Visualizations

.gitignore

@@ -1,6 +1,7 @@
 /venv/
 /.idea/
-/__pycache__/
+**/__pycache__/
+/wannadb_ui/__pycache__/
 /.pytest_cache/
 
 /models/
@@ -10,4 +11,5 @@
 .bson
 /evaluation/datasets/aviation/documents
 /evaluation/datasets/nobel/documents
-/evaluation/results/
+/evaluation/results/
+/logs/

main.py

@@ -1,6 +1,7 @@
 import logging
 import sys
 
+from PyQt6.QtCore import Qt
 from PyQt6.QtWidgets import QApplication
 
 from wannadb.resources import ResourceManager
@@ -14,6 +15,7 @@
 
     with ResourceManager() as resource_manager:
         # set up PyQt application
+        QApplication.setAttribute(Qt.ApplicationAttribute.AA_ShareOpenGLContexts)
         app = QApplication(sys.argv)
 
         window = MainWindow()

requirements.txt

@@ -69,7 +69,7 @@ murmurhash==1.0.9
     #   thinc
 nltk==3.8
     # via sentence-transformers
-numpy==1.21.4
+numpy==1.26.4
     # via
     #   aset (setup.py)
     #   blis
@@ -148,7 +148,7 @@ scikit-learn==1.0.1
     # via
     #   aset (setup.py)
     #   sentence-transformers
-scipy==1.7.2
+scipy==1.13.1
     # via
     #   aset (setup.py)
     #   scikit-learn
@@ -231,5 +231,11 @@ wasabi==0.10.1
     #   spacy
     #   thinc
 
+pyqtgraph==0.13.7
+
+PyOpenGL==3.1.7
+
+PyOpenGL_accelerate==3.1.7
+
 # The following packages are considered to be unsafe in a requirements file:
 # setuptools

scripts/preprocess.py

@@ -6,11 +6,12 @@
 from wannadb.configuration import Pipeline
 from wannadb.data.data import Document, DocumentBase
 from wannadb.interaction import EmptyInteractionCallback
+from wannadb.preprocessing.dimension_reduction import PCAReducer
 from wannadb.preprocessing.embedding import BERTContextSentenceEmbedder, RelativePositionEmbedder, SBERTTextEmbedder, SBERTLabelEmbedder
 from wannadb.preprocessing.extraction import StanzaNERExtractor, SpacyNERExtractor
 from wannadb.preprocessing.label_paraphrasing import OntoNotesLabelParaphraser, SplitAttributeNameLabelParaphraser
 from wannadb.preprocessing.normalization import CopyNormalizer
-from wannadb.preprocessing.other_processing import ContextSentenceCacher
+from wannadb.preprocessing.other_processing import ContextSentenceCacher, DuplicatedNuggetsCleaner
 from wannadb.resources import ResourceManager
 from wannadb.statistics import Statistics
 from wannadb.status import EmptyStatusCallback
@@ -68,7 +69,9 @@ def main() -> None:
             SBERTLabelEmbedder("SBERTBertLargeNliMeanTokensResource"),
             SBERTTextEmbedder("SBERTBertLargeNliMeanTokensResource"),
             BERTContextSentenceEmbedder("BertLargeCasedResource"),
-            RelativePositionEmbedder()
+            RelativePositionEmbedder(),
+            DuplicatedNuggetsCleaner(),
+            PCAReducer()
         ])
 
         document_base = DocumentBase(documents, [])

wannadb/change_captor.py

@@ -0,0 +1,220 @@
+"""
+Class providing model classes which can be utilized to capture the changes due a user feedback and propagate them to the
+UI.
+These changes are computed after every feedback of the user.
+"""
+
+from typing import Optional, Union, List
+
+from PyQt6.QtGui import QColor
+
+from wannadb.data.data import InformationNugget
+from wannadb_ui.common import ThresholdPosition, AddedReason
+
+
+class BestMatchUpdate:
+    """
+    Instances of this class represent an update of the best match of a document.
+
+    Each instance provide the old best match and the new best match of a document as well as the count specifying how
+    often similar changes of best guesses happened.
+    Another best match change is considered as similar if it happened in the same feedback round and the new best guess
+    is equal.
+
+    Methods
+    -------
+    old_best_match()
+        Returns the old best match of the related document.
+    new_best_match()
+        Returns the new best match of the related document.
+    count()
+        Returns the count of similar best match changes happened in the same feedback round.
+    """
+
+    def __init__(self, old_best_match: str, new_best_match: str, count: int):
+        """
+        Parameters
+        ----------
+        old_best_match: str
+            The old best match of the related document.
+        new_best_match: str
+            The new best match of the related document.
+        count: int
+            The count of similar best match changes happened in the same feedback round.
+        """
+
+        self._old_best_match: str = old_best_match
+        self._new_best_match: str = new_best_match
+        self._count: int = count
+
+    @property
+    def old_best_match(self) -> str:
+        return self._old_best_match
+
+    @property
+    def new_best_match(self) -> str:
+        return self._new_best_match
+
+    @property
+    def count(self) -> int:
+        return self._count
+
+
+class ThresholdPositionUpdate:
+    """
+    Instances of this class represent an update of the position of a nugget's distance relative to the threshold.
+
+    Each instance provide the text of the nugget whose position changed, the old position (above or below), the new
+    position (above or below), the old and new distance of the nugget as well as a count indicating how often similar
+    changes happened in the same feedback round.
+    A change is considered as similar if it happened in the same feedback round, the text represented by the nugget is
+    equal, and it has the same type of the update (above -> below or below -> above).
+
+    As mentioned, an instance of this class can cover multiple changes if the text of the nuggets with a change are
+    equal.
+    In this case the distance related properties are None as we don't refer to a single nugget.
+    """
+
+    def __init__(self,
+                 nugget_text: str,
+                 old_position: Optional[ThresholdPosition], new_position: ThresholdPosition,
+                 old_distance: Optional[float], new_distance: Optional[float],
+                 count: int):
+        """
+        Parameters
+        ----------
+        nugget_text: str
+            Text of the nuggets whose position relative to the threshold changed.
+        old_position: ThresholdPosition
+            Previous position of the covered nuggets relative to the threshold (above or below).
+        new_position: ThresholdPosition
+            New position of the covered nuggets relative to the threshold (above or below).
+        old_distance: float
+            Old distance associated with the nugget. If multiple nuggets are covered by this instance, this will be
+            None.
+        new_distance: float
+            New distance associated with the nugget. If multiple nuggets are covered by this instance, this will be
+            None.
+        count: int
+            Number of similar changes happened in the same feedback round.
+        """
+
+        self._best_guess: str = nugget_text
+        self._old_position: Optional[ThresholdPosition] = old_position
+        self._new_position: ThresholdPosition = new_position
+        self._old_distance: float = old_distance
+        self._new_distance: float = new_distance
+        self._count: int = count
+
+    @property
+    def nugget_text(self) -> str:
+        return self._best_guess
+
+    @property
+    def old_position(self) -> Optional[ThresholdPosition]:
+        return self._old_position
+
+    @property
+    def new_position(self) -> ThresholdPosition:
+        return self._new_position
+
+    @property
+    def old_distance(self) -> Optional[float]:
+        return self._old_distance
+
+    @property
+    def new_distance(self) -> Optional[float]:
+        return self._new_distance
+
+    @property
+    def count(self) -> int:
+        return self._count
+
+
+class NewlyAddedNuggetContext:
+    """
+    Instances of this class represent a newly added nugget to the document overview.
+    Each instance provide information about the old and new distance of the nugget as well as the reason why the system
+    newly added the nugget.
+    """
+
+    def __init__(self,
+                 nugget: InformationNugget,
+                 old_distance: Union[float, None],
+                 new_distance: float,
+                 added_reason: AddedReason):
+        """
+        Parameters
+        ----------
+        nugget: InformationNugget
+            Newly added nugget.
+        old_distance: float
+            Old distance associated with the nugget.
+        new_distance: float
+            New distance associated with the nugget.
+        added_reason: AddedReason
+            Reason for the nugget being newly added.
+        """
+
+        self._nugget = nugget
+        self._old_distance = old_distance
+        self._new_distance = new_distance
+        self._added_reason = added_reason
+
+    @property
+    def nugget(self):
+        return self._nugget
+
+    @property
+    def old_distance(self):
+        return self._old_distance
+
+    @property
+    def new_distance(self):
+        return self._new_distance
+
+    @property
+    def added_reason(self):
+        return self._added_reason
+
+
+class NuggetUpdatesContext:
+    """
+    Wrapper class wrapping multiple types of nugget related updates.
+    Nugget related updates refer to `NewlyAddedNuggetContext`, `ThresholdPositionUpdate` and `BestMatchUpdate`. Each
+    instance holds a list of updates for all of these 3 update types.
+    """
+
+    def __init__(self,
+                 newly_added_nugget_contexts: List[NewlyAddedNuggetContext],
+                 best_match_updates: List[BestMatchUpdate],
+                 threshold_position_updates: List[ThresholdPositionUpdate]):
+        """
+        Parameters
+        ----------
+        newly_added_nugget_contexts: List[NewlyAddedNuggetContext]
+            List of all `NewlyAddedNuggetContext` instances wrapped by this instance.
+        best_match_updates: List[BestMatchUpdate]
+            List of all `BestMatchUpdate` instances wrapped by this instance.
+        threshold_position_updates: List[ThresholdPositionUpdate]
+            List of all `ThresholdPositionUpdate` instances wrapped by this instance.
+        """
+
+        self._newly_added_nugget_contexts: List[NewlyAddedNuggetContext] = newly_added_nugget_contexts
+        self._best_match_updates: List[BestMatchUpdate] = best_match_updates
+        self._threshold_position_updates: List[ThresholdPositionUpdate] = threshold_position_updates
+
+    @property
+    def newly_added_nugget_contexts(self) -> List[NewlyAddedNuggetContext]:
+        return self._newly_added_nugget_contexts
+
+    @property
+    def best_match_updates(self) -> List[BestMatchUpdate]:
+        return self._best_match_updates
+
+    @property
+    def threshold_position_updates(self) -> List[ThresholdPositionUpdate]:
+        return self._threshold_position_updates
+
+
+