I had a few hours in the morning before the conference schedule kicked in, so I decided to write a little HDF 2.0 flow to grab all the tweets about Strata Hadoop conference.
First up, I used GetTwitter to read tweets and filtered on these terms:
strata,stratahadoop,strataconf,NIFI,FutureOfData,ApacheNiFi,Hortonworks,Hadoop,ApacheHive,HBase,ApacheSpark,ApacheTez,MachineLearning,ApachePhoenix,ApacheCalcite,ApacheStorm,ApacheAtlas,ApacheKnox,Apache Ranger, HDFS, Apache Pig, Accumulo, Apache Flume, Sqoop, Apache Falcon
Input:
InvokeHttp: I used this to download the first image URL from tweets.
GetTwitter: This is our primary source of data and the most important. You must have a twitter account, a twitter developer account and create a twitter application. Then you can access the keywords and hashtags above. So far I've ingested 14,211 tweets into Phoenix. This included many times I've shut it down for testing and moving things around. I've had this run live as I've added pieces. I do not recommend this development process, but it's good for exploring data.
Processing:
RouteOnAttribute: To only process tweets with an actual messages, sometimes they are damaged or missing. Don't waste our time.
ExecuteStreamCommand: To call shell scripts that call TensorFlow C++ binaries and Python scripts. Many ways to do this, but this is the easiest.
UpdateAttribute: To change the file name for files I downloaded to HDFS.
For output sinks:
PutHDFS: Saved to HDFS in a few different directories (the first attached image); the raw JSON tweet, a limited set of fields such as handle, message, geolocation and a fully processed file that I added TensorFlow Inception v3 image recognition for images attached to Strata tweets and sentiment analysis using VADER on the text of the tweet.
PutSQL: I upserted all tweets that were enriched with HDF called TensorFlow & Python Sentiment Analysis into a Phoenix Table;
PutSlack: https://nifi-se.slack.com/messages/general/
Visualization:
There are a ton of ways to look at this data now.
I used Apache Zeppelin since it was part of my HDP 2.5 cluster and it's so easy to use. I added a few tables, charts and did quick SQL exploration of the data in Phoenix.
Linux Shell Scripts
source /usr/local/lib/bazel/bin/bazel-complete.bash export JAVA_HOME=/opt/jdk1.8.0_101/ /bin/rm -rf /tmp/$@ hdfs dfs -get /twitter/rawimage/$@ /tmp/ /opt/demo/tensorflow/bazel-bin/tensorflow/examples/label_image/label_image --image="/tmp/$@" --output_layer="softmax:0" --input_layer="Mul:0" --input_std=128 --input_mean=128 --graph=/opt/demo/tensorflow/tensorflow/examples/label_image/data/tensorflow_inception_graph.pb --labels=/opt/demo/tensorflow/tensorflow/examples/label_image/data/imagenet_comp_graph_label_strings.txt 2>&1| cut -c48- /bin/rm -rf /tmp/$@
python /opt/demo/sentiment/sentiment2.py "$@"
Python Script
If you have Python 2.7 installed, in previous articles I have shown how to install PiP and NLTK. Very easy to do some simple Sentiment Analysis. I also have a version where I just return the polarity_scores (compound, negative, neutral and positive).
from nltk.sentiment.vader import SentimentIntensityAnalyzer
import sys
sid = SentimentIntensityAnalyzer()
ss = sid.polarity_scores(sys.argv[1])
if ss['compound'] == 0.00:
print('Neutral')
elif ss['compound'] < 0.00:
print ('Negative')
else:
print('Positive')NIFI 1.0.0 Flow Template
Resources:
http://conferences.oreilly.com/strata/hadoop-big-data-ny/public/schedule/grid/public/
