of data files, which add would suggest.
    • participant_codes -> participants. Before this version, the methods participant_codes (for CHI, MOT, etc) and participants (for, say, Eve, Mother, Investigator, etc) co-existed, but in practice we mostly only care about CHI, MOT, etc. So the method participants for Eve etc has been removed, and participant_codes has been renamed as participants.
  • Each participant's info in a header dictionary has these keys renamed:
    • participant_name -> name
    • participant_role -> role
    • SES -> ses (socioeconomic status)
  • The class DependencyGraph has been made private (i.e., now _DependencyGraph with a leading underscore). Its functionality hasn't really changed (it's used in the computation of IPSyn). It may be made more visible again in the future if more functionality related to grammatical relations is developed in the package.
  • Switched to sphinx-rtd-theme as the documentation theme.
  • Switched to CircleCI orbs; update dev requirements' versions.

Deprecated

  • The following Reader methods have been deprecated:
    • tagged_sents (use tokens with by_utterances=True instead)
    • tagged_words (use tokens with by_utterances=False instead)
    • sents (use words with by_utterances=True instead)

Removed

  • The following methods of the Reader class have been removed:
    • abspath. Use file_paths instead.
    • index_to_tiers. All the unparsed tiers are now available from utterances.
    • participant_codes. It's been renamed as participants, another method now removed; see "Changed" above.
    • part_of_speech_tags
    • update and remove. A reader is a list-like collection of CHAT data files, not a set (which update and remove would suggest).
    • search and concordance. To search, use one of the words, tokens, and utterances methods to walk through a reader's CHAT data and keep track of elements of interest.
    • date_of_birth. The info is now available under headers, in each participant's "dob" key.

Fixed

  • Handled [/-] in cleaning utterances.
  • [x <number>] means a repetition of the previous word/item, not repetition of the entire utterance.

[0.12.0] - 2020-10-11

Added

  • Added support for Python 3.9.
  • Enabled black to enforce styling consistency.

[0.11.0] - 2020-07-02

Added

  • Started testing Python 3.7 and 3.8 on continuous integration. (#9)
  • Add time marker support (available at _SingleReader), originally contributed at #3 by @hellolzc. (#8)

Changed

  • Switched from Travis CI to CircleCI for autobuilds. (#9)
  • Switched README from reStructuredText to Markdown. (#9)
  • Removed conversational quotes in utterance processing; updated test CHAT file to match the latest CHILDES data. (#7)

Removed

  • Dropped support for Python 2.7, 3.4, and 3.5. All code related to Python 2+3 cross compatibility was removed. (#9)

[0.10.0] - 2017-11-02

  • Fixed unicode handling across Python 2 and 3
  • Renamed method find_filename of Reader as abspath.
  • Fixed bug in Reader method decorators
  • Handled multiple dates of recording in one CHAT file. The method dates_of_recording of a Reader instance now returns a list of dates.
  • Implemented the exclude parameter in various Reader methods for excluding specific participants.
  • Fixed bug in IPSyn.

[0.9.0] - 2017-10-25

  • Python 2 and 3 cross compatibility
  • Renamed the grammar.py module as dependency.py
    • Rewrite the class DependencyGraph; do not subclass from networkx's DiGraph anymore (and we remove networkx as a dependency of this library)
  • Removed multiprocessing in reading data files. Datasets are usually small enough that the performance gain, if any, wouldn't be worth it for the potential issues w.r.t. spawning multiple processes)
  • Developed capabilities to handle PhonBank data for handling %pho and %mod tiers
  • Improved clean_utterance()
  • Added parameter encoding in read_chat()
  • Added get_lemma_from_mor()
  • Added date_of_recording() and date_of_birth(); remove date()
  • Added clean_word()
  • Restricted get_IPSyn() to only the first 100 utterances
  • Added tests

[0.8] - 2016-01-30

  • Library now compatible only with Python 3.4 or above

  • For class Reader:

    • Defined read_chat() for initializing a Reader object
    • Added parameter by_files to various methods; remove the "all_" methods
    • Added reader manipulation methods: update(), add(), remove(), clear()
    • Added parameter sorted_by_age in filenames()
    • Added parameter month in age()
    • Added word_ngrams()
    • Added find_filename()
    • Added language development measures: MLUm(), MLUw(), TTR(), IPSyn()
    • Added search() and concordance()
    • Allowed regular expression matching for parameter participant
    • Added output formats for dependency graphs: to_tikz() and to_conll()
    • Distinguished participant_name and participant_role in metadata
    • The @Languages header contents are treated as a list but not a set now for ordering in bi/multilingualism
    • Undid collapses in transcriptions such as [x 4]
    • Various bug fixes

[0.7] - 2016-01-06

  • Added part_of_speech_tags() in SingleReader
  • Added "all X" methods in Reader
  • Bug fixes: clean_utterance(), DependencyGraph

[0.6] - 2015-12-27

  • cha_lines optimized
  • Methods added: tagged_words(), words(), tagged_sents(), sents()
  • Tier detection revamped. tier_sniffer() method removed, with self.tier_markers in SingleReader now being a set of %-tier markers.
  • len() for SingleReader added
  • word_frequency() for SingleReader added
  • Module grammar added, with class DependencyGraph being set up
  • Static methods in classes pulled out

[0.5] - 2015-12-16

  • New utterances() method for extracting utterances from transcripts
  • _clean_utterance method developed for filtering CHAT annotations away in utterances
  • Standardizing terminology: use "participant(s)" consistently instead of "speaker(s)"

[0.4] - 2015-12-13

  • New number_of_utterances() method for both Reader and SingleReader
  • To avoid confusion, metadata() method is removed.
  • Extraction of utterances and tiers with dict index_to_tiers

[0.3] - 2015-12-09

  • Class Reader can read multiple .cha files. The methods associated with Reader are mostly a dict mapping from a absolute-path filename to something. Reader depends on the class SingleReader for a single CHAT file.
  • Following the conventional CHILDES and CHAT terminology, the metadata() method in Reader is renamed headers() (though a "new" metadata() method is defined and points to headers() for convenience).

[0.2] - 2015-12-05

  • new methods for class Reader: languages(), date(), participants(), participant_codes()

[0.1] - 2015-12-04

  • first commit; set up the chat submodule
  • class Reader defined for reading CHAT files, with methods cha_lines(), metadata(), and age()