PAL

Installation & Running

Requirements

Python 2.7 with pip
- Required Python packages in requirements.txt
PostgreSQL 9.3 required for directory service
Node.js > v0.10 for pizza service
Web server capable of running wsgi applications for deployment

Base Installation

We recommend installing PAL in a python virtual environment by using the virtualenv package. Use pip to install virtualenv then create a new virtualenv in the pal directory

virtualenv env

and activate the virtual environment

source env/bin/activate

If not using the directory service and/or PostgreSQL is not installed, remove the line containing psycopg2 from the requirements.txt file. Install the required packages

pip install -r requirements.txt

Next, install the required nltk libraries. Open a python shell and type

>>> import nltk
>>> nltk.download()

and install maxent_ne_chunker, maxent_treebank_pos_tagger, punkt, qc, and words.

Obtain API keys for TMDB, Wolfram, and Yelp and insert them into config.py. Obtain a Facebook app ID and insert it into static/home.js.

PAL is now ready to run locally. Start it by running python server.py. PAL can be accessed by connecting to localhost:5000 in a browser.

Hill Climb Instructions

To ensure that queries are properly sorted into the proper services, hill climbing should be run. From the base directory, run

python -m pal.heuristics/hill_climb.hill_climb

After running hill climbing, updated heuristics values can be found in the pal/heuristics/hill_climb/climbed_values directory, and will be automatically referenced by the rest of PAL. Hill climbing should be re-run whenever a new service is added to PAL.

Pizza Service Installation

The pizza service requires Node.js or io.js. Install one of these and npm. Navigate to the api/dominos directory and run

npm install

Next, install forever with

npm install -g forever

Start the pizza server with

forever start server.js

Scraping the Directory

In order for the directory service to function, the directory must be pre-scraped. Navigate to the api/directory directory. Insert the username and password for the database in the stalkernet_scraper.py file with the format postgresql://username:password@database_url/table_name

Create the database schema by running

python models.py

Populate the buildings, majors, and departments tables by running

python stalkernet_scraper.py -b buildings.txt python stalkernet_scraper.py -m majors.txt python stalkernet_scraper.py -d depts.txt

Finally, scrape the directory with

python stalkernet_scraper.py -s

Installing in Apache

PAL can be run by Apache HTTP Server by using the mod_wsgi module. The Flask documentation has pretty detailed instructions on how to set this up. Please see http://flask.pocoo.org/docs/0.10/deploying/mod_wsgi/

Services

Bon Appetit (actively scraped from [Bon Appetit website] (http://carleton.cafebonappetit.com/cafe/))
- Single dining hall, single day, any or all meals
Dictionary/Thesaurus (actively scraped from [dictionary] (http://dictionary.reference.com/browse/) or [thesaurus] (http://www.thesaurus.com/browse/))
- Definitions
- Synonyms
- Antonyms
Directory (pre-scraped from [Carleton campus directory] (http://apps.carleton.edu/campus/directory/))
- Single-person queries
  - Professor offices / Student rooms
  - email addresses (both)
  - phone numbers (both)
  - departments/majors (professors/students)
  - Class year
- Multi-person queries
  - All students on a floor / in a dorm
  - All students for a given major
  - All professors in a department
  - "Show me all the Brians on campus"
Dominos (requests to [Dominos online order] (https://order.dominos.com/en/pages/order/))
- Pizza cost
  - price of a single pizza
  - price of multiple same pizzas
  - price of multiple different pizzas
- Pizza orders
  - order a single pizza
  - order multiple same pizzas
  - order multiple different pizzas
Facebook (requests to [Facebook Graph API] (https://developers.facebook.com/docs/graph-api))
- Post to timeline on behalf of user
Movies (using the [TMDB API] (https://www.themoviedb.org/documentation/api))
- What movies was this person involved in (acting, directing, etc.)
- Was this person involved in this movie?
- How many movies was this person involved in?
- When did this movie come out?
- Who acted/directed/etc. in this movie?
- Who played this character in this movie?
Translations (requests through [UltraLingua REST API] (http://api.ultralingua.com/ulapi/rest)
- From English to {Spanish, French, German, Italian, Portuguese}
- From {Spanish, French, German, Italian, Portuguese} to English
- Between {Spanish, French, German, Italian, Portuguese}
- Use correct SpeechSynthesisVoice for the destination language
Weather (requests to [Yahoo weather API] (https://query.yahooapis.com/v1/public/yql?))
- High/low temperatures
- Snow/rain
- General forecast
- Geolocation
Wolfram|Alpha (requests using [Wolfram|Alpha's API] (http://products.wolframalpha.com/api/))
- Run queries on natural language and get numerical output
- Keep track of our limited number of queries (difficult due to concurrency issues)
Yelp (requests to [Yelp API] (http://www.yelp.com/developers/documentation))
- Businesses by search terms
- Ratings, URL, Phone Number
- Find by location
- Businesses appear on map

Machyne/pal

PAL