Spaces:
Sleeping
Sleeping
Update utils/target_classifier.py
Browse files- utils/target_classifier.py +29 -10
utils/target_classifier.py
CHANGED
@@ -69,21 +69,40 @@ def target_classification(haystack_doc:pd.DataFrame,
|
|
69 |
x: Series object with the unique SDG covered in the document uploaded and
|
70 |
the number of times it is covered/discussed/count_of_paragraphs.
|
71 |
"""
|
72 |
-
|
|
|
|
|
|
|
|
|
73 |
if not classifier_model:
|
|
|
74 |
classifier_model = st.session_state['target_classifier']
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
|
76 |
-
results = classifier_model(list(haystack_doc.text))
|
77 |
-
labels_= [(l[0]['label'],
|
78 |
-
|
79 |
|
80 |
|
81 |
-
df1 = DataFrame(labels_, columns=["Target Label","Target Score"])
|
82 |
-
df = pd.concat([haystack_doc,df1],axis=1)
|
83 |
|
84 |
-
df = df.sort_values(by="Target Score", ascending=False).reset_index(drop=True)
|
85 |
-
df['Target Score'] = df['Target Score'].round(2)
|
86 |
-
df.index += 1
|
87 |
-
# df['Label_def'] = df['Target Label'].apply(lambda i: _lab_dict[i])
|
88 |
|
89 |
return df
|
|
|
69 |
x: Series object with the unique SDG covered in the document uploaded and
|
70 |
the number of times it is covered/discussed/count_of_paragraphs.
|
71 |
"""
|
72 |
+
|
73 |
+
logging.info("Working on target/action identification")
|
74 |
+
|
75 |
+
haystack_doc['Vulnerability Label'] = 'NA'
|
76 |
+
|
77 |
if not classifier_model:
|
78 |
+
|
79 |
classifier_model = st.session_state['target_classifier']
|
80 |
+
|
81 |
+
# Get predictions
|
82 |
+
predictions = classifier_model(list(haystack_doc.text))
|
83 |
+
|
84 |
+
# Get labels for predictions
|
85 |
+
pred_labels = getlabels(predictions)
|
86 |
+
|
87 |
+
# Save labels
|
88 |
+
haystack_doc['Target Label'] = pred_labels
|
89 |
+
|
90 |
+
|
91 |
+
# logging.info("Working on action/target extraction")
|
92 |
+
# if not classifier_model:
|
93 |
+
# classifier_model = st.session_state['target_classifier']
|
94 |
|
95 |
+
# results = classifier_model(list(haystack_doc.text))
|
96 |
+
# labels_= [(l[0]['label'],
|
97 |
+
# l[0]['score']) for l in results]
|
98 |
|
99 |
|
100 |
+
# df1 = DataFrame(labels_, columns=["Target Label","Target Score"])
|
101 |
+
# df = pd.concat([haystack_doc,df1],axis=1)
|
102 |
|
103 |
+
# df = df.sort_values(by="Target Score", ascending=False).reset_index(drop=True)
|
104 |
+
# df['Target Score'] = df['Target Score'].round(2)
|
105 |
+
# df.index += 1
|
106 |
+
# # df['Label_def'] = df['Target Label'].apply(lambda i: _lab_dict[i])
|
107 |
|
108 |
return df
|