Tum Chaturapruek and Raphael Townshend

Autumn 2014

Install the following through pip install

• scikit-learn
• nltk (also download data after installing)
• senticnet

Download MaltParser 1.8.1, place it at ~/maltparser-1.8.1, and create a symbolic link ln -s ~/maltparser-1.8.1/maltparser-1.8.1.jar /usr/local/bin/malt.jar

Download concept parser from the senticNet website and place it at ~/concept-parser.

Most results are stored at static/test_data/sample_data.json.

Caches are located in caches

Helper script is in scripts

Python code is in code

Visualization tools are in static

Other assorted data is in data

To use our visualizer, open static/map.html. Use Firefox instead of Safari for a better support of AngularJS.

static/map.js and static/admin.js contain all the logic components related to mapping and visualizing.

To run the scorers, run python code/process_reviews.py [foodName] 0 199

foodName can be any of beer, burger, burrito, lobster. 0 and 199 indicate that we're running from data sample 0 to data sample 199. You can change these numbers.

To evaluate the scorer output, run

python code/evaluator.py static/test_data/sample_data.json

To run the classification system, ensure that static/test/sample_data.json has been generated (revert, or regenerate if necessary). Then run python code/feature_based_sentiment_classifier.py [foodName]. Where foodName can be any of those specified in the rule-based section.