update guidelines

Author: snukky

EvalView/templates/EvalView/_scalar_instructions.html

@@ -1,18 +1,11 @@
 <div class="row">
   <div class="col-md-12">
-    <p>Assess the translation quality on the 10-point scale using quality levels as follows:</p>
-
-
     <ul class="list-unstyled">
-      <li><strong>Not acceptable (1-2)</strong>: The translation is misleading or incomprehensible, with severe meaning errors and broken language that make it unusable.</li>
-      <li><strong>Borderline (3-4)</strong>: The text conveys some meaning but contains frequent errors and awkward phrasing; it requires heavy editing to be useful.</li>
-      <li><strong>Acceptable (5-6)</strong>: The message is clear and mostly accurate, though minor issues in grammar or style remain; suitable with light to moderate editing.</li>
-      <li><strong>Good (7-8)</strong>: The translation is accurate and fluent with only rare awkwardness; tone and style are appropriate, needing minimal polishing.</li>
-      <li><strong>Excellent (9-10)</strong>: The text reads like a native-level, publication-ready piece: fully accurate, idiomatic, stylistically polished, and coherent.</li>
+      <li><strong>Very Good (9–10)</strong>: Complete meaning transfer. Sounds perfectly natural. (no or minimal proofreading)</li>
+      <li><strong>Good (7–8)</strong>: Near‑complete transfer, minor inaccuracies. Mostly natural, minor awkwardness. (needs light proofreading)</li>
+      <li><strong>Acceptable (5–6)</strong>: Main ideas conveyed but with noticeable inaccuracies or omissions. Uneven naturalness, awkward phrasing. (usable only after substantial revision)</li>
+      <li><strong>Borderline (3–4)</strong>: Partial transfer; frequent misinterpretation or omission that confuses the message. Often unnatural. (requires major rewrite)</li>
+      <li><strong>Not acceptable (1–2)</strong>: Violation of meaning; large portions mistranslated, missing, or incoherent. (unusable without complete retranslation)</li>
     </ul>
-
-    {% if doclvl %}
-    <p>Note that sentences in each paragraph can be separated by the <code>&lt;eos&gt;</code> or <code>&lt;br/&gt;</code> tags for convenience and this should not impact your assessment.</p>
-    {% endif %}
   </div>
 </div>

EvalView/templates/EvalView/pairwise-assessment-document.html

@@ -25,9 +25,10 @@
 .quotelike .row, .pseudoquotelike .row { font-size: 16px; margin: 0; padding: 10px 20px; }
 .quotelike.active { background-color: #f7f7f7; }
 
-.question-box { margin-bottom:10px; }
-.question-box p { font-size: 120%; margin: 10px 0 20px; font-style: italic; color: #31708f; }
-.question-box li { font-size: 100%; font-style: italic; color: #31708f; }
+.question-box { margin:20px 0; }
+.question-box p { font-size: 120%; margin: 10px 10px 0; color: #555; }
+.question-box ul { margin: 10px; }
+.question-box li { font-size: 100%; color: #555; }
 .document-box { padding-top: 10px; }
 .document-box p { font-size: 16px; }
 
@@ -864,13 +865,13 @@
                     {% if trusted_user %}
                     <span class="glyphicon glyphicon-ok-sign" aria-hidden="true"></span>
                     {% endif %}
-                    {{completed_blocks}}/{{total_blocks}} documents,
+                    {{completed_blocks}}/{{total_blocks}} screens,
                     <span id="items-left-counter">{{items_left_in_block}}</span>
-                    items left in document
+                    items left on this screen 
                 </span>
             </td>
             <td style="width:33%;text-align:center;">
-                {{campaign}}:#{{datask_id}}:Document #{{document_id}}-<span id="current-item-id">{{item_id}}</span>
+                {{campaign}}:{{datask_id}}:{{document_id}}-<span id="current-item-id">{{item_id}}</span>
             </td>
             <td style="width:33%;text-align:right;">
                 {% if source_language %}{{source_language}} &rarr; {% endif %}{{target_language}}
@@ -879,11 +880,11 @@
     </table>
 </div>
 
-<div class="question-box">
+<div class="question-box alert alert-success">
     <div class="row">
         <div class="col-sm-12">
             {% for text in priming_question_texts %}
-            <p>{{text|safe}}</p>
+            {{text|safe}}
             {% endfor %}
             {% if scalar_slider %}
                 {% with mono=monolingual doclvl=doc_guidelines %}
@@ -1121,9 +1122,9 @@
 
         {% else %}
 
-        <div class="question-box">
+        <div class="question-box alert alert-success">
             <div class="row">
-                <div class="col-sm-12 question-box">
+                <div class="col-sm-12">
                     {% for text in document_question_texts %}
                     <p>{{text|safe}}</p>
                     {% endfor %}

EvalView/views.py

@@ -2391,6 +2391,25 @@ def pairwise_assessment_document(request, code=None, campaign_name=None):
         'in {0} (right column) convey the original semantics of the source document '
         'in {1} (left column)? '.format(target_language, source_language),
     ]
+    
+    # new guidelines
+    priming_question_texts = [
+        '<p>'
+        f'Below is a document in {source_language} presented sentence by sentence. Each source sentence has been translated by two different systems, A and B, into {target_language}. '
+        'Your task is to rate each translation using the scale below, based on three criteria: <br/>'
+        '</p>'
+        '<p>'
+        f'<strong>Naturalness</strong>: Does the translation sound fluent in {target_language}?<br/>'
+        f'<strong>Accuracy</strong>: Does the translation correctly preserve the meaning of the source text?<br/>'
+        f'<strong>Coherence</strong>: Does the sentence translation fit well in the document context?<br/>'
+        '</p>'
+        '<p><em>Note</em>: You first evaluate each sentence individually, but you should use the entire translation as context.</p>'
+        '<p>Rating scale:</p>'
+    ]
+    document_question_texts = [
+        'For the final step, please look again at each translated document. ' 
+        'Provide one final, overall rating for each translation candidate, judging it as a whole. '
+    ]
 
     monolingual_task = 'monolingual' in campaign_opts
     use_sqm = 'sqm' in campaign_opts