Sentiment Analysis - #iPython #Dato #GraphLab Create - Rotten Tomatoes - Public Kaggle Data

In [1]:

import graphlab
import os 
os.getcwd()

Out[1]:

'C:\\Anaconda2\\envs\\dato-env'

In [2]:

df_1=graphlab.SFrame("train_1.tsv")

[INFO] Start server at: ipc:///tmp/graphlab_server-7380 - Server binary: C:\Anaconda2\envs\dato-env\lib\site-packages\graphlab\unity_server.exe - Server log: C:\Users\Rohit\AppData\Local\Temp\graphlab_server_1455190805.log.0
[INFO] GraphLab Server Version: 1.8.1

PROGRESS: Finished parsing file C:\Anaconda2\envs\dato-env\train_1.tsv
PROGRESS: Parsing completed. Parsed 100 lines in 0.483028 secs.
------------------------------------------------------
Inferred types from first line of file as 
column_type_hints=[long,long,str,long]
If parsing fails due to incorrect types, you can correct
the inferred type list above and pass it to read_csv in
the column_type_hints argument
------------------------------------------------------
PROGRESS: Finished parsing file C:\Anaconda2\envs\dato-env\train_1.tsv
PROGRESS: Parsing completed. Parsed 156060 lines in 0.227013 secs.

In [3]:

convert_func = lambda x:1 if x>3 else 0

In [4]:

df_1[100:110]

Out[4]:

PhraseId	SentenceId	Phrase	Sentiment
101	3	would have a hard time sitting through this one ...	1
102	3	would have a hard time sitting through this one ...	0
103	3	would	2
104	3	have a hard time sitting through this one ...	0
105	3	have	2
106	3	a hard time sitting through this one ...	1
107	3	a hard time	1
108	3	hard time	1
109	3	hard	2
110	3	time	2

[10 rows x 4 columns]

In [5]:

df_1[1000:1010]
# within 1000 Rows of the DF - the Sentence ID ahs moved from - 3 to 36 and Phrase ID 110 to 1001

Out[5]:

PhraseId	SentenceId	Phrase	Sentiment
1001	36	to avoid	1
1002	36	avoid	0
1003	37	It almost feels as if the movie is more interested ...	1
1004	37	almost feels as if the movie is more interested ...	0
1005	37	feels as if the movie is more interested in ...	1
1006	37	feels as if the movie is more interested in ...	1
1007	37	feels	2
1008	37	as if the movie is more interested in ...	1
1009	37	if the movie is more interested in ...	1
1010	37	if	2

[10 rows x 4 columns]

In [6]:

df_1['Target'] = df_1['Sentiment'].apply(convert_func)
#
# Creates a Variable named Target - adds a column at end of DF named Target .
# Values of Sentiment are converted as defined in the - "convert_func"
# 
df_1[100:110]

Out[6]:

PhraseId	SentenceId	Phrase	Sentiment	Target
101	3	would have a hard time sitting through this one ...	1	0
102	3	would have a hard time sitting through this one ...	0	0
103	3	would	2	0
104	3	have a hard time sitting through this one ...	0	0
105	3	have	2	0
106	3	a hard time sitting through this one ...	1	0
107	3	a hard time	1	0
108	3	hard time	1	0
109	3	hard	2	0
110	3	time	2	0

[10 rows x 5 columns]

In [8]:

df_1['word_count'] = graphlab.text_analytics.count_words(df_1['Phrase'])
#

In [12]:

df_1['Phrase'][1:2]
# Seen below we get 1 Row of the DataType String [str] , printed from the Variable Phrase in our Data Frame 
# As seen Frequency of Ocurrence of all Words is - ONCE -- besides  'the' which occurs Twice...

Out[12]:

dtype: str
Rows: 1
['A series of escapades demonstrating the adage that what is good for the goose']

In [9]:

df_1['word_count'][1:2]
# As seen below - a Row of the Dictionary , the dtype: dict - is printed within the Squiggly Braces ..
# Seen below the Frequency of Ocurrence of the Words is -- 1L -- for every word besides 'the' :2L

Out[9]:

dtype: dict
Rows: 1
[{'a': 1L, 'what': 1L, 'good': 1L, 'escapades': 1L, 'for': 1L, 'that': 1L, 'series': 1L, 'is': 1L, 'goose': 1L, 'adage': 1L, 'demonstrating': 1L, 'of': 1L, 'the': 2L}]

In [13]:

df_1['Phrase'][2:4]
# Two Rows are printed out - Row 1 has Two Words == 'A series'
# Row 2 has One Word == 'A'

Out[13]:

dtype: str
Rows: 2
['A series', 'A']

In [10]:

df_1['word_count'][2:4]
# Seen below - Two Row's of Dictionary printed , One Row each within the Squiggly Braces.
# Square Brackets surround COMPLETE output we have sought from the - Data Type == dtype: dict. 

Out[10]:

dtype: dict
Rows: 2
[{'a': 1L, 'series': 1L}, {'a': 1L}]

In [11]:

df_1['word_count'][2:10]

Out[11]:

dtype: dict
Rows: 8
[{'a': 1L, 'series': 1L}, {'a': 1L}, {'series': 1L}, {'what': 1L, 'good': 1L, 'for': 1L, 'escapades': 1L, 'that': 1L, 'of': 1L, 'is': 1L, 'goose': 1L, 'adage': 1L, 'demonstrating': 1L, 'the': 2L}, {'of': 1L}, {'what': 1L, 'good': 1L, 'escapades': 1L, 'for': 1L, 'that': 1L, 'is': 1L, 'goose': 1L, 'adage': 1L, 'demonstrating': 1L, 'the': 2L}, {'escapades': 1L}, {'what': 1L, 'good': 1L, 'for': 1L, 'that': 1L, 'is': 1L, 'goose': 1L, 'adage': 1L, 'demonstrating': 1L, 'the': 2L}]