Multiline, file mode, node threshold and plot improvments

app.py

@@ -1,8 +1,8 @@
 import gradio as gr
 
-from transformers import pipeline
 from functools import partial
 
+import pandas as pd
 import numpy as np
 
 import matplotlib.pyplot as plt
@@ -63,7 +63,8 @@ class FramingLabels:
 
         colors = [cp[0] if s > 0.5 else cp[1] for s in scores_ordered]
         ax.barh(label_names[::-1], scores_ordered[::-1], color=colors[::-1], **kwargs)
-
+        plt.xlim(left=0)
+        plt.tight_layout()
         return fig, ax
 
 class FramingDimensions:
@@ -108,6 +109,7 @@ class FramingDimensions:
         axi = ax.twinx()
         axi.set_ylim(ax.get_ylim())
         axi.set_yticks(ax.get_yticks(), labels=name_right)
+        plt.tight_layout()
         return fig
 
 class FramingStructure:
@@ -121,6 +123,7 @@ class FramingStructure:
             try:
                 return penman.decode(x["generated_text"])
             except:
+                # print(f"Decode error for {res}")
                 return None
         graphs = list(filter(lambda item: item is not None, [try_decode(x) for x in res]))
         return graphs
@@ -167,6 +170,7 @@ class FramingStructure:
         nx.draw_networkx(G_sub, pos, node_size=node_sizes, edge_color=edge_colors)
         nx.draw_networkx_labels(G_sub, pos)
         nx.draw_networkx_edge_labels(G_sub, pos, edge_labels=nx.get_edge_attributes(G_sub, "role"))
+        plt.tight_layout()
         return fig
 
 # Specify the models
@@ -217,26 +221,49 @@ framing_label_model = FramingLabels(base_model_1, candidate_labels)
 framing_dimen_model = FramingDimensions(base_model_2, dimensions, pole_names)
 framing_struc_model = FramingStructure(base_model_3)
 
-import pandas as pd
+def framing_multi(texts, min_node_threshold=1):
+    res1 = pd.DataFrame(framing_label_model(texts))
+    fig1, _ = framing_label_model.visualize(res1.mean().to_dict(), xerr=res1.sem())
+    fig2 = framing_dimen_model.visualize(pd.DataFrame(framing_dimen_model(texts)))
+    fig3 = framing_struc_model.visualize(framing_struc_model(texts), min_node_threshold=min_node_threshold)
 
-async def framing_single(text):
+    return fig1, fig2, fig3
+
+def framing_single(text, min_node_threshold=1):
     fig1, _ = framing_label_model.visualize(framing_label_model(text))
     fig2 = framing_dimen_model.visualize(pd.DataFrame({k: [v] for k, v in framing_dimen_model(text).items()}))
-    fig3 = framing_struc_model.visualize(framing_struc_model(text))
+    fig3 = framing_struc_model.visualize(framing_struc_model(text), min_node_threshold=min_node_threshold)
 
     return fig1, fig2, fig3
 
-example_list = ["In 2021, doctors prevented the spread of the virus by vaccinating with Pfizer.",
-                "We must fight for our freedom.",
-                "The government prevents our freedom.",
-                "They prevent the spread.",
-                "We fight the virus.",
-                "I believe that we should act now. There is no time to waste."
+async def framing_textbox(text, split, min_node_threshold):
+    texts = text.split("\n")
+    if split and len(texts) > 1:
+        return framing_multi(texts, min_node_threshold)
+    return framing_single(text, min_node_threshold)
+
+async def framing_file(file_obj, min_node_threshold):
+    with open(file_obj.name, "r") as f:
+        texts = f.readlines()
+    if len(texts) > 1:
+        return framing_multi(texts, min_node_threshold)
+    return framing_single(texts, min_node_threshold)
+
+example_list = [["In 2010, CFCs were banned internationally due to their harmful effect on the ozone layer.", False, 1],
+                ["In 2021, doctors prevented the spread of the virus by vaccinating with Pfizer.", False, 1],
+                ["We must fight for our freedom.", False, 1],
+                ["The government prevents our freedom.", False, 1],
+                ["They prevent the spread.", False, 1],
+                ["We fight the virus.", False, 1],
+                ["I believe that we should act now. There is no time to waste.", True, 1],
                 ]
 
-demo = gr.Interface(fn=framing_single,
-                           title="FrameFinder",
-                    inputs=gr.Textbox(label="Text to analyze."),
+textbox_inferface = gr.Interface(fn=framing_textbox,
+                    inputs=[
+                        gr.Textbox(label="Text to analyze."),
+                        gr.Checkbox(True, label="Split on newlines? (To enter newlines type shift+Enter)"),
+                        gr.Number(1, label="Min node threshold for framing structure.")
+                    ],
                     description="A simple tool that helps you find (discover and detect) frames in text.",
                     examples=example_list,
                     article="Check out the preliminary article in the [Web Conference Symposium](https://dl.acm.org/doi/pdf/10.1145/3543873.3587534), will be updated to currently in review article after publication.",
@@ -245,4 +272,19 @@ demo = gr.Interface(fn=framing_single,
                              gr.Plot(label="Structure")
                             ])
 
+file_interface = gr.Interface(fn=framing_file,
+                    inputs=[
+                        gr.File(label="File of texts to analyze."),
+                        gr.Number(1, label="Min node threshold for framing structure."),
+                    ],
+                    description="A simple tool that helps you find (discover and detect) frames in text.",
+                    article="Check out the preliminary article in the [Web Conference Symposium](https://dl.acm.org/doi/pdf/10.1145/3543873.3587534), will be updated to currently in review article after publication.",
+                    outputs=[gr.Plot(label="Label"),
+                             gr.Plot(label="Dimensions"),
+                             gr.Plot(label="Structure")])
+
+demo = gr.TabbedInterface([textbox_inferface, file_interface], 
+                        tab_names=["Single Mode", "File Mode"],
+                        title="FrameFinder",)
+
 demo.launch()