Skip to content

Commit

Permalink
Merge pull request #55 from inception-project/feature-51-phi-annotati…
Browse files Browse the repository at this point in the history 
…on-on-feature-level

Feature 51 phi annotation on feature level
  • Loading branch information
@serwarde
serwarde committed Aug 22, 2024
2 parents c71b9eb + f67441c commit 9a91a27
Show file tree
Hide file tree
Showing 2 changed files with 137 additions and 47 deletions.
4 changes: 2 additions & 2 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,12 +9,12 @@ In case the project is part of a bigger project, the project progress can be exp
## Getting started

1. Ensure you have python 3.11 or higher (check it using python --version)
2. Install the package using ``pip install inception-reports``
2. Install the package using ``pip install -U inception-reports``
3. Run the script using ``inception_reports --help`` this will show you the options you have:
1. Run it with ``inception_reports --manager`` if you have one python project, or are responsible for one location.
2. Run it with ``inception_reports --lead`` if you're leading multiple projects, or multiple locations.
4. Regardless of the mode you run the package with, it should open a browser window. Follow the instructions on the page to proceed.

## Issues

This package is under early developments, if you notice any bugs, or face any issues, do not hesitate to open a [Github issue](https://github.com/inception-project/inception-reporting-dashboard/issues).
This package is under early development, if you notice any bugs, or face any issues, do not hesitate to open a [Github issue](https://github.com/inception-project/inception-reporting-dashboard/issues).
180 changes: 135 additions & 45 deletions inception_reports/generate_reports_manager.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,7 @@ def startup():
</style>
<style>
section.main > div {max-width:90%}
section.main > div {max-width:95%}
</style>
""",
unsafe_allow_html=True,
Expand Down Expand Up @@ -269,7 +269,9 @@ def select_method_to_import_data():
st.sidebar.write(
"Please input the path to the folder containing the INCEpTION projects."
)
projects_folder = st.sidebar.text_input("Projects Folder:", value="")
projects_folder = st.sidebar.text_input(
"Projects Folder:", value="data/gemtex_demo_projects"
)
button = st.sidebar.button("Generate Reports")
if button:
st.session_state["method"] = "Manually"
Expand Down Expand Up @@ -345,7 +347,6 @@ def find_element_by_name(element_list, name):
return element.uiName
return name.split(".")[-1]


def get_type_counts(annotations):
"""
Calculate the count of each type in the given annotations. Each annotation is a CAS object.
Expand All @@ -357,40 +358,81 @@ def get_type_counts(annotations):
dict: A dictionary containing the count of each type, both total and per document.
The structure is {type_name: {'total': count, 'documents': {doc_id: count}}}.
"""
type_count_dict = {}

type_count = {}

# Assuming that all documents have the same layer definition
first_doc = next(iter(annotations.values()))
layer_definitions = first_doc.select(
"de.tudarmstadt.ukp.clarin.webanno.api.type.LayerDefinition"
)

# Define a set of types to exclude for clarity and performance
excluded_types = {
"uima.tcas.DocumentAnnotation",
"de.tudarmstadt.ukp.clarin.webanno.api.type.LayerDefinition",
"de.tudarmstadt.ukp.clarin.webanno.api.type.FeatureDefinition",
"de.tudarmstadt.ukp.dkpro.core.api.lexmorph.type.morph.MorphologicalFeatures",
"de.tudarmstadt.ukp.dkpro.core.api.metadata.type.DocumentMetaData",
"de.tudarmstadt.ukp.dkpro.core.api.metadata.type.TagDescription",
"de.tudarmstadt.ukp.dkpro.core.api.metadata.type.TagsetDescription",
None,
}

for doc_id, cas in annotations.items():
# Get the list of types for the current CAS object
type_names = [
(find_element_by_name(layer_definitions, t.name), len(cas.select(t.name)))
for t in cas.typesystem.get_types()
if t.name != "de.tudarmstadt.ukp.clarin.webanno.api.type.LayerDefinition"
# Get the list of relevant types for the current CAS object
relevant_types = [
t for t in cas.typesystem.get_types()
if t.name not in excluded_types
]

for type_name, count in type_names:
if type_name is None:
for t in relevant_types:
cas_select = cas.select(t.name)
count = len(cas_select)
if count == 0:
continue

if type_name not in type_count_dict:
type_count_dict[type_name] = {"total": 0, "documents": {}}
type_count_dict[type_name]["total"] += count
# Filter for the features that are relevant
annotations_features = [
feature for feature in t.all_features
if feature.name not in {
cassis.typesystem.FEATURE_BASE_NAME_END,
cassis.typesystem.FEATURE_BASE_NAME_BEGIN,
cassis.typesystem.FEATURE_BASE_NAME_SOFA,
}
]

# Get UI Name for layer type
type_name = find_element_by_name(layer_definitions, t.name)

if doc_id not in type_count_dict[type_name]["documents"]:
type_count_dict[type_name]["documents"][doc_id] = 0
type_count_dict[type_name]["documents"][doc_id] += count
if type_name not in type_count:
type_count[type_name] = {
"total": 0,
"documents": {},
"features": {}
}

type_count_dict = {k: v for k, v in type_count_dict.items() if v["total"] > 0}
type_count_dict = dict(
sorted(type_count_dict.items(), key=lambda item: item[1]["total"])
)
type_count[type_name]["total"] += count
type_count[type_name]["documents"].setdefault(doc_id, 0)
type_count[type_name]["documents"][doc_id] += count

return type_count_dict
# Count the feature occurrences within the selected CAS
for feature in annotations_features:
for cas_item in cas_select:
feature_value = cas_item.get(feature.name)
if feature_value is None:
continue
if feature_value not in type_count[type_name]["features"]:
type_count[type_name]["features"][feature_value] = {}

type_count[type_name]["features"][feature_value].setdefault(doc_id, 0)
type_count[type_name]["features"][feature_value][doc_id] += 1


for type_name, type_data in type_count.items():
type_data["features"] = dict(sorted(type_data["features"].items(), key=lambda x: sum(x[1].values()), reverse=True))
type_count = dict(sorted(type_count.items(), key=lambda item: item[1]["total"], reverse=True))
return type_count


def export_data(project_data, output_directory=None):
Expand Down Expand Up @@ -524,15 +566,7 @@ def plot_project_progress(project) -> None:
{"Labels": pie_labels, "Sizes": data_sizes_tokens}
).sort_values(by="Labels", ascending=True)

df_bar = pd.DataFrame(
{
"Types": type_counts.keys(),
"Counts": [type_counts[type_name]["total"] for type_name in type_counts],
}
)

pie_chart = go.Figure()

pie_chart.add_trace(
go.Pie(
labels=df_pie_docs["Labels"],
Expand Down Expand Up @@ -591,21 +625,63 @@ def plot_project_progress(project) -> None:
}
],
)

bar_chart = go.Figure()

for _, row in df_bar.iterrows():
bar_chart.add_trace(
go.Bar(
y=[row["Types"]],
x=[row["Counts"]],
orientation="h",
name=row["Types"],
legendgroup=row["Types"],
showlegend=True,
hoverinfo="x+y",
main_traces = 0
feature_traces = 0

# Add bar traces for the total counts by category
for category, details in type_counts.items():
bar_chart.add_trace(go.Bar(
y=[category],
x=[details["total"]],
text=[details["total"]],
textposition='auto',
name=category.capitalize(),
visible=True,
orientation="h",
hoverinfo="x+y"
))
main_traces += 1

feature_buttons = []
for category, details in type_counts.items():
if len(details['features']) >= 2:
for subcategory, subvalues in details['features'].items():
bar_chart.add_trace(go.Bar(
y=[subcategory],
x=[sum(subvalues.values())],
text=[sum(subvalues.values())],
textposition='auto',
name=subcategory,
visible=False,
orientation="h",
hoverinfo="x+y"
))
feature_traces += 1

visibility = [False] * main_traces + [True] * feature_traces

feature_buttons.append(
{
"args": [
{"visible": visibility}
],
"label": category,
"method": "update"
}
)
)

bar_chart_buttons = [
{
"args": [
{"visible": [True] * main_traces + [False] * feature_traces}
],
"label": "Overview",
"method": "update"
}
] + feature_buttons

bar_chart.update_layout(
title=dict(
Expand All @@ -617,15 +693,29 @@ def plot_project_progress(project) -> None:
),
xaxis_title="Number of Annotations",
barmode="overlay",
height=60 * len(df_bar),
height=160 * len(type_counts),
font=dict(size=18),
legend=dict(font=dict(size=12)),
legend=dict(font=dict(size=10)),
paper_bgcolor="rgba(0,0,0,0)",
plot_bgcolor="rgba(0,0,0,0)",
margin=dict(l=40, r=40),
margin=dict(l=10, r=10),
colorway=px.colors.qualitative.Plotly,
)

bar_chart.update_layout(
updatemenus=[
{
"buttons": bar_chart_buttons,
"direction": "down",
"showactive": True,
"x": 0.45,
"y": 1.15,
"xanchor": "center",
"yanchor": "top",
}
]
)

col1, _, col3 = st.columns([1, 0.1, 1])
with col1:
st.plotly_chart(pie_chart, use_container_width=True)
Expand Down

0 comments on commit 9a91a27

Please sign in to comment.