HISTORY.rst

History

0.9.5 (2024-02-20)

• Sets language_level compiler directive for Cython. (In response to #63).
• Updates to spaCy 3.7.4 and igraph 0.11.3.
• Removes references to conda-forge package as it has become difficult to keep updated.
• Adds dependency on pyarrow to satisfy future pandas releases.

0.9.4 (2024-01-16)

• Updates to spaCy 3.7.2 and to igraph 0.10.5.
• Adds Python 3.12 compatibility. (In response to #61).
• Fixes a bug in how tf-idf is calculated without sublinear scaling.
• If the corpus contains long documents, adjust spaCy's maximum document length parameter as needed.

0.9.3 (2023-07-11)

• Updates to igraph 0.10.5.
• Updates to spaCy 3.6.0, bringing support for Slovenian.
• Add dependency on spacy-lookups-data for better support of languages without pre-trained models (e.g., Turkish).

0.9.2 (2023-06-28)

• Improves documentation.
• Updates to scipy 1.10.
• Fixes how the compiled extension is built.

0.9.1 (2023-06-16)

• Only fixes deployment to PyPI, otherwise unchanged from previous version.

0.9.0 (2023-06-15)

• Improves documentation.
• Parallelizes execution of NLP and backbone extraction for large corpora and networks.
• Uses sparse matrices for some operations to avoid running out of memory.
• Makes other improvements to efficiency.
• Introduces a NodeType enum to differentiate document and term nodes.
• Updates to pandas 2.0 and spaCy 3.5.3.
• Now also tested for compatibility with PyPy 3.9!

0.8.8 (2023-03-21)

• Fixes bug in disparity filter.
• Updates to spaCy 3.5.1, along various other dependencies.
• Improves testing.

0.8.7 (2023-02-10)

• Updates to spaCy 3.5 and igraph 0.10.4.
• When initializing Corpus with data that includes duplicated document labels, issue a warning, and concatenate documents with shared labels. (Implemented in response to #54).
• Various code quality improvements.

0.8.6 (2022-11-30)

• Improves documentation.
• Progress bar for long-running operations (NLP and disparity filter). The progress bar is not shown during non-interactive use. To disable, set tn.params["progress_bar"] to False.
• Python 3.11 compatibility now confirmed on Windows, too.

0.8.5 (2022-10-28)

• Python 3.11 compatibility!
• Adds an optional parameter remove_weak_edges to Textnet to create a sparser graph.
• Adds ProjectedTextnet.m property to access one-mode graphs' adjacency matrices.
• Adds Stoltz and Taylor's textual spanning measure.
• Adds TextnetBase.cluster_strength and TextnetBase.cluster_local_cc to calculate the weighted degree and local clustering coefficient of nodes within the subgraph formed by the cluster they belong to.
• Improves display of top nodes per cluster (top_cluster_nodes).
• Adds Korean language code.
• Fixes several bugs that occurred when setting document attributes.

0.8.4 (2022-10-05)

• Updates to spaCy 3.4.1, igraph 0.10.1, and leidenalg 0.9.
• Adds Ukrainian language code.
• Improves type checking.

0.8.3 (2022-07-15)

• Updates to spaCy 3.4, bringing support for Croatian to textnets.
• Updates various other dependencies.
• Adds optional dependency for experimental Formal Concept Analysis features. To install, run pip install textnets[fca]. Graphviz must also be installed separately for lattice visualization purposes.

0.8.2 (2022-06-28)

• Makes TextnetBase an abstract base class, since it is not meant to be instantiated. Implements separate graph partition methods for the classes Textnet and ProjectedTextnet to work around an occasional bug.
• Adds Finnish and Swedish language codes.
• Improves type hints.
• Adds dark mode to docs!

0.8.1 (2022-06-27)

• Fixes bug #36.
• Updates dependencies, including igraph and spaCy.

0.8.0 (2022-05-06)

• Removes Python 3.7 compatibility.
• Fixes a bug in the HTML representation of the top-level module.
• Updates dependencies, including spaCy.

0.7.1 (2022-02-08)

• Fixes #35 (invisible edges when scaling by weight).
• Updates some dependencies.

0.7.0 (2021-11-12)

• Adds abilitiy to save and load an instance of Corpus, Textnet and params to and from file using Corpus.save, load_corpus, Textnet.save, load_textnet, params.save and params.load. The same file can be used for all three kinds of objects, so all relevant data for a project can be saved in one file.
• Some further optimization of backbone extraction.
• Adds bipartite centrality measures (HITS, CoHITS and BiRank) and a bipartite clustering coefficient.
• Improved testing and type hints.
• Expanded documentation with advanced topics, including the new save/load feature and interacting with other libraries for network analysis and machine learning. Docs now use the PyData theme.
• Improvements to visualization. When plotting, nodes and edges can now be scaled by any attribute.
• Breaking change: Term weighing now happens in the corpus submodule, so the sublinear argument has to be passed to the methods for term extraction (tokenized, noun_phrases and ngrams). This change will make it easier to add additional term extraction and weighing options.
• Adds tn.init_seed() utility to quickly initialize pseudorandom number generator.
• Adds Python 3.10 compatibility.
• Updates dependencies, including igraph with some relevant upstream changes contributed by yours truly, as well as spaCy.

0.6.0 (2021-10-14)

• Adds params as a container for global parameters. This makes it possible to fix the random seed and to change the resolution parameter for the community detection algorithm, among others. If the parameter autodownload is set to true, textnets will attempt to download all required spaCy language models automatically.
• Added HTML representation for the root module that displays versions of key dependencies.
• Added back string representations of Corpus and TextnetBase-derived classes.
• Adds a Corpus.from_dict method.
• Corpus now exposes the lang attribute, so the corpus language can be set after initialization of a class instance.
• The bipartite layout optionally used by Textnet.plot is now horizontal, so node types are arranged in columns rather than rows. That way node labels are less likely to overlap.
• Adds label_nodes argument to the Textnet.plot method to label both types of nodes. Defaults to False.
• Adds node_opacity and edge_opacity arguments for Textnet.plot.
• Makes polygons marking clusters more visually appealing by adding opacity.
• Probably fixes a bug that would occasionally result in an exception being raised during plotting (IndexError: color index too large).
• When initializing an instance of the Textnet class, you can now optionally pass the argument connected=True, in which case only the largest component of the underlying network is kept. When creating a one-mode projection using Textnet.project, a connected argument can also be passed.
• Adds TextnetBase.save_graph to save the underlying graph (for instance, for further processing in Gephi).
• Improved and extended documentation and docstrings.
• Update dependencies.

0.5.4 (2021-09-24)

• Fix the cross-platform build and deploy pipeline.
• Create binary packages for conda-forge.
• Otherwise, no substantive change from previous release.

0.5.3 (2021-09-24)

• Adds Catalan, Macedonian and Russian language models.
• Significantly speeds up backbone extraction by implementing the disparity filter integrand in Cython. (If the compiled extension cannot be loaded for some reason, it falls back on an interpreted function.)
• PyPI should now receive binary wheels for Mac, Windows and Linux (via GitHub Actions) to ease installation on each platform.
• Improved type annotations.
• Update several dependencies.

0.5.2 (2021-08-24)

• Improve the handling of edge cases when initializing the Corpus and Textnet classes, such as empty data being provided.
• Added ability to run the tutorial in the documentation interactively using thebe.
• Update to spacy 3.1 and bump other dependencies.

0.5.1 (2021-07-06)

• Adds Corpus.ngrams method as alternative to Corpus.noun_phrases. This is useful when working in languages that do not have noun chunks, such as Chinese.
• Fixes a bug in Corpus.from_files.
• Introduces HTML representations of core classes for nicer integration in Jupyter notebooks.
• Updates several dependencies.

0.5.0 (2021-06-28)

• Migrate continuous integration testing from Travis to GitHub Actions.
• Continuous integration tests now run for MacOS and Windows too.
• Update to Spacy 3 and bump other dependency versions.
• Improvements to documentation.
• Handle dependencies and build project using Poetry (PEP 517 and 518).
• Remove deprecated command-line interface.

0.4.11 (2020-11-09)

• Python 3.9 compatibility!
• Updated documentation with conda-forge installation option.
• Bump versions for numerous dependencies.

0.4.10 (2020-09-14)

• Add cairocffi dependency and update installation docs.
• Bump leidenalg dependency to version 0.8.1.

0.4.9 (2020-07-15)

• Add color_clusters option to Textnet plotting methods. This colors nodes according to their partition using a bespoke color palette.

0.4.8 (2020-07-10)

• The Corpus class now handles missing data (#13).
• Support for more corpus languages. If no statistical language model is available, Corpus tries to use a basic ("blank") model.
• Improved documentation around dependencies and language support.
• Added tests.

0.4.7 (2020-07-01)

• No substantive change from previous release.

0.4.6 (2020-07-01)

• Bump spacy dependency to version 2.3 because it includes several new language models.

0.4.5 (2020-06-29)

• Textnet.plot and ProjectedTextnet.plot now accept arguments to selectively suppress node or edge labels. node_label_filter and edge_label_filter take a function that is mapped to the iterator of nodes and edges. Only nodes or edges for which the function returns True are displayed in the plot.
• Corpus now has a useful string representation.
• Documentation updates, particularly to show the label filter functionality.

0.4.4 (2020-06-19)

• Methods to report centrality measures in TextnetBase now return pandas.Series objects. This has some nice benefits, like seeing node labels alongside centrality measures and being able to call .hist() on them to visualize the distribution.
• Scaling of nodes by centrality in plots should bring out differences more clearly now.
• Improved and expanded tutorial. Among other things, it now uses short codes to specify language models.

0.4.3 (2020-06-17)

• Python 3.7 compatibility is here.
• New circular_layout option for Textnet.plot. This is based on "Tidier Drawings" and looks very nice for some bipartite graphs.
• String representation of Textnet instances now gives helpful information.
• Updated documentation to note changed Python version requirement.

0.4.2 (2020-06-16)

• ProjectedTextnet.plot now takes an argument, alpha, that allows for pruning the graph in order to visualize its "backbone." This is useful when working with hairball graphs, which is common when creating textnets. Right now, it uses Serrano et al.'s disparity filter. That means that edges with an alpha value greater than the one specified are discarded, so lower values mean more extreme pruning.
• Language models can now be specified using a short ISO language code.
• Bipartite networks can now be plotted using a layered layout (by Kozo Sugiyama). Simply pass sugiyama_layout=True to Textnet.plot.
• Incremental improvements to documentation.

0.4.1 (2020-06-12)

• Documented TextnetBase methods to output lists of nodes ranked by various centrality measures: top_betweenness and several more.
• Added top_cluster_nodes to output list of top nodes per cluster found via community detection. This is useful when trying to interpret such clusters as themes/topics (in the projected word-to-word graph) or as groupings (in the document-to-document graph).
• Small additions to documentation.

0.4.0 (2020-06-11)

Lots of changes, some of them breaking, but overall just providing nicer abstractions over the underlying pandas and igraph stuff.

• Introduced TextnetBase and ProjectedTextnet classes, and made Textnet a descendant of the former.
• Improved code modularity to make it easier to add features.
• Corpus is now based on a Series rather than a DataFrame.
• Added methods for creating an instance of Corpus: from_df, from_csv, from_sql.
• Expanded and improved documentation.
• Added bibliography to documentation using a Sphinx bibtex plugin.
• A first contributor!

0.3.6 (2020-06-03)

• Small change to finally get automatic deployments to PyPI to work.

0.3.5 (2020-06-03)

• Overall improvements to documentation.
• Added label_edges argument to Textnet.plot.

0.3.4 (2020-06-02)

• Integrated self-contained example that can be downloaded as Jupyter notebook into tutorial.
• Still trying to get automatic deployments to PyPI working.

0.3.3 (2020-06-02)

• More documentation.
• Attempt to get automatic deployments to PyPI working.

0.3.2 (2020-06-02)

• Set up continuous integration with Travis CI.
• Set up pyup.io dependency safety checks.
• Expanded documentation.
• A logo!

0.3.2 (2020-05-31)

• Further improvements to documentation.

0.3.1 (2020-05-31)

• Improvements to documentation.

0.3.0 (2020-05-31)

• First release on PyPI.