refactor: don't use suggestion

README.md

@@ -3,7 +3,7 @@
 [![CI](https://github.com/spraakbanken/sparv-ocr-correction-plugin/actions/workflows/ci.yml/badge.svg)](https://github.com/spraakbanken/sparv-ocr-correction-plugin/actions/workflows/ci.yml)
 [![PyPI version](https://badge.fury.io/py/sparv-ocr-correction-plugin.svg)](https://pypi.org/project/sparv-ocr-correction-plugin)
 
-Sparv plugin to annotate suggestions to OCR:ed documents.
+Sparv plugin to annotate corrections to OCR:ed documents.
 
 ## Install
 

examples/hello-ocr/config.yaml

@@ -10,7 +10,7 @@ export:
         - <sentence>
         # - <token:word>
         - <token>:stanza.pos
-        - <token>:ocr_correction.ocr-correction
+        - <token>:ocr_correction.ocr-correction--viklofg-swedish-ocr
 
 sparv:
     compression: none

src/ocr_correction/__init__.py

@@ -2,7 +2,6 @@
 
 from sparv.api import (  # type: ignore [import-untyped]
     Annotation,
-    Config,
     Output,
     annotator,
     get_logger,
@@ -17,18 +16,7 @@
 
 DEFAULT_MODEL_NAME = "viklofg/swedish-ocr-correction"
 DEFAULT_TOKENIZER_NAME = "google/byt5-small"
-__config__ = [
-    Config(
-        "ocr_correction.model",
-        description="Huggingface pretrained model name",
-        default=DEFAULT_MODEL_NAME,
-    ),
-    Config(
-        "ocr_correction.tokenizer",
-        description="HuggingFace pretrained tokenizer name",
-        default=DEFAULT_TOKENIZER_NAME,
-    ),
-]
+
 
 __version__ = "0.1.0"
 
@@ -42,18 +30,18 @@
 )
 def annotate_ocr_correction(
     out_ocr_correction: Output = Output(
-        "<token>:ocr_correction.ocr-correction",
+        "<token>:ocr_correction.ocr-correction--viklofg-swedish-ocr",
         cls="ocr_correction",
         description="Neighbours from masked BERT (format: '|<word>:<score>|...|)",
     ),
     word: Annotation = Annotation("<token:word>"),
     sentence: Annotation = Annotation("<sentence>"),
-    model_name: str = Config("ocr_correction.model"),
-    tokenizer_name: str = Config("ocr_correction.tokenizer"),
 ) -> None:
+    tokenizer_name = DEFAULT_TOKENIZER_NAME
+    model_name = DEFAULT_MODEL_NAME
     tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
     model = T5ForConditionalGeneration.from_pretrained(model_name)
-    ocr_suggestor = OcrSuggestor(model=model, tokenizer=tokenizer)
+    ocr_corrector = OcrCorrector(model=model, tokenizer=tokenizer)
 
     sentences, _orphans = sentence.get_children(word)
     token_word = list(word.read())
@@ -64,14 +52,14 @@ def annotate_ocr_correction(
         logger.progress()  # type: ignore
         sent_to_tag = [token_word[token_index] for token_index in sent]
 
-        ocr_corrections = ocr_suggestor.calculate_suggestions(sent_to_tag)
+        ocr_corrections = ocr_corrector.calculate_corrections(sent_to_tag)
         out_ocr_correction_annotation[:] = ocr_corrections
 
     logger.info("writing annotations")
     out_ocr_correction.write(out_ocr_correction_annotation)
 
 
-class OcrSuggestor:
+class OcrCorrector:
     TEXT_LIMIT: int = 127
 
     def __init__(self, *, tokenizer, model) -> None:
@@ -81,7 +69,7 @@ def __init__(self, *, tokenizer, model) -> None:
             "text2text-generation", model=model, tokenizer=tokenizer
         )
 
-    def calculate_suggestions(self, text: list[str]) -> list[Optional[str]]:
+    def calculate_corrections(self, text: list[str]) -> list[Optional[str]]:
         logger.debug("Analyzing '%s'", text)
         parts = []
         curr_part: list[str] = []

tests/conftest.py

@@ -2,7 +2,7 @@
 from ocr_correction import (
     DEFAULT_MODEL_NAME,
     DEFAULT_TOKENIZER_NAME,
-    OcrSuggestor,
+    OcrCorrector,
 )
 from transformers import (  # type: ignore [import-untyped]
     AutoTokenizer,
@@ -11,7 +11,7 @@
 
 
 @pytest.fixture(scope="session")
-def ocr_suggestor() -> OcrSuggestor:
+def ocr_corrector() -> OcrCorrector:
     tokenizer = AutoTokenizer.from_pretrained(DEFAULT_TOKENIZER_NAME)
     model = T5ForConditionalGeneration.from_pretrained(DEFAULT_MODEL_NAME)
-    return OcrSuggestor(model=model, tokenizer=tokenizer)
+    return OcrCorrector(model=model, tokenizer=tokenizer)

tests/test_ocr_suggestor.py

@@ -1,7 +1,7 @@
-from ocr_correction import OcrSuggestor
+from ocr_correction import OcrCorrector
 
 
-def test_short_text(ocr_suggestor: OcrSuggestor):
+def test_short_text(ocr_corrector: OcrCorrector):
     text = [
         "Den",
         "i",
@@ -16,7 +16,7 @@ def test_short_text(ocr_suggestor: OcrSuggestor):
         "Frölnndaviken",
         ".",
     ]
-    actual = ocr_suggestor.calculate_suggestions(text)
+    actual = ocr_corrector.calculate_corrections(text)
 
     expected = [
         None,
@@ -35,7 +35,7 @@ def test_short_text(ocr_suggestor: OcrSuggestor):
     assert actual == expected
 
 
-def test_long_text(ocr_suggestor: OcrSuggestor):
+def test_long_text(ocr_corrector: OcrCorrector):
     text1 = [
         "Förvaltningen",
         "af",
@@ -71,8 +71,8 @@ def test_long_text(ocr_suggestor: OcrSuggestor):
     # blifvit dertill utsedd; tillkommande Fullmäktige att sjelfva bland sig välja en
     # vice Ordförande att föra ordet, när hinder för Ordföranden inträffar."""
     # print(f"{len(text2)=}, {len(text2.encode())=}")
-    actual = ocr_suggestor.calculate_suggestions(text1)
-    # actual = ocr_suggestor.calculate_suggestions(text2)
+    actual = ocr_corrector.calculate_corrections(text1)
+    # actual = ocr_corrector.calculate_corrections(text2)
 
     expected = [
         None,