ultima versao

app.py

@@ -1,11 +1,9 @@
 import gradio as gr
-from sklearn.feature_extraction.text import TfidfVectorizer
-from sklearn.feature_selection import SelectKBest
 import pandas as pd
 from textblob import TextBlob
 import textstat
 from huggingface_hub import hf_hub_download
-import joblib
+from joblib import load
 
 titulo1 = """CLONE - Studio Dashboard: "default" and "Default Project" does not give clear information about Alloy and Project unless description is read."""
 descricao1 = """Steps To Reproduce: 1. On dashboard on studio 3.0, navigate to Develop tab. 2. Notice "default" and "Default Project" & "two-tabbed" and "Tabbed Application" names. Actual: User does not get clear information from names that one is alloy project and another one is Titanium project unless he reads the description below. Expected: Naming convention or icon corresponding must suggest type"""
@@ -22,32 +20,24 @@ descricao4 = """During the compile process Alloy will attempt to remove files fr
 titulo5 = """Resolve suboptimal compression from uglify-js v2 update"""
 descricao5 = """The v2 update of uglify-js in Alloy, specifically version 2.2.5, has some suboptimal compressions, which are causing the optimizer.js test spec to fail in certain cases. Specifically the issues are around booleans and cascading of variables in assignments. These issues have been logged with the Uglifyjs2 project in the following links:    * https://github.com/mishoo/UglifyJS2/issues/137  * https://github.com/mishoo/UglifyJS2/issues/138    When these issues are resolved and distributed in an npm release, we need to revisit these compressions and testing to ensure that the fixes are in place, and that new uglify-js version has no regressions that impact alloy."""
 
-
-from huggingface_hub import hf_hub_download
-import joblib
-
 def calcula_MbR(titulo, descricao):
     context = titulo + descricao
     d = {"context": [context]}
     df = pd.DataFrame(data=d, columns=["context"])
-
-    model = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_mbr.joblib")
-    )
-
-    #model = load("model/model_tawos_aloy_mbr.pkl")
-
+    model = load(hf_hub_download("model_effort_tawos", "model_tawos_aloy_mbr.joblib"))
     story_points_MbR = model.predict(df["context"])
     return story_points_MbR
 
-def calcula_neosp(titulo, descricao):
-
-    model = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_neosp.joblib")
-    )
-    
-    # model = load("model/model_tawos_aloy_neosp.pkl")
+def calcula_Median(titulo, descricao):
+    context = titulo + descricao
+    d = {"context": [context]}
+    df = pd.DataFrame(data=d, columns=["context"])
+    model = load(hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_median.joblib"))
+    story_points_MbR = model.predict(df["context"])
+    return story_points_MbR
 
+def calcula_NEOSP_SVR(titulo, descricao):
+    model = load(hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_neosp_svr.joblib"))
     context = titulo + descricao
     d = {"context": [context]}
     df = pd.DataFrame(data=d, columns=["context"])
@@ -73,59 +63,65 @@ def calcula_neosp(titulo, descricao):
 
     story_points = model.predict(X)
     return story_points
-    
-def calculaTFIDFSVM(titulo, descricao):
-
-    model = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_tfidfsvm.joblib")
-    )
-
-    # model = load("model/model_tawos_aloy_tfidfsvm.pkl")
 
+def calcula_NEOSP_Linear(titulo, descricao):
+    model = load(hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_neosp_linear.joblib"))
     context = titulo + descricao
     d = {"context": [context]}
     df = pd.DataFrame(data=d, columns=["context"])
 
-    vectorizer = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "vectorizer_tfidfsvm.joblib")
-    )
+    # features de legibilidade
+    df["gunning_fog"] = df['context'].apply(textstat.gunning_fog)#
+    df["flesch_reading_ease"] = df['context'].apply(textstat.flesch_reading_ease)#
+    df["flesch_kincaid_grade"] = df['context'].apply(textstat.flesch_kincaid_grade)#
+    df["smog_index"] = df['context'].apply(textstat.smog_index)
+    df["coleman_liau_index"] = df['context'].apply(textstat.coleman_liau_index)#
+    df["automated_readability_index"] = df['context'].apply(textstat.automated_readability_index) #
+    df["dale_chall_readability_score"] = df['context'].apply(textstat.dale_chall_readability_score)#
+    df["difficult_words"] = df['context'].apply(textstat.difficult_words)
+    df["linsear_write_formula"] = df['context'].apply(textstat.linsear_write_formula)#
 
-    # vectorizer = load("model/vectorizer_tfidfsvm.pkl")
+    # feature de sentimento
+    df["polarity"] = df["context"].apply(lambda x: TextBlob(x).sentiment.polarity)
+    df["subjectivity"] = df["context"].apply(lambda x: TextBlob(x).sentiment.subjectivity)
 
+    X = df[["gunning_fog", "flesch_reading_ease", "flesch_kincaid_grade", "smog_index", "coleman_liau_index", 
+            "automated_readability_index", "dale_chall_readability_score", "difficult_words", "linsear_write_formula", 
+            "polarity", "subjectivity"]]
+
+    story_points = model.predict(X)
+    return story_points
+    
+def calcula_TFIDF_SVR(titulo, descricao):
+    model = load(hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_tfidf_svr.joblib"))
+    context = titulo + descricao
+    d = {"context": [context]}
+    df = pd.DataFrame(data=d, columns=["context"])
+    vectorizer = load(hf_hub_download("giseldo/model_effort_tawos", "vectorizer_tfidf.joblib"))
     X = vectorizer.transform(df["context"])
     story_points = model.predict(X)
     return story_points
 
-def calculaTFIDFLinear(titulo, descricao):
-
-    model = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_tfidflinear.joblib")
-    )
-
-    # model = load("model/model_tawos_aloy_tfidflinear.pkl")
-
+def calcula_TFIDF_Linear(titulo, descricao):
+    model = load(hf_hub_download("giseldo/model_effort_tawos", "model_tawos_aloy_tfidf_linear.joblib"))
     context = titulo + descricao
     d = {"context": [context]}
     df = pd.DataFrame(data=d, columns=["context"])
-
-    vectorizer = joblib.load(
-	    hf_hub_download("giseldo/model_effort_tawos", "vectorizer_tfidflinear.joblib")
-    )
-
-    # vectorizer = load("model/vectorizer_tfidflinear.pkl")
-
+    vectorizer = load(hf_hub_download("giseldo/model_effort_tawos", "vectorizer_tfidf.joblib"))
     X = vectorizer.transform(df["context"])
     story_points = model.predict(X)
     return story_points
 
 def calcula(titulo, descricao):
-    return calcula_MbR(titulo, descricao), calcula_neosp(titulo, descricao), calculaTFIDFSVM(titulo, descricao), calculaTFIDFLinear(titulo, descricao)
+    return calcula_MbR(titulo, descricao), calcula_Median(titulo, descricao), calcula_NEOSP_SVR(titulo, descricao), calcula_NEOSP_Linear(titulo, descricao), calcula_TFIDF_SVR(titulo, descricao), calcula_TFIDF_Linear(titulo, descricao)
 
 demo = gr.Interface(fn=calcula, 
                     inputs=[gr.Textbox(placeholder="Título", label="Título"), 
                             gr.Textbox(lines=10, placeholder="Descrição", label="Descrição")], 
-                    outputs=[gr.Textbox(label="Story Points Estimado MbR"),
+                    outputs=[gr.Textbox(label="Story Points Estimado Média"),
+                             gr.Textbox(label="Story Points Estimado Mediana"),
                              gr.Textbox(label="Story Points Estimado NEOSP-SVR"),
+                             gr.Textbox(label="Story Points Estimado NEOSP-Linear"),
                              gr.Textbox(label="Story Points Estimado TFIDF-SVR"),
                              gr.Textbox(label="Story Points Estimado TFIDF-Linear")],
                     title="Agile Task Story Point Estimator",