Added missing classes from final year at OSU

OSU Coursework/CS 331 - Intro to Artificial Intelligence/Programming Assignment 3/reference/README.md

@@ -0,0 +1,23 @@
+# assignment3 - Sentiment Analysis with Naive Bayes
+- Rutger Farry & Simon Wu
+- CS331
+- Dr. Rebecca Hutchinson
+- 8 June 2017
+
+## Running the code
+The program looks for two files, named trainingSet.txt and testSet.txt, and outputs the results to stdout. If these files don't exist, the program will crash. Specs for these files are found here: It is built in Python 3, so running them should be simple on most computers:
+```bash
+python3 main.py
+```
+
+## About
+### Construction
+The program is split into a library and executable. All the computations and processing happen in `lib.py`, while the executable, `main.py` just calls the correct functions and provides a CLI.
+
+The code has been tested for compatibility down to Python 3.3.2. It may work with older versions, but they are not supported. :warning: The script will not work on Python 2.
+
+### Accuracy
+:chart: In our tests, the sentiment analyzer achieved **79.7% accuracy**. Some ideas for improving accuracy are: 
+1. Increasing our training data
+2. Spell-correcting mispelled words
+3. Trimming excess characters. For example `soooooooo good` should be considered the same as `soo good`.

OSU Coursework/CS 331 - Intro to Artificial Intelligence/Programming Assignment 3/reference/lib.py

@@ -0,0 +1,89 @@
+import re
+from math import log
+from collections import defaultdict
+
+non_alphanum = re.compile(r'[\W_]+')
+flatten = lambda l: [item for sublist in l for item in sublist]
+
+
+def process_word(word):
+    return non_alphanum.sub('', word.lower())
+
+
+def process_document(document):
+    """ Returns a tuple. The first element is the review split into array of words.
+        The second element is an int indicating whether the review was positive. """
+    return (
+        [process_word(word) for word in document.split()][:-1],
+        int(document.split()[-1])
+    )
+
+
+def format_document(document, vocab):
+    return (
+            ','.join(['1' if word in document[0] else '0' for word in vocab])
+            + ',' + str(document[1])
+    )
+
+
+def preprocess(infilename, outfilename):
+    """ Returns a tuple in the form of (vocab, documents).
+        Vocab: a set of words, e.g. { 'great', 'steak', 'gross', ... }
+        Documents: a tuple containing the array of words in the document followed
+        by the sentiment of the document, represented by a 1 or 0, e.g.
+        (['i', 'liked', 'the', 'food'], 1) """
+
+    # Process input file
+    infile = open(infilename, 'r')
+    raw_documents = infile.read().splitlines()
+    documents = [process_document(document) for document in raw_documents]
+    vocab = {process_word(word) for document in raw_documents for word in document.split()[:-1]}
+    vocab.discard('')
+    infile.close()
+
+    # Format text for outfile and write
+    outfile = open(outfilename, 'w')
+    formatted_vocab = ','.join(sorted(list(vocab)))
+    formatted_documents = '\n'.join([format_document(document, vocab) for document in documents])
+    outfile.write(formatted_vocab + '\n' + formatted_documents)
+    outfile.close()
+
+    return vocab, documents
+
+
+def train_naive_bayes(vocab, documents, classes):
+    logprior = {}
+    loglikelihood = defaultdict(dict)
+    big_doc = {}
+
+    def get_logprior(documents, c):
+        class_document_count = len([doc for doc in documents if doc[1] == c])
+        total_document_count = len(documents)
+        return log(class_document_count / total_document_count)
+
+    def get_bigdoc(documents, c):
+        return flatten([doc[0] for doc in documents if doc[1] == c])
+
+    for c in classes:
+        logprior[c] = get_logprior(documents, c)
+        big_doc[c] = get_bigdoc(documents, c)
+        words_in_class_count = sum([big_doc[c].count(w) for w in vocab])
+
+        for word in vocab:
+            word_count = big_doc[c].count(word)
+            loglikelihood[word][c] = log((word_count + 1) / (words_in_class_count + 1))
+
+    return logprior, loglikelihood
+
+
+def test_naive_bayes(document, logprior, loglikelihood, classes, vocab):
+    # Filter words not in the vocab
+    document = [word for word in document[0] if word in vocab]
+    prob = {}
+
+    for c in classes:
+        prob[c] = logprior[c]
+        for word in document:
+            prob[c] = prob[c] + loglikelihood[word][c]
+
+    return max(prob.keys(), key=(lambda key: prob[key]))

OSU Coursework/CS 331 - Intro to Artificial Intelligence/Programming Assignment 3/reference/main.py

@@ -0,0 +1,54 @@
+import sys
+from lib import preprocess, train_naive_bayes, test_naive_bayes
+
+def print_test_results(results, test_docs):
+    misclassified_docs = []
+    for result, doc in zip(results, test_docs):
+        if result != doc[1]:
+            misclassified_docs.append(doc + tuple([result]))
+
+    for doc in misclassified_docs:
+        print('misclassified {}: actual: {}, expected: {}'.format(doc[0], doc[2], doc[1]))
+
+    correct = len(test_docs) - len(misclassified_docs)
+    incorrect = len(misclassified_docs)
+    accuracy = correct / len(test_docs)
+
+    print("\n")
+    print("****************************************")
+    print("* RESULTS:")
+    print("*     Correct:   {}".format(correct))
+    print("*     Incorrect: {}".format(incorrect))
+    print("*")
+    print("*     Accuracy:  {}".format(accuracy))
+    print("****************************************")
+
+def test(test_docs, prior, likelihood, classes, vocab):
+    results = []
+    for test_doc in test_docs:
+        results.append(test_naive_bayes(test_doc, prior, likelihood, classes, vocab))
+    print_test_results(results, test_documents)
+
+# Prevent running if imported as a module
+if __name__ == '__main__':
+    classes = [0, 1]
+    # perform test on the training data
+    sys.stdout = open('output.txt', 'wt')
+    vocab, documents = preprocess('trainingSet.txt', 'preprocessed_train.txt')
+    _, test_documents = preprocess('trainingSet.txt', 'preprocessed_train.txt')
+
+    prior, likelihood = train_naive_bayes(vocab, documents, classes)
+
+    print("\n")
+    print("testing on training data")
+    test(documents, prior, likelihood, classes, vocab)
+
+    # perform test on the testing data
+    vocab, documents = preprocess('trainingSet.txt', 'preprocessed_train.txt')
+    _, test_documents = preprocess('testSet.txt', 'preprocessed_test.txt')
+
+    prior, likelihood = train_naive_bayes(vocab, documents, classes)
+    print("\n")
+    print("testing on testing data")
+    test(test_documents, prior, likelihood, classes, vocab)
+