Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,140 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pandas as pd
|
2 |
+
import plotly.express as px
|
3 |
+
import plotly.graph_objects as go
|
4 |
+
import streamlit as st
|
5 |
+
import tweepy
|
6 |
+
from plotly.subplots import make_subplots
|
7 |
+
from transformers import pipeline
|
8 |
+
|
9 |
+
consumer_key = "sHz78Xj5Dl41cqfzEHVoRcaKo"
|
10 |
+
consumer_secret = "3y5caZfu91nmB2MNH7mDSu5Cgf5qaVRpMfbDoCPW4dU7E46k03"
|
11 |
+
access_key = "1116912581434695680-x359MscPSdqEcJzoIlg4jMsCZRdyNX"
|
12 |
+
access_secret = "wEsALFUava2TnYXWnuacrzSK4eiYfJUFLBRWPqGuMRnTz"
|
13 |
+
|
14 |
+
auth = tweepy.OAuthHandler(consumer_key,consumer_secret)
|
15 |
+
auth.set_access_token(access_key,access_secret)
|
16 |
+
api = tweepy.API(auth)
|
17 |
+
|
18 |
+
|
19 |
+
def get_tweets(username, count):
|
20 |
+
tweets = tweepy.Cursor(
|
21 |
+
api.user_timeline,
|
22 |
+
screen_name=username,
|
23 |
+
tweet_mode="extended",
|
24 |
+
exclude_replies=True,
|
25 |
+
include_rts=False,
|
26 |
+
).items(count)
|
27 |
+
|
28 |
+
tweets = list(tweets)
|
29 |
+
response = {
|
30 |
+
"tweets": [tweet.full_text.replace("\n", "").lower() for tweet in tweets],
|
31 |
+
"timestamps": [str(tweet.created_at) for tweet in tweets],
|
32 |
+
"retweets": [tweet.retweet_count for tweet in tweets],
|
33 |
+
"likes": [tweet.favorite_count for tweet in tweets],
|
34 |
+
}
|
35 |
+
return response
|
36 |
+
|
37 |
+
|
38 |
+
def get_sentiment(texts):
|
39 |
+
preds = pipe(texts)
|
40 |
+
|
41 |
+
response = dict()
|
42 |
+
response["labels"] = [pred["label"] for pred in preds]
|
43 |
+
response["scores"] = [pred["score"] for pred in preds]
|
44 |
+
return response
|
45 |
+
|
46 |
+
|
47 |
+
def neutralise_sentiment(preds):
|
48 |
+
for i, (label, score) in enumerate(zip(preds["labels"], preds["scores"])):
|
49 |
+
if score < 0.5:
|
50 |
+
preds["labels"][i] = "neutral"
|
51 |
+
preds["scores"][i] = 1.0 - score
|
52 |
+
|
53 |
+
|
54 |
+
def get_aggregation_period(df):
|
55 |
+
t_min, t_max = df["timestamps"].min(), df["timestamps"].max()
|
56 |
+
t_delta = t_max - t_min
|
57 |
+
if t_delta < pd.to_timedelta("30D"):
|
58 |
+
return "1D"
|
59 |
+
elif t_delta < pd.to_timedelta("365D"):
|
60 |
+
return "7D"
|
61 |
+
else:
|
62 |
+
return "30D"
|
63 |
+
|
64 |
+
|
65 |
+
@st.cache_data
|
66 |
+
def load_model():
|
67 |
+
pipe = pipeline(task="sentiment-analysis", model="bhadresh-savani/distilbert-base-uncased-emotion")
|
68 |
+
return pipe
|
69 |
+
|
70 |
+
|
71 |
+
"""
|
72 |
+
# Twitter Emotion Analyser
|
73 |
+
"""
|
74 |
+
|
75 |
+
|
76 |
+
pipe = load_model()
|
77 |
+
twitter_handle = st.sidebar.text_input("Twitter handle:", "elonmusk")
|
78 |
+
twitter_count = st.sidebar.selectbox("Number of tweets:", (10, 30, 50, 100))
|
79 |
+
|
80 |
+
|
81 |
+
if st.sidebar.button("Get tweets!"):
|
82 |
+
tweets = get_tweets(twitter_handle, twitter_count)
|
83 |
+
preds = get_sentiment(tweets["tweets"])
|
84 |
+
# neutralise_sentiment(preds)
|
85 |
+
tweets.update(preds)
|
86 |
+
# dataframe creation + preprocessing
|
87 |
+
df = pd.DataFrame(tweets)
|
88 |
+
df["timestamps"] = pd.to_datetime(df["timestamps"])
|
89 |
+
# plots
|
90 |
+
agg_period = get_aggregation_period(df)
|
91 |
+
ts_sentiment = (
|
92 |
+
df.groupby(["timestamps", "labels"])
|
93 |
+
.count()["likes"]
|
94 |
+
.unstack()
|
95 |
+
.resample(agg_period)
|
96 |
+
.count()
|
97 |
+
.stack()
|
98 |
+
.reset_index()
|
99 |
+
)
|
100 |
+
ts_sentiment.columns = ["timestamp", "label", "count"]
|
101 |
+
|
102 |
+
fig = make_subplots(rows=1, cols=2, horizontal_spacing=0.15)
|
103 |
+
|
104 |
+
# TODO: check that stacking makes sense!
|
105 |
+
for label in ts_sentiment["label"].unique():
|
106 |
+
fig.add_trace(
|
107 |
+
go.Scatter(
|
108 |
+
x=ts_sentiment.query("label == @label")["timestamp"],
|
109 |
+
y=ts_sentiment.query("label == @label")["count"],
|
110 |
+
mode="lines",
|
111 |
+
name=label,
|
112 |
+
stackgroup="one",
|
113 |
+
hoverinfo="x+y",
|
114 |
+
),
|
115 |
+
row=1,
|
116 |
+
col=1,
|
117 |
+
)
|
118 |
+
|
119 |
+
likes_per_label = df.groupby("labels")["likes"].mean().reset_index()
|
120 |
+
|
121 |
+
fig.add_trace(
|
122 |
+
go.Bar(
|
123 |
+
x=likes_per_label["labels"],
|
124 |
+
y=likes_per_label["likes"],
|
125 |
+
showlegend=False,
|
126 |
+
marker_color=px.colors.qualitative.Plotly,
|
127 |
+
opacity=0.6,
|
128 |
+
),
|
129 |
+
row=1,
|
130 |
+
col=2,
|
131 |
+
)
|
132 |
+
|
133 |
+
fig.update_yaxes(title_text="Number of Tweets", row=1, col=1)
|
134 |
+
fig.update_yaxes(title_text="Number of Likes", row=1, col=2)
|
135 |
+
fig.update_layout(height=350, width=750)
|
136 |
+
|
137 |
+
st.plotly_chart(fig)
|
138 |
+
|
139 |
+
# tweet sample
|
140 |
+
st.markdown(df.to_markdown())
|