added my notebooks and additions to utils.py

evalstudent/utils.py

@@ -1,3 +1,4 @@
+import os
 import pandas as pd
 pd.options.mode.chained_assignment = None # Disabling pandas warnings triggered by display_classes
 from IPython.core.display import HTML
@@ -20,15 +21,15 @@ def display_classes(essay_id, train_df):
     Prints the essay text (keeping its exact original formatting) using colors to highlight discourse elements and their classes.
     Uses only `predictionstring`, which is useful to display models predictions.
     '''
-    
+
     # Handling submission format :
     discourse_type = "class" if "discourse_type" not in train_df.columns else "discourse_type"
-    
+
     elements_df = train_df[train_df["id"] == essay_id]
     essay_text = open(f'../../raw_data/train/{essay_id}.txt').read()
     essay_words = essay_text.split()
     formatted_essay = ""
-    
+
     # First we make sure discourse elements are in the text order
     elements_df["prediction_list"] = elements_df["predictionstring"].map(lambda x : x.split())
     elements_df["start_word_index"] = elements_df["prediction_list"].map(lambda x : int(x[0]))
@@ -38,7 +39,7 @@ def display_classes(essay_id, train_df):
     # and then we highlight the exact part of the essay corresponding to the discourse class.
     end_char = 0
     for i, element in elements_df.iterrows():
-        start_word = essay_words[element["start_word_index"]] 
+        start_word = essay_words[element["start_word_index"]]
         start_char = essay_text[end_char:].find(start_word) + len(essay_text[:end_char])
         formatted_essay += essay_text[end_char:start_char]
         for word_index in element["prediction_list"]:
@@ -68,4 +69,116 @@ def generate_predictionstring(discourse_start, discourse_end, essay_text):
     word_end = word_start + len(essay_text[char_start:char_end].split())
     word_end = min( word_end, len(essay_text.split()) )
     predictionstring = " ".join( [str(x) for x in range(word_start,word_end)] )
-    return predictionstring
+    return predictionstring
+
+## ADD ARTHUR ##
+
+def get_essay(id,mode='train'):
+    """Function to get the full text of an essay from the .txt file.
+
+    Args:
+        id_ (str): id of the essay
+        mode (str, optional): determines whether to access *train* or *test* texts. \
+            Defaults to 'train'.
+
+    Returns:
+        str: Returns the full text of the id
+    """
+    with open(os.path.join(os.path.dirname(os.path.dirname(__file__)),
+                           'raw_data',
+                           mode,
+                           f'{id}.txt'),'r') as file:
+        txt = file.read()
+    return txt
+
+def slicering(ps,txt):
+    """
+    Allow for predictionstring to match with corresponding words of an essay.
+    Given a predictionstring of a portion of a text and the full text, the
+    function returns the portion of the text corresponding to the predictionstring.
+
+    Args:
+        ps (str): predictionstring of a discourse
+        txt (str): full text of an essay
+
+    Returns:
+        str: portion of the text corresponding to the predictionstring
+    """
+    ps_l = ps.split()
+    txt = txt.split()
+
+    return ' '.join(txt[int(ps_l[0]):int(ps_l[-1])+1])
+
+
+def css():
+    """
+    Apply custom.css into the notebook
+
+    Returns:
+        str: HTML style tag
+    """
+    styles = open("./styles/custom.css", "r").read()
+    return HTML('<style>'+styles+'</style>')
+
+
+
+def render_html(df):
+    """
+    Transforms each discourse into a html string with appropriates tags for
+    visualization.
+
+    Args:
+        df (DataFrame): dataframe containing discourse_type and discourse_text
+
+    Returns:
+        str: html string
+    """
+    if 'class' in df.keys():
+        class_='class'
+    else:
+        class_='discourse_type'
+
+    html = "<{0} style='padding: 2px'>{1} <strong> [{0}] </strong></{0}>"\
+            .format(df[class_],df['discourse_text'])
+
+    return html
+
+
+def comparison_text(prediction, ground_truth):
+    """
+    Allow for visual comparison of an essay with predicted classes vs the essay
+    with the true classes
+
+    Args:
+        prediction (str): essay with predicted classes in html formatting
+        ground_truth (str): essay with true classes in html formatting
+
+    Returns:
+        html: visual table
+    """
+
+
+    html = f"""
+    <div class="content">
+     <span style="font-size:16px">Legend --></span>
+      <lead>Lead</lead>
+      <Position>Position</Position>
+      <Claim>Claim</Claim>
+      <Counterclaim>Counterclaim</Counterclaim>
+      <Rebuttal>Rebuttal</Rebuttal>
+      <Evidence>Evidence</Evidence>
+      <Concluding_Statement>Concluding_Statement</Concluding_Statement>
+    </div>
+
+    <div class="row">
+      <div class="column">
+        <h2 class="title">Prediction</h2>
+        <p style="text-align:justify">{prediction}</p>
+      </div>
+      <div class="column">
+        <h2 class="title">Ground Truth</h2>
+        <p style="text-align:justify">{ground_truth}</p>
+      </div>
+    </div>
+    """
+    return HTML(html)