release: v0.2.1

v0.2.1

Author: severinsimmler

README.md

@@ -1,25 +1,38 @@
-# A Lightweight Conditional Random Field
+# Chaine
 
-This is a modern Python library without any third-party dependencies and a backend written in C implementing conditional random fields for natural language processing tasks like named entity recognition or part-of-speech tagging.
+A linear-chain conditional random field implementation.
+
+Chaine is a modern Python library without any third-party dependencies and a backend written in C implementing conditional random fields for natural language processing tasks like named entity recognition or part-of-speech tagging.
+
+- **Lightweight:** explain
+- **Fast:** explain
+- **Easy to use:** explain
 
 You can install the latest stable version from [PyPI](https://pypi.org/project/chaine):
 
 ```
 $ pip install chaine
 ```
 
-If you are interested in the theoretical concepts behind conditional random fields, I can recommend the introducing paper by [Lafferty et al](https://repository.upenn.edu/cgi/viewcontent.cgi?article=1162&context=cis_papers).
+If you are interested in the theoretical concepts behind conditional random fields, refer to the introducing paper by [Lafferty et al](https://repository.upenn.edu/cgi/viewcontent.cgi?article=1162&context=cis_papers).
 
 
-## Example
+## How it works
 
-```python
+```
 >>> import chaine
->>> sequences = [[["a", "a"], ["b", "b"]]]
->>> labels = [["0", "1"]]
->>> model = chaine.train(sequences, labels)
->>> model.predict(sequences)
-[['0', '1']]
+>>> tokens = [["John", "Lennon", "was", "rhythm", "guitarist" "of", "The", "Beatles"]]
+>>> labels = [["B-PER", "I-PER", "O", "O", "O", "O", "B-ORG", "I-ORG"]]
+>>> model = chaine.train(tokens, labels, max_iterations=5)
+Loading data
+Start training
+Iteration 1, train loss: 14.334076
+Iteration 2, train loss: 14.334064
+Iteration 3, train loss: 14.334053
+Iteration 4, train loss: 14.334041
+Iteration 5, train loss: 14.334029
+>>> model.predict(tokens)
+[['B-PER', 'I-PER', 'O', 'O', 'O', 'B-ORG', 'I-ORG']]
 ```
 
 Check out the introducing [Jupyter notebook](https://github.com/severinsimmler/chaine/blob/master/notebooks/tutorial.ipynb).

chaine/__init__.py

@@ -1,2 +1,2 @@
-from chaine.core import train
+from chaine.training import train
 from chaine.crf import Model, Trainer

chaine/crfsuite/include/crfsuite.hpp

@@ -364,13 +364,13 @@ namespace CRFSuite
 
         if (model == NULL)
         {
-            throw std::invalid_argument("The tagger is not opened");
+            throw std::invalid_argument("The tagger is not opened.");
         }
 
         // Obtain the dictionary interface representing the labels in the model.
         if ((ret = model->get_labels(model, &labels)))
         {
-            throw std::runtime_error("Failed to obtain the dictionary interface for labels");
+            throw std::runtime_error("Failed to obtain the dictionary interface for labels.");
         }
 
         // Collect all label strings to lseq.
@@ -405,13 +405,13 @@ namespace CRFSuite
 
         if (model == NULL || tagger == NULL)
         {
-            throw std::invalid_argument("The tagger is not opened");
+            throw std::invalid_argument("The tagger is not opened.");
         }
 
         // Obtain the dictionary interface representing the attributes in the model.
         if ((ret = model->get_attrs(model, &attrs)))
         {
-            throw std::runtime_error("Failed to obtain the dictionary interface for attributes");
+            throw std::runtime_error("Failed to obtain the dictionary interface for attributes.");
         }
 
         // Build an instance.
@@ -468,7 +468,7 @@ namespace CRFSuite
         // Obtain the dictionary interface representing the labels in the model.
         if ((ret = model->get_labels(model, &labels)))
         {
-            throw std::runtime_error("Failed to obtain the dictionary interface for labels");
+            throw std::runtime_error("Failed to obtain the dictionary interface for labels.");
         }
 
         // Run the Viterbi algorithm.

chaine/data.py

@@ -8,8 +8,6 @@
 import re
 from dataclasses import dataclass
 
-from chaine.typing import Iterable
-
 
 @dataclass
 class Token:

chaine/core.py renamed to chaine/training.py

@@ -68,6 +68,27 @@
    "source": [
     "crf.predict(tokens)"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Feature extraction\n",
+    "\n",
+    "```\n",
+    "identity of wi, identity of neighboring words\n",
+    "embeddings for wi, embeddings for neighboring words\n",
+    "part of speech of wi, part of speech of neighboring words\n",
+    "base-phrase syntactic chunk label of wi and neighboring words\n",
+    "presence of wi in a gazetteer\n",
+    "wi contains a particular prefix (from all prefixes of length ≤ 4)\n",
+    "wi contains a particular suffix (from all suffixes of length ≤ 4)\n",
+    "wi is all upper case\n",
+    "word shape of wi, word shape of neighboring words\n",
+    "short word shape of wi, short word shape of neighboring words\n",
+    "presence of hyphen\n",
+    "```"
+   ]
   }
  ],
  "metadata": {