In [ ]:

    

# We're using the pymongo module to connect to the mLab database where the
# tweets are being stored
import pymongo


    

In [ ]:

    

# Set up the "client" we want to connect to
client = pymongo.MongoClient("mongodb://reader:reader@ds123614.mlab.com:23614/bmlm-parking-tweets")

# Get the correct database (there's only one!)
db = client["bmlm-parking-tweets"]

# Now get the right collection (there is also db.test)
collection = db.tweets


    

In [ ]:

    

# Make a list where each element is the "text" part of one tweet
tweets = [obj["text"] for obj in collection.find()]


    

In [ ]:

    

# Have a look at some of the tweets
tweets[0:10]


    

In [ ]:

    




    

In [ ]:

    

# Define some words to do with traffic and business
buzzwords = ["busy", "cars", "traffic", "crash", "accident", "standstill",
             "queue", "lorry", "roadworks"]


    

In [ ]:

    

# Get all the tweets where any of the buzzwords are present
traffic_tweets = [tweet for tweet in tweets if any(bw in tweet.split() for bw in buzzwords)]

# How many do we have?
print(len(traffic_tweets))


    

In [ ]:

    

# OK... what proportion of all the tweets we have is this?
print(len(traffic_tweets)/len(tweets))


    

In [ ]:

    




    

In [ ]:

    

# Huh. OK.


    

In [ ]:

    




    

In [ ]:

    

# Let's have a look at these tweets then...
traffic_tweets


    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    

# Bum.


    

In [ ]:

    




    

In [ ]:

    

# Let's do some sentiment analysis anyway :P
# Lots of this is borrowed from Roger's workshop: https://github.com/Bath-ML/sentiment-analysis-workshop
import nltk
print(nltk.__version__)


    

In [ ]:

    

# We're using the vader sentiment analysis library
from nltk.sentiment import vader
vanalyser = vader.SentimentIntensityAnalyzer()


    

In [ ]:

    

# Define a small function to return the sentiment scores
def vanalyse(sample):
    return vanalyser.polarity_scores(sample)


    

In [ ]:

    

# Try this out
vanalyse("I hate being stuck in traffic")


    

In [ ]:

    




    

In [ ]:

    

# Small function to extract compound score:
#      -1 ............0.............1
# super-negative   neutral   super-positive
def vaderSentiment(review):
    return vanalyse(review)["compound"]


    

In [ ]:

    

# e.g.
vaderSentiment("town is stupidly busy today")


    

In [ ]:

    

# Get scores for all our buzzword tweets
[vaderSentiment(tweet) for tweet in traffic_tweets]