Jupyter app

2022-11-22_ybc/sentiment_analysis_eval_comparison_.py

@@ -0,0 +1,162 @@
+#####################################################################
+# 1. code is adapted from the code below.                           #
+#   https://github.com/PetrKorab/The-Most-Favorable-Pre-trained     #
+#   -Sentiment-Classifiers-in-Python/blob/main/analysis.ipynb       #
+#   the purpose of the code:                                        #
+#   is to compare various pretrained analyzer                       #
+# 2. please make sure the libraries "vaderSentiment",               #
+#    "textblob","happytransformer" and "emoji" are installed        #
+#####################################################################
+from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
+import pandas as pd
+import numpy as np
+import requests
+import time
+from textblob import TextBlob
+from happytransformer import HappyTextClassification
+import matplotlib.pyplot as plt
+
+analyser = SentimentIntensityAnalyzer()
+
+playlist_df= pd.read_csv("my_playlist.csv")
+def get_lyric(artist_name,track_name):
+    # musixmatch api base url
+    base_url = "https://api.musixmatch.com/ws/1.1/"
+    # your api key
+    api_key = "&apikey=95f7e1aa1421c9807a18a2abdef5ce6e"
+    #api method
+    lyrics_matcher = "matcher.lyrics.get"
+    "&q_artist="
+    "&q_track="
+    api_call = base_url + lyrics_matcher + "?format=json&callback=callback" + "&q_artist=" + artist_name + "&q_track=" + track_name + api_key
+    request = requests.get(api_call)
+    data = request.json()
+    lyric = data['message']['body']['lyrics']['lyrics_body']
+    return lyric
+
+
+
+###########
+# 1.VADER #
+###########
+lyric_list = []
+sentiment_list = []
+sentiment_score_list = []
+
+for track0 in playlist_df[['artist_name','track_name']].values:
+
+
+    try:
+        song = get_lyric(track0[0], track0[1])
+        sentiment_score = analyser.polarity_scores(song)
+
+        if sentiment_score['compound'] >= 0.05:
+            sentiment_percentage = sentiment_score['compound']
+            sentiment = 'Positive'
+        elif sentiment_score['compound'] > -0.05 and sentiment_score['compound'] < 0.05:
+            sentiment_percentage = sentiment_score['compound']
+            sentiment = 'Neutral'
+        elif sentiment_score['compound'] <= -0.05:
+            sentiment_percentage = sentiment_score['compound']
+            sentiment = 'Negative'
+
+        sentiment_list.append(sentiment)
+        sentiment_score_list.append((abs(sentiment_percentage) * 100))
+
+    except:
+        sentiment_list.append('None')
+        sentiment_score_list.append(0)
+    lyric_list.append(song)
+    time.sleep(0.2)
+    #the end of the for loop
+playlist_df['lyric'] = lyric_list
+playlist_df['VADER_sentiment'] = sentiment_list
+playlist_df['VADER_sentiment_score'] = sentiment_score_list
+
+playlist_df.loc[playlist_df['VADER_sentiment']=="None",'VADER_sentiment'] = "Neutral"
+#########################
+# 2.textblob            #
+# pip install textblob  #
+#########################
+#The sentiment property returns a namedtuple of the form Sentiment(polarity, subjectivity).
+#The polarity score is a float within the range [-1.0, 1.0]. 
+#The subjectivity is a float within the range [0.0, 1.0] 
+#where 0.0 is very objective and 1.0 is very subjective.
+sentiment_list = []
+sentiment_score_list = []
+for lyric in lyric_list:
+    classifier = TextBlob(lyric)
+    polarity = classifier.sentiment.polarity
+    #subjectivity = classifier.sentiment.subjectivity
+    sentiment_score_list.append(polarity)
+
+playlist_df['textblob_sentiment_score'] = sentiment_score_list
+
+
+playlist_df.loc[playlist_df['textblob_sentiment_score']>0,'textblob_sentiment'] = "Positive"
+playlist_df.loc[playlist_df['textblob_sentiment_score']==0,'textblob_sentiment'] = "Neutral"
+playlist_df.loc[playlist_df['textblob_sentiment_score']<0,'textblob_sentiment'] = "Negative"
+################################
+# 3.Happy Transformer          #
+# pip install happytransformer #
+# pip3 install emoji==0.6.0    #
+################################
+#https://www.youtube.com/watch?v=Ew72EAgM7FM
+#models are found from the website https://huggingface.co/
+sentiment_list = []
+sentiment_score_list = []
+for lyric in lyric_list:
+    classifier = HappyTextClassification(model_type="RoBERTa", model_name="finiteautomata/bertweet-base-sentiment-analysis", num_labels=3)
+    try:
+        if len(lyric)>128:
+            lyric= lyric[:128]
+        polarity = classifier.classify_text(lyric)
+        sentiment_list.append(polarity.label)
+        sentiment_score_list.append(polarity.score)
+    except:
+        sentiment_list.append("NA")
+        sentiment_score_list.append(404)
+
+playlist_df['HF_sentiment'] = sentiment_list
+playlist_df['HF_sentiment_score'] = sentiment_score_list
+playlist_df.loc[playlist_df['HF_sentiment']=='POS','HF_sentiment'] = "Positive"
+playlist_df.loc[playlist_df['HF_sentiment']=='NEU','HF_sentiment'] = "Neutral"
+playlist_df.loc[playlist_df['HF_sentiment']=='NEG','HF_sentiment'] = "Negative"
+
+################################
+# 4. sentiment_consistency     #
+################################
+x =np.array(playlist_df['VADER_sentiment']==playlist_df['HF_sentiment'])*1
+y =np.array(playlist_df['VADER_sentiment']==playlist_df['textblob_sentiment'])*1
+z =np.array(playlist_df['HF_sentiment']==playlist_df['textblob_sentiment'])*1
+playlist_df['sentiment_consistency']=x+y+z
+# 3 means all the sentiments agrees
+# 1 means two sentiments agrees
+# 0 means no sentiments agrees
+# Pie chart, where the slices will be ordered and plotted counter-clockwise:
+labels = 'All agree', 'Two agrees', 'no agrees'
+x0=playlist_df.loc[playlist_df['sentiment_consistency']==3].shape[0]/playlist_df.shape[0]
+y0=playlist_df.loc[playlist_df['sentiment_consistency']==1].shape[0]/playlist_df.shape[0]
+z0=playlist_df.loc[playlist_df['sentiment_consistency']==0].shape[0]/playlist_df.shape[0]
+sizes = [x0, y0, z0]
+explode = (0, 0, 0.1)  # only "explode" the 3nd slice
+
+fig1, ax1 = plt.subplots()
+ax1.pie(sizes, explode=explode, labels=labels, autopct='%1.1f%%',
+        shadow=True, startangle=90)
+ax1.axis('equal')  # Equal aspect ratio ensures that pie is drawn as a circle.
+plt.title(f"total songs #: {playlist_df.shape[0]}")
+plt.savefig("sentiment_consistency.png", dpi =300,
+            transparent = True,
+            bbox_inches = 'tight')
+plt.show()
+
+
+#####################
+# 4.Export the data #
+#####################
+playlist_df=playlist_df.loc[:,['artist_name', 'artist_id', 'track_name', 'track_id', 'lyric',
+       'VADER_sentiment', 'VADER_sentiment_score', 
+       'textblob_sentiment', 'textblob_sentiment_score',
+       'HF_sentiment','HF_sentiment_score','sentiment_consistency']]
+playlist_df.to_csv('playlist_sentiment_comparison.csv',header=True, index=False)

README.md

@@ -1,3 +1,16 @@
 # CourseProject
 
-Please fork this repository and paste the github link of your fork on Microsoft CMT. Detailed instructions are on Coursera under Week 1: Course Project Overview/Week 9 Activities.
+1. find your user id
+go to the spotify profile webpage and check the webpage address. https://open.spotify.com/user/****
+
+**** is your spotify user id
+
+2. get a fresh OAuth Token
+from an spotify console. e.g. https://developer.spotify.com/console/get-playlists/
+
+3. edit the personal info in the "spotify_api.py" and export a playlist as "my_playlist.csv". I have a playlist of 401 songs.
+
+4. sentiment analysis
+try "sentiment_analysis_2.py" and export a table as "my_playlist_sentiment2.csv"
+
+