From c1b55192e3ad829d63ef64beaaa4329fa6cd9994 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 10:41:49 -0400
Subject: [PATCH 1/9] Added basic recipe logging
---
src/corppa/poetry_detection/annotation/recipe.py | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index abd35a9..e6f2ddd 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -18,6 +18,7 @@
import spacy
from prodigy.components.loaders import JSONL
from prodigy.core import Arg, recipe
+from prodigy.util import log
#: reference to current directory, for use as Prodigy CSS directory
CURRENT_DIR = Path(__file__).parent.absolute()
@@ -86,7 +87,7 @@ def annotate_text_and_image(
"""Annotate text and image side by side: allows adding manual spans
to both image and text. Intended for page-level annotation.
"""
-
+ log("RECIPE: Starting recipe annotate_text_and_image", locals())
stream = JSONL(source) # load jsonlines into stream
# tokenize for span annotation and add image prefix
tokenized_stream = tokenize_stream(stream, image_prefix)
@@ -136,7 +137,7 @@ def annotate_page_text(
with text for reference only (image cannot be annotated).
Intended for page-level annotation.
"""
-
+ log("RECIPE: Starting recipe annotate_page_text", locals())
stream = JSONL(source) # load jsonlines into stream
# tokenize for span annotation and add image prefix
tokenized_stream = tokenize_stream(stream, image_prefix)
From 1c12cd93d9051d62f44d26ee0da8c2684ae62bdb Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 12:25:29 -0400
Subject: [PATCH 2/9] Modify recipes to use Prodigy's get_label utility
---
.../poetry_detection/annotation/recipe.py | 18 +++++++++++++-----
1 file changed, 13 insertions(+), 5 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index e6f2ddd..635eacd 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -18,7 +18,7 @@
import spacy
from prodigy.components.loaders import JSONL
from prodigy.core import Arg, recipe
-from prodigy.util import log
+from prodigy.util import get_labels, log
#: reference to current directory, for use as Prodigy CSS directory
CURRENT_DIR = Path(__file__).parent.absolute()
@@ -78,7 +78,11 @@ def tokenize_stream(stream, image_prefix=None):
@recipe(
"annotate_text_and_image",
dataset=Arg(help="path to input dataset"),
- labels=Arg("--label", "-l", help="Comma-separated label(s)"),
+ label=Arg(
+ "--label",
+ "-l",
+ help="Comma-separated label(s) to annotate or text file with one label per line",
+ ),
image_prefix=Arg("--image-prefix", "-i", help="Base URL for images"),
)
def annotate_text_and_image(
@@ -93,7 +97,7 @@ def annotate_text_and_image(
tokenized_stream = tokenize_stream(stream, image_prefix)
# split labels by commas and strip any whitespace
- label_list = [label.strip() for label in labels.split(",")]
+ label_list = get_labels(label)
blocks = [
{
@@ -127,7 +131,11 @@ def annotate_text_and_image(
@recipe(
"annotate_page_text",
dataset=Arg(help="path to input dataset"),
- labels=Arg("--label", "-l", help="Comma-separated label(s)"),
+ label=Arg(
+ "--label",
+ "-l",
+ help="Comma-separated label(s) to annotate or text file with one label per line",
+ ),
image_prefix=Arg("--image-prefix", "-i", help="Base URL for images"),
)
def annotate_page_text(
@@ -143,7 +151,7 @@ def annotate_page_text(
tokenized_stream = tokenize_stream(stream, image_prefix)
# split labels by commas and strip any whitespace
- label_list = [label.strip() for label in labels.split(",")]
+ label_list = get_label(label)
blocks = [
{
From 6d19b2fb01b4cd064343cdb593b1f304a5d84dcb Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 12:28:13 -0400
Subject: [PATCH 3/9] Updated top-level documentation
---
src/corppa/poetry_detection/annotation/recipe.py | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index 635eacd..b1f1c71 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -1,8 +1,12 @@
"""
-This module provides custom recipes for Prodigy annotation. They were
-created with page-level text annotation in mind, and support annotating
-text with a reference image displayed beside the text (`annotate_page_text`),
-or annotating both text and image side by side (`annotate_text_and_image`).
+This module provides custom recipes for Prodigy annotation. These were
+created with page-level annotation in mind, and assume a page is associated
+with both an image and text.
+
+Recipes:
+ * `annotate_page_text`: Annotate a page's text with the page's image
+ displayed side-by-side for reference.
+ * `annotate_text_and_image`: Annotate both a page's text and image side-by-side.
Referenced images must be served out independently for display; the image url
prefix for images should be specified when initializing the recipe.
@@ -10,7 +14,7 @@
Example use:
```
prodigy annotate_page_text poetry_spans poetry_pages.jsonl --label POETRY,PROSODY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix http://localhost:8000/
-prodigy annotate_text_and_image poetry_text_image poetry_pages.jsonl --label POETRY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix http://localhost:8000/
+prodigy annotate_text_and_image poetry_text_image poetry_pages.jsonl -l POETRY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix http://localhost:8000/
"""
from pathlib import Path
From a961c1e31949a7b4e8837ff548c240db058374f5 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 12:33:03 -0400
Subject: [PATCH 4/9] Fixed label(s) typos
---
src/corppa/poetry_detection/annotation/recipe.py | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index b1f1c71..da2c76a 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -101,7 +101,7 @@ def annotate_text_and_image(
tokenized_stream = tokenize_stream(stream, image_prefix)
# split labels by commas and strip any whitespace
- label_list = get_labels(label)
+ label_list = get_labels(labels)
blocks = [
{
@@ -155,7 +155,7 @@ def annotate_page_text(
tokenized_stream = tokenize_stream(stream, image_prefix)
# split labels by commas and strip any whitespace
- label_list = get_label(label)
+ label_list = get_labels(labels)
blocks = [
{
From 6c5d9388663a7f0e27f0af90ebf57132af27ce48 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 21:32:35 -0400
Subject: [PATCH 5/9] Update to use Stream component's get_stream
---
src/corppa/poetry_detection/annotation/recipe.py | 13 +++++++------
1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index da2c76a..7766422 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -20,9 +20,10 @@
from pathlib import Path
import spacy
-from prodigy.components.loaders import JSONL
+from prodigy import log
+from prodigy.components.stream import get_stream
from prodigy.core import Arg, recipe
-from prodigy.util import get_labels, log
+from prodigy.util import get_labels
#: reference to current directory, for use as Prodigy CSS directory
CURRENT_DIR = Path(__file__).parent.absolute()
@@ -82,7 +83,7 @@ def tokenize_stream(stream, image_prefix=None):
@recipe(
"annotate_text_and_image",
dataset=Arg(help="path to input dataset"),
- label=Arg(
+ labels=Arg(
"--label",
"-l",
help="Comma-separated label(s) to annotate or text file with one label per line",
@@ -96,7 +97,7 @@ def annotate_text_and_image(
to both image and text. Intended for page-level annotation.
"""
log("RECIPE: Starting recipe annotate_text_and_image", locals())
- stream = JSONL(source) # load jsonlines into stream
+ stream = get_stream(source)
# tokenize for span annotation and add image prefix
tokenized_stream = tokenize_stream(stream, image_prefix)
@@ -135,7 +136,7 @@ def annotate_text_and_image(
@recipe(
"annotate_page_text",
dataset=Arg(help="path to input dataset"),
- label=Arg(
+ labels=Arg(
"--label",
"-l",
help="Comma-separated label(s) to annotate or text file with one label per line",
@@ -150,7 +151,7 @@ def annotate_page_text(
Intended for page-level annotation.
"""
log("RECIPE: Starting recipe annotate_page_text", locals())
- stream = JSONL(source) # load jsonlines into stream
+ stream = get_stream(source)
# tokenize for span annotation and add image prefix
tokenized_stream = tokenize_stream(stream, image_prefix)
From 27a569db27c3c8156fca98be852d9c05e9d05681 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Tue, 22 Oct 2024 22:39:07 -0400
Subject: [PATCH 6/9] Update recipes to use stream.apply & fetch_media
---
.../poetry_detection/annotation/recipe.py | 104 +++++++++++-------
1 file changed, 66 insertions(+), 38 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index 7766422..4b61209 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -21,6 +21,8 @@
import spacy
from prodigy import log
+from prodigy.components.preprocess import add_tokens
+from prodigy.components.preprocess import fetch_media as fetch_media_preprocessor
from prodigy.components.stream import get_stream
from prodigy.core import Arg, recipe
from prodigy.util import get_labels
@@ -48,36 +50,28 @@
}
-def tokenize_stream(stream, image_prefix=None):
- """Takes a stream of Prodigy tasks and tokenizes text for span annotation,
- and optionally adds an image prefix URL to any image paths present.
- Stream is expected to contain `text` and may contain image_path` and a `meta`
- dictionary. Returns a generator of the stream.
- """
+def add_image(task, image_prefix=None):
+ if image_prefix is None:
+ task["image"] = task["image_path"]
+ else:
+ path_pfx = image_prefix.rstrip("/")
+ task["image"] = f"{path_pfx}/{task['image_path']}"
+ return task
- nlp = spacy.blank("en") # use blank spaCy model for tokenization
- # ensure image prefix URL does not have a trailing slash
- if image_prefix is None:
- image_prefix = ""
- image_prefix = image_prefix.rstrip("/")
-
- for task in stream:
- if task.get("text"):
- doc = nlp(task["text"])
- task["tokens"] = [
- {
- "text": token.text,
- "start": token.idx,
- "end": token.idx + len(token.text),
- "id": i,
- }
- for i, token in enumerate(doc)
- ]
- # add image prefix URL for serving out images
- if "image_path" in task:
- task["image"] = f"{image_prefix}/{task['image_path']}"
- yield task
+def add_images(examples, image_prefix=None):
+ for task in examples:
+ yield add_image(task, image_prefix=image_prefix)
+
+
+def remove_images(examples, image_prefix=None):
+ for task in examples:
+ # If "image" is a base64 string and "image_path" is present in the task,
+ # remove the image data
+ if task["image"].startswith("data:") and "image_path" in task:
+ # Replace image with full image path
+ add_image(task, image_prefix=image_prefix)
+ return examples
@recipe(
@@ -89,17 +83,29 @@ def tokenize_stream(stream, image_prefix=None):
help="Comma-separated label(s) to annotate or text file with one label per line",
),
image_prefix=Arg("--image-prefix", "-i", help="Base URL for images"),
+ fetch_media=Arg(
+ "--fetch-media", "-FM", help="Load images from local paths or URLs"
+ ),
)
def annotate_text_and_image(
- dataset: str, source: str, labels: str, image_prefix: str = None
+ dataset: str,
+ source: str,
+ labels: str,
+ image_prefix: str = None,
+ fetch_media: bool = False,
):
"""Annotate text and image side by side: allows adding manual spans
to both image and text. Intended for page-level annotation.
"""
log("RECIPE: Starting recipe annotate_text_and_image", locals())
stream = get_stream(source)
- # tokenize for span annotation and add image prefix
- tokenized_stream = tokenize_stream(stream, image_prefix)
+ # add tokens tokenize
+ stream.apply(add_tokens, nlp=spacy.blank("en"), stream=stream)
+ # add image prefix
+ stream.apply(add_images, image_prefix=image_prefix)
+ # optionally fetch media
+ if fetch_media:
+ stream.apply(fetch_media_preprocessor, ["image"], skip=True)
# split labels by commas and strip any whitespace
label_list = get_labels(labels)
@@ -125,13 +131,18 @@ def annotate_text_and_image(
}
)
- return {
+ components = {
"dataset": dataset,
- "stream": tokenized_stream,
+ "stream": stream,
"view_id": "blocks",
"config": config,
}
+ if fetch_media:
+ components["before_db"] = lambda x: remove_images(x, image_prefix=image_prefix)
+
+ return components
+
@recipe(
"annotate_page_text",
@@ -142,9 +153,16 @@ def annotate_text_and_image(
help="Comma-separated label(s) to annotate or text file with one label per line",
),
image_prefix=Arg("--image-prefix", "-i", help="Base URL for images"),
+ fetch_media=Arg(
+ "--fetch-media", "-FM", help="Load images from local paths or URLs"
+ ),
)
def annotate_page_text(
- dataset: str, source: str, labels: str, image_prefix: str = None
+ dataset: str,
+ source: str,
+ labels: str,
+ image_prefix: str = None,
+ fetch_media: bool = False,
):
"""Annotate text with manual spans; displays an image side by side
with text for reference only (image cannot be annotated).
@@ -152,8 +170,13 @@ def annotate_page_text(
"""
log("RECIPE: Starting recipe annotate_page_text", locals())
stream = get_stream(source)
- # tokenize for span annotation and add image prefix
- tokenized_stream = tokenize_stream(stream, image_prefix)
+ # add tokens tokenize
+ stream.apply(add_tokens, nlp=spacy.blank("en"), stream=stream)
+ # add image prefix
+ stream.apply(add_images, stream, image_prefix=image_prefix)
+ # optionally fetch media
+ if fetch_media:
+ stream.apply(fetch_media_preprocessor, ["image"], skip=True)
# split labels by commas and strip any whitespace
label_list = get_labels(labels)
@@ -175,9 +198,14 @@ def annotate_page_text(
}
)
- return {
+ components = {
"dataset": dataset,
- "stream": tokenized_stream,
+ "stream": stream,
"view_id": "blocks",
"config": config,
}
+
+ if fetch_media:
+ components["before_db"] = lambda x: remove_images(x, image_prefix=image_prefix)
+
+ return components
From eafd313fe16b9aa262e7de0fc45a1d9b8a9e0c78 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Wed, 23 Oct 2024 11:44:06 -0400
Subject: [PATCH 7/9] Added review recipe
---
.../poetry_detection/annotation/recipe.py | 286 +++++++++++++++---
1 file changed, 252 insertions(+), 34 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index 4b61209..cc6db10 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -1,12 +1,14 @@
"""
This module provides custom recipes for Prodigy annotation. These were
-created with page-level annotation in mind, and assume a page is associated
-with both an image and text.
+created with page-level annotation in mind and assume a page is associated
+with both text and an image. Each recipe displays a page's image and text
+side-by-side.
Recipes:
- * `annotate_page_text`: Annotate a page's text with the page's image
- displayed side-by-side for reference.
+ * `annotate_page_text`: Annotate a page's text.
* `annotate_text_and_image`: Annotate both a page's text and image side-by-side.
+ * `review_page_spans`: Review existing page-level text annotations to produce
+ a final, adjudicated set of annotations.
Referenced images must be served out independently for display; the image url
prefix for images should be specified when initializing the recipe.
@@ -14,18 +16,24 @@
Example use:
```
prodigy annotate_page_text poetry_spans poetry_pages.jsonl --label POETRY,PROSODY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix http://localhost:8000/
-prodigy annotate_text_and_image poetry_text_image poetry_pages.jsonl -l POETRY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix http://localhost:8000/
+prodigy annotate_text_and_image poetry_text_image poetry_pages.jsonl -l POETRY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix ../ppa-web-images -FM
+prodigy review_page_spans adjudicate poetry_spans -l POETRY -F ../corppa/poetry_detection/annotation/recipe.py --image-prefix ../ppa-web-images -FM --sessions alice,bob
"""
+from collections import defaultdict
+from copy import deepcopy
from pathlib import Path
+from typing import Dict, Iterable, List, Optional
import spacy
-from prodigy import log
+from prodigy import log, set_hashes
+from prodigy.components.db import connect
from prodigy.components.preprocess import add_tokens
from prodigy.components.preprocess import fetch_media as fetch_media_preprocessor
from prodigy.components.stream import get_stream
from prodigy.core import Arg, recipe
-from prodigy.util import get_labels
+from prodigy.types import LabelsType, RecipeSettingsType, StreamType, TaskType
+from prodigy.util import INPUT_HASH_ATTR, SESSION_ID_ATTR, get_labels
#: reference to current directory, for use as Prodigy CSS directory
CURRENT_DIR = Path(__file__).parent.absolute()
@@ -49,22 +57,51 @@
"global_css_dir": CURRENT_DIR,
}
+#: color palette for predefined session names
+PALETTE = [
+ "#c5bdf4",
+ "#ffd882",
+ "#d9fbad",
+ "#c2f2f6",
+ "#ffdaf9",
+ "#b5c6c9",
+ "#96e8ce",
+ "#ffd1b2",
+]
-def add_image(task, image_prefix=None):
+
+def add_image(example: TaskType, image_prefix: Optional[str] = None):
+ """
+ Set an example's image field to its existing image_path with an optional prefix
+
+ Note: Assumes filepaths use forward slash
+ """
if image_prefix is None:
- task["image"] = task["image_path"]
+ example["image"] = example["image_path"]
else:
path_pfx = image_prefix.rstrip("/")
- task["image"] = f"{path_pfx}/{task['image_path']}"
- return task
+ example["image"] = f"{path_pfx}/{example['image_path']}"
+ return example
-def add_images(examples, image_prefix=None):
- for task in examples:
- yield add_image(task, image_prefix=image_prefix)
+def add_images(examples: StreamType, image_prefix: Optional[str] = None) -> StreamType:
+ """
+ Set the image field for each example in the stream
+
+ Calls: `add_image`
+ """
+ for example in examples:
+ yield add_image(example, image_prefix=image_prefix)
-def remove_images(examples, image_prefix=None):
+def remove_images(
+ examples: Iterable[TaskType], image_prefix: Optional[str] = None
+) -> List[TaskType]:
+ """
+ For each example, replace base64 data URIs with image filepath or URL
+
+ Calls: `add_image`
+ """
for task in examples:
# If "image" is a base64 string and "image_path" is present in the task,
# remove the image data
@@ -90,10 +127,10 @@ def remove_images(examples, image_prefix=None):
def annotate_text_and_image(
dataset: str,
source: str,
- labels: str,
+ labels: LabelsType = [],
image_prefix: str = None,
fetch_media: bool = False,
-):
+) -> RecipeSettingsType:
"""Annotate text and image side by side: allows adding manual spans
to both image and text. Intended for page-level annotation.
"""
@@ -105,25 +142,21 @@ def annotate_text_and_image(
stream.apply(add_images, image_prefix=image_prefix)
# optionally fetch media
if fetch_media:
- stream.apply(fetch_media_preprocessor, ["image"], skip=True)
-
- # split labels by commas and strip any whitespace
- label_list = get_labels(labels)
+ stream.apply(fetch_media_preprocessor, ["image"])
blocks = [
{
"view_id": "image_manual",
- "labels": label_list,
+ "labels": labels,
},
- {"view_id": "spans_manual", "labels": label_list},
+ {"view_id": "spans_manual", "labels": labels},
]
# copy the common config options and add blocks and labels
- config = PRODIGY_COMMON_CONFIG.copy()
+ config = deepcopy(PRODIGY_COMMON_CONFIG)
config.update(
{
"blocks": blocks,
- "labels": label_list,
"ner_manual_highlight_chars": True,
"image_manual_spans_key": "image_spans",
# limit image selection to rectangle only, no polygon or freehand
@@ -160,10 +193,10 @@ def annotate_text_and_image(
def annotate_page_text(
dataset: str,
source: str,
- labels: str,
+ labels: LabelsType = [],
image_prefix: str = None,
fetch_media: bool = False,
-):
+) -> RecipeSettingsType:
"""Annotate text with manual spans; displays an image side by side
with text for reference only (image cannot be annotated).
Intended for page-level annotation.
@@ -176,24 +209,20 @@ def annotate_page_text(
stream.apply(add_images, stream, image_prefix=image_prefix)
# optionally fetch media
if fetch_media:
- stream.apply(fetch_media_preprocessor, ["image"], skip=True)
-
- # split labels by commas and strip any whitespace
- label_list = get_labels(labels)
+ stream.apply(fetch_media_preprocessor, ["image"])
blocks = [
{
"view_id": "html",
"html_template": "
",
},
- {"view_id": "spans_manual", "labels": label_list},
+ {"view_id": "spans_manual", "labels": labels},
]
# copy the common config options and add blocks and labels
- config = PRODIGY_COMMON_CONFIG.copy()
+ config = deepcopy(PRODIGY_COMMON_CONFIG)
config.update(
{
"blocks": blocks,
- "labels": label_list,
"ner_manual_highlight_chars": True,
}
)
@@ -209,3 +238,192 @@ def annotate_page_text(
components["before_db"] = lambda x: remove_images(x, image_prefix=image_prefix)
return components
+
+
+class ReviewStream:
+ """
+ Stream of review examples. This mostly exists to expose a __len__ to show
+ total progress in the web interface.
+ """
+
+ def __init__(
+ self,
+ data: Dict[int, List[TaskType]],
+ image_prefix: Optional[str] = None,
+ fetch_media: bool = False,
+ ) -> None:
+ """
+ Initialize a review stream.
+
+ data: Merged data, with examples grouped by input hash.
+ image_prefix: Image prefix for creating image (full) paths
+ fetch_media: Whether to fetch task images.
+ """
+ self.n_examples = len(data)
+ self.data = self.get_data(data, image_prefix, fetch_media)
+
+ def __len__(self) -> int:
+ return self.n_examples
+
+ def __iter__(self) -> StreamType:
+ for example in self.data:
+ yield example
+
+ def create_review_example(self, versions: List[TaskType]) -> TaskType:
+ """
+ Create review example from several annotated versions.
+ """
+ # TODO: Make sure that no unmerged version content is preserved.
+ review_example = deepcopy(versions[-1])
+ # Merge spans
+ merged_spans = []
+ session_counts = {}
+ sessions = []
+
+ for version in versions:
+ session_id = version[SESSION_ID_ATTR]
+ # Assume: session name does not contain -
+ session_name = session_id.rsplit("-", maxsplit=1)[1]
+ if session_id not in session_counts:
+ session_counts[session_id] = 1
+ else:
+ session_name += f"-{session_counts[session_id]}"
+ session_counts[session_id] += 1
+ sessions.append(session_name)
+ if "spans" not in version:
+ # Not sure when an annotated example would be missing a spans field
+ continue
+ for span in version["spans"]:
+ new_span = span.copy()
+ span_label = span["label"]
+ new_span["label"] = f"{session_name}: {span_label}"
+ merged_spans.append(new_span)
+ review_example["spans"] = merged_spans
+ review_example["sessions"] = sessions
+ return review_example
+
+ def get_data(
+ self,
+ data: Dict[int, List[TaskType]],
+ image_prefix: Optional[str],
+ fetch_media: bool,
+ ) -> List[TaskType]:
+ """
+ Build review examples from data. Add images to each example.
+ """
+ examples = []
+ for _, versions in data.items():
+ review_example = self.create_review_example(versions)
+ review_example = add_image(review_example, image_prefix)
+ examples.append(review_example)
+ if fetch_media:
+ return fetch_media_preprocessor(examples, ["image"])
+ return examples
+
+
+def get_review_stream(
+ examples: Iterable[TaskType],
+ image_prefix: Optional[str] = None,
+ fetch_media: bool = False,
+) -> StreamType:
+ # Group examples by input (page_id, text)
+ grouped_examples = defaultdict(list)
+ for example in examples:
+ # Reset hashes
+ example = set_hashes(
+ example, overwrite=True, input_keys=["id", "text"], task_keys=["spans"]
+ )
+ input_hash = example[INPUT_HASH_ATTR]
+ grouped_examples[input_hash].append(example)
+ return ReviewStream(
+ grouped_examples, image_prefix=image_prefix, fetch_media=fetch_media
+ )
+
+
+@recipe(
+ "review_page_spans",
+ dataset=Arg(help="Dataset to save annotations to"),
+ input_dataset=Arg(help="Name of dataset to review"),
+ labels=Arg(
+ "--label",
+ "-l",
+ help="Comma-separated label(s) to annotate or text file with one label per line",
+ ),
+ image_prefix=Arg("--image-prefix", "-i", help="Base URL for images"),
+ fetch_media=Arg(
+ "--fetch-media", "-FM", help="Load images from local paths or URLs"
+ ),
+ sessions=Arg("--sessions", help="Comma-separated session names for coloring"),
+)
+def review_page_spans(
+ dataset: str,
+ input_dataset: str,
+ labels: LabelsType = [],
+ image_prefix: str = None,
+ fetch_media: bool = False,
+ sessions: List[str] = [],
+) -> RecipeSettingsType:
+ """
+ Review input text span annotations and annotate with manual spans to create
+ final, adjudicated annotations. Loads and displays input text span
+ annotations.
+ """
+ # Load annotations
+ DB = connect()
+ if input_dataset not in DB:
+ raise RecipeError(f"Can't find input dataset '{input_dataset}' in database")
+ annotations = DB.get_dataset_examples(input_dataset)
+
+ blocks = [
+ {
+ "view_id": "html",
+ "html_template": "",
+ },
+ {"view_id": "spans_manual", "labels": labels},
+ ]
+
+ def before_db(examples):
+ """
+ Modifies annotated examples before saving to the database:
+ * Remove image spans & tokens (unneeded fields)
+ * Reset image to (full) image path if image fetched
+ """
+ for example in examples:
+ # remove image spans
+ del example["image_spans"]
+ # remove tokens
+ del example["tokens"]
+ if fetch_media:
+ # reset image to path
+ example = add_image(example, image_prefix=image_prefix)
+ return examples
+
+ # Set label colors
+ label_colors = PRODIGY_COMMON_CONFIG["custom_theme"]["labels"].copy()
+ if sessions:
+ # Add session-label colors
+ for i, session in enumerate(sessions):
+ session_color = PALETTE[i % len(PALETTE)]
+ for label in labels:
+ label_colors[f"{session}: {label}"] = session_color
+
+ # copy the common config options and add blocks and labels
+ config = deepcopy(PRODIGY_COMMON_CONFIG)
+ config.update(
+ {
+ "blocks": blocks,
+ "ner_manual_highlight_chars": True,
+ "global_css_dir": CURRENT_DIR,
+ "custom_theme": {"labels": label_colors},
+ }
+ )
+
+ return {
+ "dataset": dataset,
+ "view_id": "blocks",
+ "stream": get_review_stream(
+ annotations, image_prefix=image_prefix, fetch_media=fetch_media
+ ),
+ "before_db": before_db,
+ "config": config,
+ }
From 93db63f79eb394108517355d415015500735c52e Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Wed, 23 Oct 2024 16:23:38 -0400
Subject: [PATCH 8/9] Update src/corppa/poetry_detection/annotation/recipe.py
Co-authored-by: Rebecca Sutton Koeser
---
src/corppa/poetry_detection/annotation/recipe.py | 1 +
1 file changed, 1 insertion(+)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index cc6db10..f606a9a 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -283,6 +283,7 @@ def create_review_example(self, versions: List[TaskType]) -> TaskType:
for version in versions:
session_id = version[SESSION_ID_ATTR]
# Assume: session name does not contain -
+ # full session name includes the dataset id; split to get the session name without dataset id
session_name = session_id.rsplit("-", maxsplit=1)[1]
if session_id not in session_counts:
session_counts[session_id] = 1
From a0780c0528e970d2f56364902225a3a09f66a256 Mon Sep 17 00:00:00 2001
From: Laure Thompson <602628+laurejt@users.noreply.github.com>
Date: Wed, 23 Oct 2024 16:33:44 -0400
Subject: [PATCH 9/9] Rename remove_images to remove_image_data
---
src/corppa/poetry_detection/annotation/recipe.py | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/src/corppa/poetry_detection/annotation/recipe.py b/src/corppa/poetry_detection/annotation/recipe.py
index f606a9a..b23beb7 100644
--- a/src/corppa/poetry_detection/annotation/recipe.py
+++ b/src/corppa/poetry_detection/annotation/recipe.py
@@ -94,7 +94,7 @@ def add_images(examples: StreamType, image_prefix: Optional[str] = None) -> Stre
yield add_image(example, image_prefix=image_prefix)
-def remove_images(
+def remove_image_data(
examples: Iterable[TaskType], image_prefix: Optional[str] = None
) -> List[TaskType]:
"""
@@ -172,7 +172,9 @@ def annotate_text_and_image(
}
if fetch_media:
- components["before_db"] = lambda x: remove_images(x, image_prefix=image_prefix)
+ components["before_db"] = lambda x: remove_image_data(
+ x, image_prefix=image_prefix
+ )
return components
@@ -235,7 +237,9 @@ def annotate_page_text(
}
if fetch_media:
- components["before_db"] = lambda x: remove_images(x, image_prefix=image_prefix)
+ components["before_db"] = lambda x: remove_image_data(
+ x, image_prefix=image_prefix
+ )
return components