Renard 0.7.0

Makefile

@@ -0,0 +1,7 @@
+.PHONY: test
+test:
+	uv run --group dev python -m pytest tests
+
+.PHONE: ui
+ui:
+	uv run --extra ui python -m renard.ui

README.md

@@ -9,11 +9,16 @@ Renard (Relationship Extraction from NARrative Documents) is a library for creat
 
 # Installation
 
-You can install the latest version using pip:
+Currently, Renard supports Python>=3.9,<=3.12. You can install the
+latest version using pip:
 
 > pip install renard-pipeline
 
-Currently, Renard supports Python>=3.9,<=3.12
+If you have a GPU, there are accelerated versions for Nvidia CUDA and
+AMD ROCm:
+
+> pip install renard-pipeline[cuda128]
+> pip install renard-pipeline[rocm63]
 
 
 # Documentation
@@ -59,7 +64,25 @@ For more information, see `renard_tutorial.py`, which is a tutorial in the `jupy
 
 > uv run python -m pytest tests
 
-Expensive tests are disabled by default. These can be run by setting the environment variable `RENARD_TEST_ALL` to `1`.
+Alternatively, the project Makefile has a test target:
+
+> make test
+
+Expensive tests are disabled by default. These can be run by setting the environment variable `RENARD_TEST_SLOW` to `1`.
+
+
+
+# Renard UI
+
+Since version 0.7, Renard has a web interface powered by gradio. First, install the additional dependencies:
+
+> uv sync --group ui
+
+Then, simply run:
+
+> make ui
+
+And open your browser at http://127.0.0.1:7860
 
 
 # Contributing

docs/contributing.rst

@@ -7,7 +7,7 @@ issue if you encounter a problem or want to discuss a specific
 feature. If you want to contribute a patch:
 
 1. Check that your code matches our code quality guidelines and that
-   all existing tests are passing with ``RENARD_TEST_ALL=1``.
+   all existing tests are passing with ``RENARD_TEST_SLOW=1``.
 2. Create a Github pull request with your patch, explaining the
    rationale behind it and giving a high level overview of your
    code. Mention the relevant issue if applicable.
@@ -36,9 +36,7 @@ the ``tests`` directory. We use ``pytest`` to test code, and also use
 ``hypothesis`` when applicable. If you open a patch, make sure that
 all tests are passing. In particular, do not rely on the CI, as it
 does not run time costly tests! Check for yourself locally, using
-``RENARD_TEST_ALL=1 python -m pytest tests``. Note that there are
+``RENARD_TEST_SLOW=1 python -m pytest tests``. Note that there are
 specific tests and environment variable for optional dependencies such
-as *stanza* (``RENARD_TEST_STANZA_OPTDEP``). These must be explicitely
-set to ``1`` if you want to test optional dependencies, as
-``RENARD_TEST_ALL=1`` does not enable test on these optional
-dependencies.
+as *stanza* (``RENARD_TEST_OPTDEP_STANZA``). These must be explicitely
+set to ``1`` if you want to test optional dependencies.

docs/installation.rst

@@ -5,7 +5,11 @@ Installation
 Using Pip
 =========
 
-Simply use ``pip install renard-pipeline``.
+For the simplest case, use ``pip install renard-pipeline``. By default, this installs the CPU version of PyTorch. If you want GPU support to accelerate inference:
+
+- CUDA 12.8: ``pip install renard-pipeline[cuda128]``
+- ROCm 6.3: ``pip install renard-pipeline[rocm63]``
+
 
 Note that for some modules, you might need to install additional
 libraries:

docs/pipeline.rst

@@ -75,17 +75,22 @@ For simplicity, one can use one of the preconfigured pipelines:
 
 .. code-block:: python
 
-   from renard.pipeline.preconfigured import bert_pipeline
+   from renard.pipeline.preconfigured import co_occurence_pipeline
 
    with open("./my_doc.txt") as f:
        text = f.read()
 
-   pipeline = bert_pipeline(
-       graph_extractor_kwargs={"co_occurrences_dist": (1, "sentences")}
-   )
+   pipeline = co_occurrence_pipeline()
    out = pipeline(text)
 
 
+The following preconfigured pipelines are available:
+
+- :func:`.co_occurrence_pipeline`
+- :func:`.conversational_pipeline`
+- :func:`.relational_pipeline`
+
+
 Pipeline Output: the Pipeline State
 ===================================
 
@@ -137,7 +142,7 @@ Tokenization
 Tokenization is the task of cutting text in *tokens*. It is usually
 the first task to apply to a text. 2 tokenizer are available:
 
-- :class:`.NLTKTokenizer`
+- :class:`.NLTKTokenizer` is the tokenizer from NLTK.
 - :class:`.StanfordCoreNLPPipeline` does contain a tokenizer as part
   of its full NLP pipeline.
 
@@ -148,16 +153,19 @@ Named Entity Recognition
 Named entity recognition (NER) detects entities occurences in the
 text. 3 modules are available:
 
-- :class:`.NLTKNamedEntityRecognizer`
-- :class:`.BertNamedEntityRecognizer`
+- :class:`.NLTKNamedEntityRecognizer` is a lightweight NER module from
+  NLTK, based on POS tagging and rules.
+- :class:`.BertNamedEntityRecognizer` is a NER module employing a
+  finetuned BERT model.
 - :class:`.StanfordCoreNLPPipeline` contains a NER model as part of
   its full NLP pipeline.
 
 
 Coreference Resolution
 ----------------------
 
-- :class:`.SpacyCorefereeCoreferenceResolver`
+- :class:`.SpacyCorefereeCoreferenceResolver` uses the spacy coreferee
+  module.
 - :class:`.BertCoreferenceResolver`, using the Tibert library.
 - :class:`.StanfordCoreNLPPipeline` can execute a coreference
   resolution model as part of its pipeline.
@@ -166,14 +174,14 @@ Coreference Resolution
 Quote Detection
 ---------------
 
-- :class:`.QuoteDetector`
+- :class:`.QuoteDetector` detect quotes using simple logic.
 
 
 Sentiment Analysis
 ------------------
 
 - :class:`.NLTKSentimentAnalyzer` leverages NLTK's Vader for sentiment
-  analysis
+  analysis.
 
 
 Characters Extraction
@@ -183,21 +191,36 @@ Characters extraction (or alias resolution) extract characters from
 occurences detected using NER. This is done by assigning each mention
 to a unique character.
 
-- :class:`.NaiveCharacterUnifier`
-- :class:`.GraphRulesCharacterUnifier`
+- :class:`.NaiveCharacterUnifier` assigns each mention with a unique
+  form to a character.
+- :class:`.GraphRulesCharacterUnifier` uses a set of rules to assign
+  each mention to a character.
+
+
+Relation Extraction
+-------------------
+
+- :class:`.GenerativeRelationExtractor` is currently in development
+  and should not be used.
 
 
 Speaker Attribution
 -------------------
 
-- :class:`.BertSpeakerDetector`
+- :class:`.BertSpeakerDetector` detects speaker using a finetuned BERT
+  model.
 
 
 Graph Extraction
 ----------------
 
-- :class:`.CoOccurrencesGraphExtractor`
-- :class:`.ConversationalGraphExtractor`
+- :class:`.CoOccurrencesGraphExtractor` extracts a graph of
+  co-occurrence between characters.
+- :class:`.ConversationalGraphExtractor` extracts a conversational
+  graph: either conversation between characters, or of character
+  mentions.
+- :class:`.RelationalGraphExtractor` extracts a relational graph,
+  where the relation between each character is typed.
 
 
 Dynamic Graphs
@@ -240,8 +263,9 @@ When executing the above block of code, the output attribute
 >>> out.character_network
 [<networkx.classes.graph.Graph object at 0x7fd9e9115900>]
 
-See :class:`.CoOccurrencesGraphExtractor` for more details on the
-usage of the ``dynamic`` and ``dynamic_window`` arguments.
+Both :class:`.CoOccurrencesGraphExtractor` and
+:class:`.ConversationalGraphExtractor` support dynamic networks. See
+their documentation for more details.
 
 Plot and export functions work as one would expect
 intuitively. :meth:`.PipelineState.plot_graph` allow to visualize the
@@ -255,10 +279,9 @@ dynamic graph to the Gephi format.
 Custom Segmentation
 -------------------
 
-The ``dynamic_window`` parameter of
-:class:`.CoOccurencesGraphExtractor` determines the segmentation of
-the dynamic networks, in number of interactions. In the example above,
-a new graph will be created for each 20 interactions.
+The ``dynamic_window`` parameter determines the segmentation of the
+dynamic networks, in number of interactions. In the example above, a
+new graph will be created for each 20 interactions.
 
 While one can rely on the arguments of the graph extractor of the
 pipeline to determine the dynamic window, Renard allows to specify a

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "renard-pipeline"
-version = "0.6.5"
+version = "0.7.0"
 description = "Relationships Extraction from NARrative Documents"
 authors = [
     {name = "Arthur Amalvy", email = "arthur.amalvy@univ-avignon.fr"},
@@ -9,20 +9,23 @@ license = { text = "GPL-3.0-only" }
 readme = "README.md"
 requires-python = ">=3.9,<3.13"
 dependencies = [
-    "torch>=2.0.0,!=2.0.1",
-    "transformers>=4.37",
-    "nltk>=3.9",
-    "tqdm>=4.62",
-    "networkx>=3.0",
-    "more-itertools>=10.5",
-    "nameparser>=1.1",
-    "matplotlib>=3.5",
-    "pandas>=2.0",
-    "pytest>=8.3.0",
-    "tibert>=0.5",
-    "grimbert>=0.1",
-    "datasets>=3.0",
+    "torch>=2.7.0",
+    "transformers>=4.57.1",
+    "nltk>=3.9.1",
+    "tqdm>=4.67.1",
+    "networkx>=3.2",
+    "more-itertools>=10.7",
+    "nameparser>=1.1.3",
+    "matplotlib>=3.9",
+    "pytest>=8.4.1",
+    "tibert>=0.5.2",
+    "grimbert>=0.1.5",
+    "datasets>=4.0.0",
     "rank-bm25>=0.2.2",
+    "accelerate>=1.10.1",
+    "scikit-learn>=1.6.1",
+    "tiktoken>=0.12.0",
+    "protobuf>=6.33.2",
 ]
 
 [build-system]
@@ -43,4 +46,66 @@ dev = [
     "Sphinx>=4.3",
     "sphinx-rtd-theme>=1.0.0",
     "sphinx-autodoc-typehints>=1.12.0",
-]
+]
+
+[project.optional-dependencies]
+ui = [
+    "gradio>=4.44.1",
+    "pyvis>=0.3.2",
+]
+cpu = [
+  "torch>=2.7.1",
+]
+cuda128 = [
+  "torch>=2.7.1",
+]
+rocm63 = [
+  "torch>=2.7.1",
+  "pytorch-triton-rocm>=3.1.0", 
+]
+rocm64 = [
+  "torch>=2.7.1",
+  "pytorch-triton-rocm>=3.1.0", 
+]
+
+[tool.uv]
+conflicts = [
+  [
+    { extra = "cpu" },
+    { extra = "cuda128" },
+    { extra = "rocm63" },
+    { extra = "rocm64" },
+  ],
+]
+
+[tool.uv.sources]
+torch = [
+  { index = "pytorch-cpu", extra = "cpu" },
+  { index = "pytorch-cuda128", extra = "cuda128" },
+  { index = "pytorch-rocm63", extra = "rocm63" },
+  { index = "pytorch-rocm64", extra = "rocm64" },
+]
+pytorch-triton-rocm = [
+  { index = "pytorch-rocm63", extra = "rocm63" },
+  { index = "pytorch-rocm64", extra = "rocm64" },
+]
+
+[[tool.uv.index]]
+name = "pytorch-cpu"
+url = "https://download.pytorch.org/whl/cpu"
+explicit = true
+
+[[tool.uv.index]]
+name = "pytorch-cuda128"
+url = "https://download.pytorch.org/whl/cu128"
+explicit = true
+
+[[tool.uv.index]]
+name = "pytorch-rocm63"
+url = "https://download.pytorch.org/whl/rocm6.3"
+explicit = true
+
+[[tool.uv.index]]
+name = "pytorch-rocm64"
+url = "https://download.pytorch.org/whl/rocm6.4"
+explicit = true

renard/pipeline/character_unification.py

@@ -21,10 +21,14 @@ class Character:
     mentions: List[Mention]
     gender: Gender = Gender.UNKNOWN
 
-    def longest_name(self) -> str:
+    def longest_name(self) -> Optional[str]:
+        if len(self.names) == 0:
+            return None
         return max(self.names, key=len)
 
-    def shortest_name(self) -> str:
+    def shortest_name(self) -> Optional[str]:
+        if len(self.names) == 0:
+            return None
         return min(self.names, key=len)
 
     def most_frequent_name(self) -> Optional[str]:
@@ -236,7 +240,6 @@ def __call__(
 
         # * link nodes based on several rules
         for name1, name2 in combinations(G.nodes(), 2):
-
             # preprocess name when needed
             pname1 = self._preprocess_name(name1)
             pname2 = self._preprocess_name(name2)
@@ -294,7 +297,6 @@ def try_remove_edges(edges):
                 pass
 
         for name1, name2 in combinations(G.nodes(), 2):
-
             # preprocess names when needed
             pname1 = self._preprocess_name(name1)
             pname2 = self._preprocess_name(name2)