Source Code for Module jazzparser.taggers.segmidi.chordclass.tagutils

  1  """Misc utilities used by the chordclass tagger. 
  2   
  3  """ 
  4  """ 
  5  ============================== License ======================================== 
  6   Copyright (C) 2008, 2010-12 University of Edinburgh, Mark Granroth-Wilding 
  7    
  8   This file is part of The Jazz Parser. 
  9    
 10   The Jazz Parser is free software: you can redistribute it and/or modify 
 11   it under the terms of the GNU General Public License as published by 
 12   the Free Software Foundation, either version 3 of the License, or 
 13   (at your option) any later version. 
 14    
 15   The Jazz Parser is distributed in the hope that it will be useful, 
 16   but WITHOUT ANY WARRANTY; without even the implied warranty of 
 17   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the 
 18   GNU General Public License for more details. 
 19    
 20   You should have received a copy of the GNU General Public License 
 21   along with The Jazz Parser.  If not, see <http://www.gnu.org/licenses/>. 
 22   
 23  ============================ End license ====================================== 
 24   
 25  """ 
 26  __author__ = "Mark Granroth-Wilding <mark.granroth-wilding@ed.ac.uk>"  
 27   
 28 -class CategoryProcessChart(object): 
 29          """ 
 30          CKY chart for the mini-parsing task we need to do to prepare the categories. 
 31          """ 
 32 -        def __init__(self, length): 
 33                  self.length = length 
 34                  # Prepare the chart representation 
 35                  chart = [] 
 36                  for i in range(length): 
 37                          row = [] 
 38                          for j in range(length): 
 39                                  row.append([]) 
 40                          chart.append(row) 
 41                  self._chart = chart 
 42                   
 43 -        def get_cell(self, start, end): 
 44                  return self._chart[start][end-1] 
 45                   
 46 -        def add_span(self, start, end, val): 
 47                  self._chart[start][end-1].append(val) 
 48           
 49 -        def __str__(self): 
 50                  # Just for debugging, doesn't need to look good 
 51                  return str(self._chart) 
 52   
 53 -def prepare_categories(timesteps): 
 54          """ 
 55          Processes the output from decoding the model in the form of a set of  
 56          top tags for each timestep. We need to find each possible sequence of  
 57          identical tags and combine then into a single span so that the  
 58          self-transition gets interpreted as a continuation of the category.  
 59          We also want to keep the smaller spans, right down to the single-timestep  
 60          categories, so that they can be considered by the parser in combination  
 61          with other spans that would otherwise overlap. 
 62           
 63          We do this using a little CKY chart parser. 
 64           
 65          """ 
 66          T = len(timesteps) 
 67          chart = CategoryProcessChart(T) 
 68          # Label the tags with their priority 
 69          timesteps = [list(enumerate(timestep)) for timestep in timesteps] 
 70           
 71          # Initialize the chart 
 72          for (time,tags) in enumerate(timesteps): 
 73                  for tag in tags: 
 74                          chart.add_span(time, time+1, tag) 
 75           
 76          # CKY loops 
 77          for end in range(2, T+1): 
 78                  for start in range(0, end-1): 
 79                          for middle in range(start+1, end): 
 80                                  # Look for identical tags between these two cells 
 81                                  first = chart.get_cell(start, middle) 
 82                                  second = chart.get_cell(middle, end) 
 83                                  for priority1,(prob1,tag1) in first: 
 84                                          for priority2,(prob2,tag2) in second: 
 85                                                  if tag1 == tag2: 
 86                                                          # Found a repeated tag: add it as a span 
 87                                                          # Take the product of the probabilities as the  
 88                                                          #  probability of the combined tag 
 89                                                          chart.add_span(start, end, \ 
 90                                                                  (max(priority1, priority2), (prob1+prob2, tag1))) 
 91           
 92          # Get all the spans that result from the parse 
 93          prioritized_spans = {} 
 94          for start in range(0, T): 
 95                  for end in range(start+1, T+1): 
 96                          for (priority,(prob,tag)) in chart.get_cell(start,end): 
 97                                  prioritized_spans.setdefault(priority, []).append((start, end, (prob,tag))) 
 98          # Return them grouped by priority and in descending order of priority 
 99          grouped_spans = [spans for (priority,spans) in sorted(prioritized_spans.items())] 
100                   
101          return grouped_spans 
102