tritopic 0.1.0__tar.gz → 1.0.0__tar.gz

{tritopic-0.1.0/tritopic.egg-info → tritopic-1.0.0}/PKG-INFO

@@ -1,14 +1,16 @@
 Metadata-Version: 2.4
 Name: tritopic
-Version: 0.1.0
-Summary: Tri-Modal Graph Topic Modeling with Iterative Refinement - A state-of-the-art topic modeling library
-Author-email: Roman Egger <roman@example.com>
+Version: 1.0.0
+Summary: Tri-Modal Graph Topic Modeling with Iterative Refinement
+Author-email: Roman Egger <roman.egger@example.com>
 License: MIT
 Project-URL: Homepage, https://github.com/roman-egger/tritopic
 Project-URL: Documentation, https://tritopic.readthedocs.io
 Project-URL: Repository, https://github.com/roman-egger/tritopic
-Keywords: topic-modeling,nlp,machine-learning,graph-clustering,leiden,embeddings,text-analysis,bertopic-alternative
+Project-URL: Issues, https://github.com/roman-egger/tritopic/issues
+Keywords: topic-modeling,nlp,machine-learning,bertopic,clustering,text-analysis,multilingual
 Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Science/Research
 Classifier: License :: OSI Approved :: MIT License
 Classifier: Programming Language :: Python :: 3
@@ -20,35 +22,42 @@ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Text Processing :: Linguistic
 Requires-Python: >=3.9
 Description-Content-Type: text/markdown
-License-File: LICENSE
 Requires-Dist: numpy>=1.21.0
-Requires-Dist: pandas>=1.3.0
 Requires-Dist: scipy>=1.7.0
 Requires-Dist: scikit-learn>=1.0.0
+Requires-Dist: pandas>=1.3.0
 Requires-Dist: sentence-transformers>=2.2.0
 Requires-Dist: leidenalg>=0.9.0
-Requires-Dist: igraph>=0.10.0
-Requires-Dist: umap-learn>=0.5.0
-Requires-Dist: hdbscan>=0.8.0
-Requires-Dist: plotly>=5.0.0
+Requires-Dist: python-igraph>=0.10.0
 Requires-Dist: tqdm>=4.60.0
-Requires-Dist: rank-bm25>=0.2.0
-Requires-Dist: keybert>=0.7.0
 Provides-Extra: llm
 Requires-Dist: anthropic>=0.18.0; extra == "llm"
 Requires-Dist: openai>=1.0.0; extra == "llm"
-Provides-Extra: full
-Requires-Dist: anthropic>=0.18.0; extra == "full"
-Requires-Dist: openai>=1.0.0; extra == "full"
-Requires-Dist: pacmap>=0.6.0; extra == "full"
-Requires-Dist: datamapplot>=0.1.0; extra == "full"
+Provides-Extra: multilingual
+Requires-Dist: langdetect>=1.0.9; extra == "multilingual"
+Requires-Dist: jieba>=0.42.1; extra == "multilingual"
+Provides-Extra: japanese
+Requires-Dist: fugashi>=1.2.0; extra == "japanese"
+Requires-Dist: unidic-lite>=1.0.8; extra == "japanese"
+Provides-Extra: korean
+Requires-Dist: konlpy>=0.6.0; extra == "korean"
+Provides-Extra: thai
+Requires-Dist: pythainlp>=4.0.0; extra == "thai"
+Provides-Extra: visualization
+Requires-Dist: plotly>=5.0.0; extra == "visualization"
+Requires-Dist: matplotlib>=3.4.0; extra == "visualization"
+Requires-Dist: umap-learn>=0.5.0; extra == "visualization"
+Provides-Extra: evaluation
+Requires-Dist: gensim>=4.0.0; extra == "evaluation"
+Provides-Extra: all
+Requires-Dist: tritopic[evaluation,llm,multilingual,visualization]; extra == "all"
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
 Requires-Dist: black>=23.0.0; extra == "dev"
 Requires-Dist: ruff>=0.1.0; extra == "dev"
 Requires-Dist: mypy>=1.0.0; extra == "dev"
-Dynamic: license-file
+Requires-Dist: sphinx>=6.0.0; extra == "dev"
 
 # 🔺 TriTopic
 
@@ -68,6 +77,8 @@ A state-of-the-art topic modeling library that consistently outperforms BERTopic
 | **Mutual kNN + SNN** | Eliminates noise bridges between unrelated documents |
 | **Leiden + Consensus** | Dramatically more stable than single-run clustering |
 | **Iterative Refinement** | Topics improve embeddings, embeddings improve topics |
+| **Multilingual Support** | 60+ languages with auto language detection |
+| **Archetype Representatives** | Rich document selection beyond simple centroids |
 | **LLM-Powered Labels** | Human-readable topic names via Claude or GPT-4 |
 
 ## 📦 Installation
@@ -79,8 +90,11 @@ pip install tritopic
 # With LLM labeling support
 pip install tritopic[llm]
 
+# With multilingual support
+pip install tritopic[multilingual]
+
 # Full installation (all features)
-pip install tritopic[full]
+pip install tritopic[all]
 ```
 
 ### From source (development)
@@ -171,6 +185,23 @@ model.generate_labels(labeler)
 print(model.get_topic_info())
 ```
 
+### Multilingual Support
+
+```python
+from tritopic import TriTopic
+
+# Auto-detect language and select appropriate model
+model = TriTopic(
+    language="auto",          # Auto-detect language
+    multilingual=False,       # Use language-specific model
+    verbose=True
+)
+
+# Works with Chinese, German, Japanese, etc.
+chinese_docs = ["机器学习正在改变医疗诊断", "深度神经网络取得超人类表现", ...]
+topics = model.fit_transform(chinese_docs)
+```
+
 ### With Metadata
 
 ```python
@@ -202,13 +233,18 @@ from tritopic import TriTopic, TriTopicConfig
 
 config = TriTopicConfig(
     # Embedding settings
-    embedding_model="all-MiniLM-L6-v2",  # or "BAAI/bge-base-en-v1.5"
+    embedding_model="all-MiniLM-L6-v2",  # or "auto", "BAAI/bge-base-en-v1.5"
     embedding_batch_size=32,
     
+    # Language settings
+    language="auto",           # or "en", "de", "zh", etc.
+    multilingual=False,        # Force multilingual model
+    language_detection_sample=100,
+    
     # Graph construction
     n_neighbors=15,
     metric="cosine",
-    graph_type="hybrid",  # "knn", "mutual_knn", "snn", "hybrid"
+    graph_type="hybrid",       # "knn", "mutual_knn", "snn", "hybrid"
     snn_weight=0.5,
     
     # Multi-view fusion weights
@@ -227,11 +263,17 @@ config = TriTopicConfig(
     use_iterative_refinement=True,
     max_iterations=5,
     convergence_threshold=0.95,
+    refinement_strength=0.15,
     
     # Keywords
     n_keywords=10,
+    keyword_method="ctfidf",   # "ctfidf", "bm25", "keybert"
+    
+    # Representatives (with archetype support)
     n_representative_docs=5,
-    keyword_method="ctfidf",  # "ctfidf", "bm25", "keybert"
+    representative_method="hybrid",  # "centroid", "medoid", "archetype", "diverse", "hybrid"
+    n_archetypes=4,
+    archetype_method="furthest_sum",
     
     # Misc
     outlier_threshold=0.1,
@@ -242,16 +284,25 @@ config = TriTopicConfig(
 model = TriTopic(config=config)
 ```
 
-### Quick Parameter Override
+### Pre-defined Configurations
 
 ```python
-# Override just what you need
-model = TriTopic(
-    embedding_model="BAAI/bge-base-en-v1.5",
-    n_neighbors=20,
-    use_iterative_refinement=True,
-    verbose=True,
-)
+from tritopic import TriTopic, get_config
+
+# Fast processing (less accurate)
+model = TriTopic(config=get_config("fast"))
+
+# High quality (slower)
+model = TriTopic(config=get_config("quality"))
+
+# Multilingual corpus
+model = TriTopic(config=get_config("multilingual"))
+
+# Chinese text
+model = TriTopic(config=get_config("chinese"))
+
+# German text
+model = TriTopic(config=get_config("german"))
 ```
 
 ## 📊 Evaluation
@@ -285,20 +336,6 @@ model = TriTopic()
 topics = model.fit_transform(documents, embeddings=embeddings)
 ```
 
-### Find Optimal Resolution
-
-```python
-from tritopic.core.clustering import ConsensusLeiden
-
-clusterer = ConsensusLeiden()
-optimal_res = clusterer.find_optimal_resolution(
-    graph=model.graph_,
-    resolution_range=(0.5, 2.0),
-    target_n_topics=15,  # Optional: target number
-)
-print(f"Optimal resolution: {optimal_res}")
-```
-
 ### Transform New Documents
 
 ```python
@@ -327,7 +364,8 @@ model = TriTopic.load("my_topic_model.pkl")
 | Views | Embeddings only | Semantic + Lexical + Metadata |
 | Refinement | None | Iterative embedding refinement |
 | Stability | Low (varies by run) | High (consensus clustering) |
-| Outlier Handling | HDBSCAN built-in | Configurable threshold |
+| Languages | Limited | 60+ with auto-detection |
+| Representatives | Centroid only | Archetypes, medoids, diverse |
 
 ### Benchmark Results
 
@@ -339,7 +377,7 @@ On 20 Newsgroups dataset (n=18,846):
 | Diversity | 0.834 | **0.891** | +7% |
 | Stability (ARI) | 0.721 | **0.934** | +30% |
 
-## 🏗️ Architecture
+## 🗂️ Architecture
 
 ```
 Documents
@@ -347,10 +385,10 @@ Documents
     ├─── Embedding Engine ──────────────┐
     │    (Sentence-BERT/BGE/Instructor) │
     │                                   │
-    ├─── Lexical Matrix ───────────────┼─── Multi-View
+    ├─── Lexical Matrix ────────────────┼─── Multi-View
     │    (TF-IDF/BM25)                  │    Graph Builder
     │                                   │         │
-    └─── Metadata Graph ───────────────┘         │
+    └─── Metadata Graph ────────────────┘         │
          (Optional)                              │
                                                  ▼
                                     ┌─────────────────────┐
@@ -369,6 +407,12 @@ Documents
                                     └──────────┬──────────┘
                                                │
                                     ┌──────────▼──────────┐
+                                    │  Representative      │
+                                    │  Selection           │
+                                    │  (Archetype/Hybrid)  │
+                                    └──────────┬──────────┘
+                                               │
+                                    ┌──────────▼──────────┐
                                     │   LLM Labeling       │
                                     │  (Claude/GPT-4)      │
                                     └─────────────────────┘

{tritopic-0.1.0 → tritopic-1.0.0}/README.md

@@ -16,6 +16,8 @@ A state-of-the-art topic modeling library that consistently outperforms BERTopic
 | **Mutual kNN + SNN** | Eliminates noise bridges between unrelated documents |
 | **Leiden + Consensus** | Dramatically more stable than single-run clustering |
 | **Iterative Refinement** | Topics improve embeddings, embeddings improve topics |
+| **Multilingual Support** | 60+ languages with auto language detection |
+| **Archetype Representatives** | Rich document selection beyond simple centroids |
 | **LLM-Powered Labels** | Human-readable topic names via Claude or GPT-4 |
 
 ## 📦 Installation
@@ -27,8 +29,11 @@ pip install tritopic
 # With LLM labeling support
 pip install tritopic[llm]
 
+# With multilingual support
+pip install tritopic[multilingual]
+
 # Full installation (all features)
-pip install tritopic[full]
+pip install tritopic[all]
 ```
 
 ### From source (development)
@@ -119,6 +124,23 @@ model.generate_labels(labeler)
 print(model.get_topic_info())
 ```
 
+### Multilingual Support
+
+```python
+from tritopic import TriTopic
+
+# Auto-detect language and select appropriate model
+model = TriTopic(
+    language="auto",          # Auto-detect language
+    multilingual=False,       # Use language-specific model
+    verbose=True
+)
+
+# Works with Chinese, German, Japanese, etc.
+chinese_docs = ["机器学习正在改变医疗诊断", "深度神经网络取得超人类表现", ...]
+topics = model.fit_transform(chinese_docs)
+```
+
 ### With Metadata
 
 ```python
@@ -150,13 +172,18 @@ from tritopic import TriTopic, TriTopicConfig
 
 config = TriTopicConfig(
     # Embedding settings
-    embedding_model="all-MiniLM-L6-v2",  # or "BAAI/bge-base-en-v1.5"
+    embedding_model="all-MiniLM-L6-v2",  # or "auto", "BAAI/bge-base-en-v1.5"
     embedding_batch_size=32,
     
+    # Language settings
+    language="auto",           # or "en", "de", "zh", etc.
+    multilingual=False,        # Force multilingual model
+    language_detection_sample=100,
+    
     # Graph construction
     n_neighbors=15,
     metric="cosine",
-    graph_type="hybrid",  # "knn", "mutual_knn", "snn", "hybrid"
+    graph_type="hybrid",       # "knn", "mutual_knn", "snn", "hybrid"
     snn_weight=0.5,
     
     # Multi-view fusion weights
@@ -175,11 +202,17 @@ config = TriTopicConfig(
     use_iterative_refinement=True,
     max_iterations=5,
     convergence_threshold=0.95,
+    refinement_strength=0.15,
     
     # Keywords
     n_keywords=10,
+    keyword_method="ctfidf",   # "ctfidf", "bm25", "keybert"
+    
+    # Representatives (with archetype support)
     n_representative_docs=5,
-    keyword_method="ctfidf",  # "ctfidf", "bm25", "keybert"
+    representative_method="hybrid",  # "centroid", "medoid", "archetype", "diverse", "hybrid"
+    n_archetypes=4,
+    archetype_method="furthest_sum",
     
     # Misc
     outlier_threshold=0.1,
@@ -190,16 +223,25 @@ config = TriTopicConfig(
 model = TriTopic(config=config)
 ```
 
-### Quick Parameter Override
+### Pre-defined Configurations
 
 ```python
-# Override just what you need
-model = TriTopic(
-    embedding_model="BAAI/bge-base-en-v1.5",
-    n_neighbors=20,
-    use_iterative_refinement=True,
-    verbose=True,
-)
+from tritopic import TriTopic, get_config
+
+# Fast processing (less accurate)
+model = TriTopic(config=get_config("fast"))
+
+# High quality (slower)
+model = TriTopic(config=get_config("quality"))
+
+# Multilingual corpus
+model = TriTopic(config=get_config("multilingual"))
+
+# Chinese text
+model = TriTopic(config=get_config("chinese"))
+
+# German text
+model = TriTopic(config=get_config("german"))
 ```
 
 ## 📊 Evaluation
@@ -233,20 +275,6 @@ model = TriTopic()
 topics = model.fit_transform(documents, embeddings=embeddings)
 ```
 
-### Find Optimal Resolution
-
-```python
-from tritopic.core.clustering import ConsensusLeiden
-
-clusterer = ConsensusLeiden()
-optimal_res = clusterer.find_optimal_resolution(
-    graph=model.graph_,
-    resolution_range=(0.5, 2.0),
-    target_n_topics=15,  # Optional: target number
-)
-print(f"Optimal resolution: {optimal_res}")
-```
-
 ### Transform New Documents
 
 ```python
@@ -275,7 +303,8 @@ model = TriTopic.load("my_topic_model.pkl")
 | Views | Embeddings only | Semantic + Lexical + Metadata |
 | Refinement | None | Iterative embedding refinement |
 | Stability | Low (varies by run) | High (consensus clustering) |
-| Outlier Handling | HDBSCAN built-in | Configurable threshold |
+| Languages | Limited | 60+ with auto-detection |
+| Representatives | Centroid only | Archetypes, medoids, diverse |
 
 ### Benchmark Results
 
@@ -287,7 +316,7 @@ On 20 Newsgroups dataset (n=18,846):
 | Diversity | 0.834 | **0.891** | +7% |
 | Stability (ARI) | 0.721 | **0.934** | +30% |
 
-## 🏗️ Architecture
+## 🗂️ Architecture
 
 ```
 Documents
@@ -295,10 +324,10 @@ Documents
     ├─── Embedding Engine ──────────────┐
     │    (Sentence-BERT/BGE/Instructor) │
     │                                   │
-    ├─── Lexical Matrix ───────────────┼─── Multi-View
+    ├─── Lexical Matrix ────────────────┼─── Multi-View
     │    (TF-IDF/BM25)                  │    Graph Builder
     │                                   │         │
-    └─── Metadata Graph ───────────────┘         │
+    └─── Metadata Graph ────────────────┘         │
          (Optional)                              │
                                                  ▼
                                     ┌─────────────────────┐
@@ -317,6 +346,12 @@ Documents
                                     └──────────┬──────────┘
                                                │
                                     ┌──────────▼──────────┐
+                                    │  Representative      │
+                                    │  Selection           │
+                                    │  (Archetype/Hybrid)  │
+                                    └──────────┬──────────┘
+                                               │
+                                    ┌──────────▼──────────┐
                                     │   LLM Labeling       │
                                     │  (Claude/GPT-4)      │
                                     └─────────────────────┘

{tritopic-0.1.0 → tritopic-1.0.0}/pyproject.toml

@@ -4,19 +4,20 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "tritopic"
-version = "0.1.0"
-description = "Tri-Modal Graph Topic Modeling with Iterative Refinement - A state-of-the-art topic modeling library"
+version = "1.0.0"
+description = "Tri-Modal Graph Topic Modeling with Iterative Refinement"
 readme = "README.md"
 license = {text = "MIT"}
 authors = [
-    {name = "Roman Egger", email = "roman@example.com"}
+    {name = "Roman Egger", email = "roman.egger@example.com"}
 ]
 keywords = [
-    "topic-modeling", "nlp", "machine-learning", "graph-clustering",
-    "leiden", "embeddings", "text-analysis", "bertopic-alternative"
+    "topic-modeling", "nlp", "machine-learning", "bertopic", 
+    "clustering", "text-analysis", "multilingual"
 ]
 classifiers = [
     "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
     "Intended Audience :: Science/Research",
     "License :: OSI Approved :: MIT License",
     "Programming Language :: Python :: 3",
@@ -30,18 +31,13 @@ classifiers = [
 requires-python = ">=3.9"
 dependencies = [
     "numpy>=1.21.0",
-    "pandas>=1.3.0",
     "scipy>=1.7.0",
     "scikit-learn>=1.0.0",
+    "pandas>=1.3.0",
     "sentence-transformers>=2.2.0",
     "leidenalg>=0.9.0",
-    "igraph>=0.10.0",
-    "umap-learn>=0.5.0",
-    "hdbscan>=0.8.0",
-    "plotly>=5.0.0",
+    "python-igraph>=0.10.0",
     "tqdm>=4.60.0",
-    "rank-bm25>=0.2.0",
-    "keybert>=0.7.0",
 ]
 
 [project.optional-dependencies]
@@ -49,11 +45,30 @@ llm = [
     "anthropic>=0.18.0",
     "openai>=1.0.0",
 ]
-full = [
-    "anthropic>=0.18.0",
-    "openai>=1.0.0",
-    "pacmap>=0.6.0",
-    "datamapplot>=0.1.0",
+multilingual = [
+    "langdetect>=1.0.9",
+    "jieba>=0.42.1",
+]
+japanese = [
+    "fugashi>=1.2.0",
+    "unidic-lite>=1.0.8",
+]
+korean = [
+    "konlpy>=0.6.0",
+]
+thai = [
+    "pythainlp>=4.0.0",
+]
+visualization = [
+    "plotly>=5.0.0",
+    "matplotlib>=3.4.0",
+    "umap-learn>=0.5.0",
+]
+evaluation = [
+    "gensim>=4.0.0",
+]
+all = [
+    "tritopic[llm,multilingual,visualization,evaluation]",
 ]
 dev = [
     "pytest>=7.0.0",
@@ -61,12 +76,14 @@ dev = [
     "black>=23.0.0",
     "ruff>=0.1.0",
     "mypy>=1.0.0",
+    "sphinx>=6.0.0",
 ]
 
 [project.urls]
 Homepage = "https://github.com/roman-egger/tritopic"
 Documentation = "https://tritopic.readthedocs.io"
 Repository = "https://github.com/roman-egger/tritopic"
+Issues = "https://github.com/roman-egger/tritopic/issues"
 
 [tool.setuptools.packages.find]
 where = ["."]
@@ -74,13 +91,15 @@ include = ["tritopic*"]
 
 [tool.black]
 line-length = 100
-target-version = ['py39', 'py310', 'py311']
+target-version = ['py39', 'py310', 'py311', 'py312']
 
 [tool.ruff]
 line-length = 100
-select = ["E", "F", "W", "I", "N"]
+select = ["E", "F", "W", "I", "N", "D", "UP"]
+ignore = ["D100", "D104"]
 
 [tool.mypy]
-python_version = "3.10"
+python_version = "3.9"
 warn_return_any = true
 warn_unused_configs = true
+ignore_missing_imports = true

tritopic-1.0.0/tritopic/__init__.py

@@ -0,0 +1,36 @@
+"""
+TriTopic: Tri-Modal Graph Topic Modeling with Iterative Refinement
+
+A state-of-the-art topic modeling library that consistently outperforms
+BERTopic and traditional approaches.
+
+Key Features:
+- Multi-view representation (semantic, lexical, metadata)
+- Hybrid graph construction (Mutual kNN + SNN)
+- Consensus Leiden clustering for stability
+- Iterative refinement for improved coherence
+- Multilingual support (60+ languages)
+- LLM-powered labeling
+
+Example:
+    >>> from tritopic import TriTopic
+    >>> model = TriTopic(verbose=True)
+    >>> topics = model.fit_transform(documents)
+    >>> print(model.get_topic_info())
+"""
+
+__version__ = "1.0.0"
+__author__ = "Roman Egger"
+
+from .model import TriTopic, Topic
+from .config import TriTopicConfig, get_config
+from .labeling import LLMLabeler, KeywordLabeler
+
+__all__ = [
+    "TriTopic",
+    "Topic",
+    "TriTopicConfig",
+    "get_config",
+    "LLMLabeler",
+    "KeywordLabeler",
+]