add ctc beam search decoder

ctc_beam_search_decoder/ctc_beam_search.py

@@ -0,0 +1,152 @@
+## This is a prototype of ctc beam search decoder
+
+import copy
+import random
+import numpy as np
+
+# vocab = English characters + blank + space
+#vocab = ['-', ' '] + [chr(i) for i in range(97, 123)]
+
+vocab = [chr(97), chr(98), chr(99), chr(100)]+[' ', '-']
+
+def ids_str2list(ids_str):
+    ids_str = ids_str.split(' ')
+    ids_list = [int(elem) for elem in ids_str]
+    return ids_list
+
+def ids_list2str(ids_list):
+    ids_str = [str(elem) for elem in ids_list]
+    ids_str = ' '.join(ids_str)
+    return ids_str
+
+def ctc_beam_search_decoder(
+    input_probs_matrix,
+    beam_size,
+    lang_model=None,
+    name=None,
+    alpha=1.0,
+    beta=1.0,
+    blank_id=0,
+    space_id=1,
+    num_results_per_sample=None):
+
+    '''
+    beam search decoder for CTC-trained network, called outside of the recurrent group. 
+    adapted from Algorithm 1 in https://arxiv.org/abs/1408.2873.
+    '''
+    if num_results_per_sample is None:
+        num_results_per_sample = beam_size
+    assert num_results_per_sample <= beam_size
+
+    max_time_steps = len(input_probs_matrix)
+    assert  max_time_steps > 0
+
+    vocab_dim = len(input_probs_matrix[0])
+    assert blank_id < vocab_dim
+    assert space_id < vocab_dim
+
+    # initialize 
+    start_id = -1
+    prefix_set_prev = {str(start_id):1.0}
+    probs_b, probs_nb = {str(start_id):1.0}, {str(start_id):0.0}
+
+    # extend prefix in loop 
+    for time_step in range(max_time_steps):
+        print "\ntime_step = %d" % (time_step+1)
+        prefix_set_next = {}
+        probs_b_cur, probs_nb_cur = {}, {}
+        for (l, prob_) in prefix_set_prev.items():
+            print "l = %s\t%f" % (l, prob_)
+            prob = input_probs_matrix[time_step]
+
+            # convert ids in string to list
+            ids_list = ids_str2list(l)
+            end_id = ids_list[-1]
+            if not probs_b_cur.has_key(l):
+                probs_b_cur[l], probs_nb_cur[l] = 0.0, 0.0
+
+            # extend prefix
+            for c in range(0, vocab_dim):
+                if c == blank_id:
+	            probs_b_cur[l] += prob[c] * (probs_b[l] + probs_nb[l])
+		else:
+	            l_plus = l + ' ' + str(c)
+                    if not probs_b_cur.has_key(l_plus):
+                        probs_b_cur[l_plus], probs_nb_cur[l_plus] = 0.0, 0.0
+
+		    if c == end_id:
+		        probs_nb_cur[l_plus] += prob[c] * probs_b[l]
+			probs_nb_cur[l] += prob[c] * probs_nb[l]
+                    elif c == space_id:
+                        lm = 1 if lang_model is None else np.power(lang_model(ids_list), alpha)
+                        probs_nb_cur[l_plus] +=  lm * prob[c] * (probs_b[l]+probs_nb[l])
+                    else:
+                        probs_nb_cur[l_plus] += prob[c] * (probs_b[l]+probs_nb[l])
+                    prefix_set_next[l_plus] = probs_nb_cur[l_plus]+probs_b_cur[l_plus]
+
+                    print "l_plus: %s\t b=%f\tnb=%f\tP=%f" % (l_plus, probs_b_cur[l_plus], probs_nb_cur[l_plus], prefix_set_next[l_plus])
+            prefix_set_next[l] = probs_b_cur[l]+probs_nb_cur[l]
+            print "l: %s\t b=%f\tnb=%f\tP=%f" % (l, probs_b_cur[l], probs_nb_cur[l], prefix_set_next[l])
+        probs_b, probs_nb = copy.deepcopy(probs_b_cur), copy.deepcopy(probs_nb_cur)
+
+	prefix_set_prev = sorted(prefix_set_next.iteritems(), key = lambda asd:asd[1], reverse=True)
+
+	if beam_size < len(prefix_set_prev):
+	    prefix_set_prev = prefix_set_prev[:beam_size]
+	prefix_set_prev = dict(prefix_set_prev)
+
+    beam_result = []
+    for (seq, prob) in prefix_set_prev.items():
+        if prob > 0.0:
+            ids_list = ids_str2list(seq)
+            log_prob = np.log(prob)
+	    beam_result.append([log_prob, ids_list[1:]])
+
+    beam_result = sorted(beam_result, key = lambda asd:asd[0], reverse=True)
+    if num_results_per_sample < beam_size:
+        beam_result = beam_result[:num_results_per_sample]
+    return beam_result
+
+def language_model(input):
+    # TODO
+    return random.uniform(0, 1)
+
+def ctc_net(input, size_vocab):
+    size = len(vocab)
+   # prob =  np.array([random.uniform(0, 1) for i in range(0, size)])
+    prob =  np.array([1.0 for i in range(0, size)])
+    prob = prob/prob.sum()
+    return prob
+
+def main():
+
+    input_probs_matrix = [[0.1, 0.3, 0.6],
+                          [0.2, 0.1, 0.7],
+                          [0.5, 0.2, 0.3]]
+
+    prob_matrix = [[0.30999, 0.309938, 0.0679938, 0.0673362, 0.0708352, 0.173908],
+            [0.215136, 0.439699, 0.0370931, 0.0393967, 0.0381581, 0.230517],
+            [0.199959, 0.489485, 0.0233221, 0.0251417, 0.0233289, 0.238763],
+            [0.279611, 0.452966, 0.0204795, 0.0209126, 0.0194803, 0.20655],
+            [0.51286, 0.288951, 0.0243026, 0.0220788, 0.0219297, 0.129878],
+            # Random entry added in at time=5
+            #[0.155251, 0.164444, 0.173517, 0.176138, 0.169979, 0.160671]
+            ]
+
+    beam_result = ctc_beam_search_decoder(
+			input_probs_matrix=prob_matrix,
+			beam_size=2,
+                        blank_id=5,
+			)
+    def ids2str(ids_list):
+        ids_str = ''
+        for ids in ids_list:
+	    ids_str += vocab[ids]
+        return ids_str
+
+    print "\nbeam search output:"
+    for result in beam_result:
+        print result[0], ids2str(result[1])
+
+if __name__ == '__main__':
+	main()