diff --git a/README.rst b/README.md similarity index 73% rename from README.rst rename to README.md index 4a69007..0f79532 100644 --- a/README.rst +++ b/README.md @@ -1,47 +1,34 @@ -Overview -======== - -A sample program I wrote to detect gibberish. It uses a 2 character markov chain. - -http://en.wikipedia.org/wiki/Markov_chain - -This is a nice (IMO) answer to this guys question on stackoverflow. -http://stackoverflow.com/questions/6297991/is-there-any-way-to-detect-strings-like-putjbtghguhjjjanika/6298040#comment-7360747 - -Usage -===== - -First train the model: - -python gib_detect_train.py - -Then try it on some sample input - -python gib_detect.py - -my name is rob and i like to hack True - -is this thing working? True - -i hope so True - -t2 chhsdfitoixcv False - -ytjkacvzw False - -yutthasxcvqer False - -seems okay True - -yay! True - -How it works -============ -The markov chain first 'trains' or 'studies' a few MB of English text, recording how often characters appear next to each other. Eg, given the text "Rob likes hacking" it sees Ro, ob, o[space], [space]l, ... It just counts these pairs. After it has finished reading through the training data, it normalizes the counts. Then each character has a probability distribution of 27 followup character (26 letters + space) following the given initial. - -So then given a string, it measures the probability of generating that string according to the summary by just multiplying out the probabilities of the adjacent pairs of characters in that string. EG, for that "Rob likes hacking" string, it would compute prob['r']['o'] * prob['o']['b'] * prob['b'][' '] ... This probability then measures the amount of 'surprise' assigned to this string according the data the model observed when training. If there is funny business with the input string, it will pass through some pairs with very low counts in the training phase, and hence have low probability/high surprise. - -I then look at the amount of surprise per character for a few known good strings, and a few known bad strings, and pick a threshold between the most surprising good string and the least surprising bad string. Then I use that threshold whenever to classify any new piece of text. - -Peter Norvig, the director of Research at Google, has this nice talk about "The unreasonable effectiveness of data" here, http://www.youtube.com/watch?v=9vR8Vddf7-s. This insight is really not to try to do something complicated, just write a small program that utilizes a bunch of data and you can do cool things. - +# GIBBERISH-DETECTOR + +A program to check if a sentence is gibberish or not using simple markov chain + +## 1. Getting Started + +Clone the repo: + + ```bash + git clone https://github.com/rrenaud/Gibberish-Detector.git + ``` +Train the model: + +``` +python gib_detect_train.py +``` +Run the model: + +``` +python gib_detect.py +``` + +## 2. How it works + +The markov chain first 'trains' or 'studies' a few MB of English text, recording how often characters appear next to each other. Eg, given the text "Rob likes hacking" it sees Ro, ob, o[space], [space]l, ... It just counts these pairs. After it has finished reading through the training data, it normalizes the counts. Then each character has a probability distribution of 27 followup character (26 letters + space) following the given initial. + +So then given a string, it measures the probability of generating that string according to the summary by just multiplying out the probabilities of the adjacent pairs of characters in that string. EG, for that "Rob likes hacking" string, it would compute prob['r']['o'] * prob['o']['b'] * prob['b'][' '] ... This probability then measures the amount of 'surprise' assigned to this string according the data the model observed when training. If there is funny business with the input string, it will pass through some pairs with very low counts in the training phase, and hence have low probability/high surprise. + +I then look at the amount of surprise per character for a few known good strings, and a few known bad strings, and pick a threshold between the most surprising good string and the least surprising bad string. Then I use that threshold whenever to classify any new piece of text. + +Peter Norvig, the director of Research at Google, has this nice talk about "The unreasonable effectiveness of data" here, http://www.youtube.com/watch?v=9vR8Vddf7-s. This insight is really not to try to do something complicated, just write a small program that utilizes a bunch of data and you can do cool things. + +![sample](/sample.png) + diff --git a/gib_detect.py b/gib_detect.py index d59d911..5e235b5 100644 --- a/gib_detect.py +++ b/gib_detect.py @@ -2,11 +2,19 @@ import pickle import gib_detect_train - +print(" ") +print(" GIBBBERISH-DETECTOR") +print(" ") model_data = pickle.load(open('gib_model.pki', 'rb')) - -while True: - l = raw_input() +check = 1 +while check == 1: + l = str(input("Enter text: ")) model_mat = model_data['mat'] threshold = model_data['thresh'] - print gib_detect_train.avg_transition_prob(l, model_mat) > threshold + if (gib_detect_train.avg_transition_prob(l, model_mat) > threshold) == True: + print("The text is not gibberish") + else: + print("The text is gibberish") + print(" ") + check = int(input("Press 1 continue: ")) + print(" ") \ No newline at end of file diff --git a/gib_detect_train.py b/gib_detect_train.py index 11fa8f1..8fd3270 100644 --- a/gib_detect_train.py +++ b/gib_detect_train.py @@ -26,7 +26,7 @@ def train(): # prior or smoothing factor. This way, if we see a character transition # live that we've never observed in the past, we won't assume the entire # string has 0 probability. - counts = [[10 for i in xrange(k)] for i in xrange(k)] + counts = [[10 for i in range(k)] for i in range(k)] # Count transitions from big text file, taken # from http://norvig.com/spell-correct.html @@ -41,7 +41,7 @@ def train(): # http://squarecog.wordpress.com/2009/01/10/dealing-with-underflow-in-joint-probability-calculations/ for i, row in enumerate(counts): s = float(sum(row)) - for j in xrange(len(row)): + for j in range(len(row)): row[j] = math.log(row[j] / s) # Find the probability of generating a few arbitrarily choosen good and @@ -72,4 +72,4 @@ def avg_transition_prob(l, log_prob_mat): - + \ No newline at end of file diff --git a/gib_model.pki b/gib_model.pki new file mode 100644 index 0000000..cf3dc3d Binary files /dev/null and b/gib_model.pki differ diff --git a/sample.png b/sample.png new file mode 100644 index 0000000..caeb81b Binary files /dev/null and b/sample.png differ