BUGFIX: Dental and Stop special keys don't match multichar sounds like tʃ

So 't' will be matched but not 'tʃ'

README.rst

@@ -31,7 +31,7 @@ What can you do with this library?
 - map an actual pronunciation to a dictionary pronunciation (can be used 
   to automatically find speech errors)::
   
-    pysle.pronunciationtools.findClosestPronunciation(isleDict, 'cat', ['kh', 'ae',]) 
+    pysle.pronunciationtools.findClosestPronunciation(isleDict, 'cat', ['k', 'æ',])
 
 - automatically syllabify a praat textgrid containing words and phones 
   (e.g. force-aligned text) -- requires my 
@@ -39,10 +39,23 @@ What can you do with this library?
   
     pysle.syllabifyTextgrid(isleDict, praatioTextgrid, "words", "phones")
 
+- search for words based on pronunciation::
 
+    e.g. Words that start with a sound, or have a sound word medially, or 
+    in stressed vowel position, etc.
+    
+    see /tests/dictionary_search.py
+    
 Major revisions
 ================
 
+Ver 1.4 (July 9, 2016)
+
+- added search functionality
+
+- ported code to use the new unicode IPA-based isledict
+  (the old one was ascii)
+
 Ver 1.3 (March 15, 2016)
 
 - added indicies for stressed vowels
@@ -64,12 +77,14 @@ Requirements
 ================
 
 - Before you use this library (before or after installing it) you will need
-  to download the ILSEX dictionary.  It can be downloaded here:
+  to download the ILSEX dictionary.  It can be downloaded here under the
+  section 'English' linked under the text 'English Pronlex'
+  (with a file name of ISLEdict.txt):
 
-  `ISLEX project page <http://www.isle.illinois.edu/sst/data/dict/>`_
+  `ISLEX project page <http://isle.illinois.edu/sst/data/g2ps/>`_
 
   `Direct link to the ISLEX file used in this project
-  <http://www.isle.illinois.edu/sst/data/dict/islex/islev2.txt>`_ (islev2.txt)
+  <http://isle.illinois.edu/sst/data/g2ps/English/ISLEdict.txt>`_ (ISLEdict.txt)
 
 - ``Python 2.7.*`` or above
 
@@ -103,7 +118,7 @@ Here is a typical common usage::
     from pysle import isle
     isleDict = isle.LexicalTool('C:\islev2.dict')
     print isleDict.lookup('catatonic')[0] # Get the first pronunciation
-    >> [['kh', '@,'], ['t_(', '&'], ['th', "A'"], ['n', 'I', 'kh']] [2]
+    >> [['k', 'ˌæ'], ['t˺', 'ə'], ['t', 'ˈɑ'], ['n', 'ɪ', 'k']] [2, 0]
 
 and another::
 
@@ -111,7 +126,7 @@ and another::
     from psyle import pronunciationTools
     
     searchWord = 'another'
-    anotherPhoneList = ['n', '@', 'th', 'r'] # Actually produced
+    anotherPhoneList = ['n', '@', 'th', 'r'] # Actually produced (ASCII or IPA ok here)
 
     returnList = pronunciationTools.findBestSyllabification(isleDict, 
                                                             searchWord, 
@@ -128,7 +143,7 @@ Citing pysle
 
 Pysle is general purpose coding and doesn't need to be cited
 (you should cite the
-`ISLEX project <http://www.isle.illinois.edu/sst/data/dict/islex/index.shtml>`_
+`ISLEX project <http://isle.illinois.edu/sst/data/g2ps/>`_
 instead) but if you would like to, it can be cited like so:
 
 Tim Mahrt. Pysle. https://github.com/timmahrt/pysle, 2016.
@@ -139,7 +154,7 @@ Acknowledgements
 
 Development of Pysle was possible thanks to NSF grant **IIS 07-03624**
 to Jennifer Cole and Mark Hasegawa-Johnson, NSF grant **BCS 12-51343**
-to Jennifer Cole, José Hualde, and Caroline Smith, and
-to the A*MIDEX project (n° **ANR-11-IDEX-0001-02**) to James Sneed German
-funded by the Investissements d’Avenir French Government program, managed
+to Jennifer Cole, José Hualde, and Caroline Smith, and
+to the A*MIDEX project (n° **ANR-11-IDEX-0001-02**) to James Sneed German
+funded by the Investissements d'Avenir French Government program, managed
 by the French National Research Agency (ANR).

pysle/isletool.py

@@ -1,11 +1,33 @@
+#encoding: utf-8
 '''
 Created on Oct 11, 2012
 
 @author: timmahrt
 '''
 
+import io
+import re
 
-vowelList = ['a', '@', 'e', 'i', 'o', 'u', '^', '&', '>', ]
+
+charList = [u'#', u'.', u'aʊ', u'b', u'd', u'dʒ', u'ei', u'f', u'g',
+            u'h', u'i', u'j', u'k', u'l', u'm', u'n', u'oʊ', u'p',
+            u'r', u's', u't', u'tʃ', u'u', u'v', u'w', u'z', u'æ',
+            u'ð', u'ŋ', u'ɑ', u'ɑɪ', u'ɔ', u'ɔi', u'ə', u'ɚ', u'ɛ', u'ɝ',
+            u'ɪ', u'ɵ', u'ɹ', u'ʃ', u'ʊ', u'ʒ', u'æ', u'ʌ', ]
+
+diacriticList = [u'˺', u'ˌ', u'̩', u'̃', ]
+
+vowelList = [u'aʊ', u'ei', u'i', u'oʊ', u'u', u'æ',
+             u'ɑ', u'ɑɪ', u'ɔ', u'ɔi', u'ə', u'ɚ', u'ɛ', u'ɝ',
+             u'ɪ', u'ʊ', u'ʌ', ]
+
+
+def isVowel(char):
+    return any([vowel in char for vowel in vowelList])
+
+
+def sequenceMatch(matchChar, searchStr):
+    return matchChar in searchStr
 
 
 class WordNotInISLE(Exception):
@@ -30,7 +52,8 @@ class LexicalTool():
         Builds the isle textfile into a dictionary for fast searching
         '''
         lexDict = {}
-        wordList = [line.rstrip('\n') for line in open(self.islePath, "rU")]
+        with io.open(self.islePath, "r", encoding='utf-8') as fd:
+            wordList = [line.rstrip('\n') for line in fd]
             
         for row in wordList:
             word, pronunciation = row.split(" ", 1)
@@ -60,6 +83,214 @@ class LexicalTool():
         
         return pronList
 
+    def search(self, matchStr, numSyllables=None, wordInitial='ok',
+               wordFinal='ok', spanSyllable='ok', stressedSyllable='ok',
+               multiword='ok'):
+        return search(self.data.items(), matchStr, numSyllables=numSyllables,
+                      wordInitial=wordInitial, wordFinal=wordFinal,
+                      spanSyllable=spanSyllable,
+                      stressedSyllable=stressedSyllable,
+                      multiword=multiword)
+
+
+def _prepRESearchStr(matchStr, wordInitial='ok', wordFinal='ok',
+                     spanSyllable='ok', stressedSyllable='ok'):
+    '''
+    Prepares a user's RE string for a search
+    '''
+    
+    # Protect sounds that are two characters
+    # After this we can assume that each character represents a sound
+    # (We'll revert back when we're done processing the RE)
+    replList = [(u'ei', u'9'), (u'tʃ', u'='), (u'oʊ', u'~'),
+                (u'dʒ', u'@'), (u'aʊ', u'%'), (u'ɑɪ', u'&'),
+                (u'ɔi', u'$')]
+
+    # Add to the replList
+    currentReplNum = 0
+    startI = 0
+    for left, right in (('(', ')'), ('[', ']')):
+        while True:
+            try:
+                i = matchStr.index(left, startI)
+            except ValueError:
+                break
+            j = matchStr.index(right, i) + 1
+            replList.append((matchStr[i:j], str(currentReplNum)))
+            currentReplNum += 1
+            startI = j
+    
+    for charA, charB in replList:
+        matchStr = matchStr.replace(charA, charB)
+    
+    # Characters to check between all other characters
+    # Don't check between all other characters if the character is already
+    # in the search string or
+    interleaveStr = None
+    stressOpt = (stressedSyllable == 'ok' or stressedSyllable == 'only')
+    spanOpt = (spanSyllable == 'ok' or spanSyllable == 'only')
+    if stressOpt and spanOpt:
+        interleaveStr = u"\.?ˈ?"
+    elif stressOpt:
+        interleaveStr = u"ˈ?"
+    elif spanOpt:
+        interleaveStr = u"\.?"
+    
+    if interleaveStr is not None:
+        matchStr = interleaveStr.join(matchStr)
+    
+    # Setting search boundaries
+    # We search on '[^\.#]' and not '.' so that the search doesn't span
+    # multiple syllables or words
+    if wordInitial == 'only':
+        matchStr = u'#' + matchStr
+    elif wordInitial == 'no':
+        # Match the closest preceeding syllable.  If there is none, look
+        # for word boundary plus at least one other character
+        matchStr = u'(?:\.[^\.#]*?|#[^\.#]+?)' + matchStr
+    else:
+        matchStr = u'[#\.][^\.#]*?' + matchStr
+    
+    if wordFinal == 'only':
+        matchStr = matchStr + u'#'
+    elif wordFinal == 'no':
+        matchStr = matchStr + u"(?:[^\.#]*?\.|[^\.#]+?#)"
+    else:
+        matchStr = matchStr + u'[^\.#]*?[#\.]'
+    
+    # For sounds that are designated two characters, prevent
+    # detecting those sounds if the user wanted a sound
+    # designated by one of the contained characters
+    
+    # Forward search ('a' and not 'ab')
+    insertList = []
+    for charA, charB in [(u'e', u'i'), (u't', u'ʃ'), (u'd', u'ʒ'),
+                         (u'o', u'ʊ'), (u'a', u'ʊ|ɪ'), (u'ɔ', u'i'), ]:
+        startI = 0
+        while True:
+            try:
+                i = matchStr.index(charA, startI)
+            except ValueError:
+                break
+            if matchStr[i + 1] != charB:
+                forwardStr = u'(?!%s)' % charB
+#                 matchStr = matchStr[:i + 1] + forwardStr + matchStr[i + 1:]
+                startI = i + 1 + len(forwardStr)
+                insertList.append((i + 1, forwardStr))
+        
+    # Backward search ('b' and not 'ab')
+    for charA, charB in [(u't', u'ʃ'), (u'd', u'ʒ'),
+                         (u'a|o', u'ʊ'), (u'e|ɔ', u'i'), (u'ɑ' u'ɪ'), ]:
+        startI = 0
+        while True:
+            try:
+                i = matchStr.index(charB, startI)
+            except ValueError:
+                break
+            if matchStr[i - 1] != charA:
+                backStr = u'(?<!%s)' % charA
+#                 matchStr = matchStr[:i] + backStr + matchStr[i:]
+                startI = i + 1 + len(backStr)
+                insertList.append((i, backStr))
+                
+    insertList.sort()
+    for i, insertStr in insertList[::-1]:
+        matchStr = matchStr[:i] + insertStr + matchStr[i:]
+    
+    # Revert the special sounds back from 1 character to 2 characters
+    for charA, charB in replList:
+        matchStr = matchStr.replace(charB, charA)
+
+    # Replace special characters
+    replDict = {"D": u"(?:t(?!ʃ)|d(?!ʒ)|[sz])",  # dentals
+                "F": u"[ʃʒfvszɵðh]",  # fricatives
+                "S": u"(?:t(?!ʃ)|d(?!ʒ)|[pbkg])",  # stops
+                "N": u"[nmŋ]",  # nasals
+                "R": u"[rɝɚ]",  # rhotics
+                "V": u"(?:aʊ|ei|oʊ|ɑɪ|ɔi|[iuæɑɔəɛɪʊʌ]):?",  # vowels
+                "B": u"\.",  # syllable boundary
+                }
+
+    for char, replStr in replDict.items():
+        matchStr = matchStr.replace(char, replStr)
+
+    return matchStr
+
+
+def search(searchList, matchStr, numSyllables=None, wordInitial='ok',
+           wordFinal='ok', spanSyllable='ok', stressedSyllable='ok',
+           multiword='ok'):
+    '''
+    Searches for matching words in the dictionary with regular expressions
+    
+    wordInitial, wordFinal, spanSyllable, stressSyllable, and multiword
+    can take three different values: 'ok', 'only', or 'no'.
+    
+    Special search characters:
+    'D' - any dental; 'F' - any fricative; 'S' - any stop
+    'V' - any vowel; 'N' - any nasal; 'R' - any rhotic
+    '#' - word boundary
+    'B' - syllable boundary
+    '.' - anything
+    
+    For advanced queries:
+    Regular expression syntax applies, so if you wanted to search for any
+    word ending with a vowel or rhotic, matchStr = '(?:VR)#', '[VR]#', etc.
+    '''
+    # Run search for words
+    
+    matchStr = _prepRESearchStr(matchStr, wordInitial, wordFinal,
+                                spanSyllable, stressedSyllable)
+    
+    compiledRE = re.compile(matchStr)
+    retList = []
+    for word, pronList in searchList:
+        newPronList = []
+        for pron in pronList:
+            searchPron = pron.replace(",", "").replace(" ", "")
+            
+            # Ignore diacritics for now:
+            for diacritic in diacriticList:
+                if diacritic not in matchStr:
+                    searchPron = searchPron.replace(diacritic, "")
+            
+            if numSyllables is not None:
+                if numSyllables != searchPron.count('.') + 1:
+                    continue
+            
+            # Is this a compound word?
+            if multiword == 'only':
+                if searchPron.count('#') == 2:
+                    continue
+            elif multiword == 'no':
+                if searchPron.count('#') > 2:
+                    continue
+            
+            matchList = compiledRE.findall(searchPron)
+            if len(matchList) > 0:
+                if stressedSyllable == 'only':
+                    if all([u"ˈ" not in match for match in matchList]):
+                        continue
+                if stressedSyllable == 'no':
+                    if all([u"ˈ" in match for match in matchList]):
+                        continue
+                
+                # For syllable spanning, we check if there is a syllable
+                # marker inside (not at the border) of the match.
+                if spanSyllable == 'only':
+                    if all(["." not in txt[1:-1] for txt in matchList]):
+                        continue
+                if spanSyllable == 'no':
+                    if all(["." in txt[1:-1] for txt in matchList]):
+                        continue
+                newPronList.append(pron)
+        
+        if len(newPronList) > 0:
+            retList.append((word, newPronList))
+    
+    retList.sort()
+    return retList
+
 
 def _parsePronunciation(pronunciationStr):
     '''
@@ -76,11 +307,11 @@ def _parsePronunciation(pronunciationStr):
     stressedPhoneList = []
     for i, syllable in enumerate(syllableList):
         for j, phone in enumerate(syllable):
-            if "'" in phone:
+            if u"ˈ" in phone:
                 stressedSyllableList.insert(0, i)
                 stressedPhoneList.insert(0, j)
                 break
-            elif '"' in phone:
+            elif u'ˌ' in phone:
                 stressedSyllableList.insert(i)
                 stressedPhoneList.insert(j)
     

pysle/praattools.py

@@ -1,3 +1,4 @@
+#encoding: utf-8
 '''
 Created on Oct 22, 2014
 
@@ -76,7 +77,7 @@ def syllabifyTextgrid(isleDict, tg, wordTierName, phoneTierName,
             stressJ = None  #
             
         if stressI is not None:
-            syllableList[stressI][stressJ] += "'"
+            syllableList[stressI][stressJ] += u"ˈ"
 
         i = 0
 #         print(syllableList)

pysle/pronunciationtools.py

@@ -1,3 +1,4 @@
+#encoding: utf-8
 '''
 Created on Oct 15, 2014
 
@@ -151,7 +152,7 @@ def _findBestPronunciation(isleDict, wordText, aPron):
         hasStress = False
         for syllable in syllableList:
             for phone in syllable:
-                hasStress = "'" in phone or hasStress
+                hasStress = u"ˈ" in phone or hasStress
         
         if hasStress:
             withStress.append(i)

setup.py

@@ -1,16 +1,19 @@
+#!/usr/bin/env python
+# encoding: utf-8
 '''
 Created on Oct 15, 2014
 
 @author: tmahrt
 '''
+import codecs
 from distutils.core import setup
 setup(name='pysle',
-      version='1.3.0',
+      version='1.4.0',
       author='Tim Mahrt',
       author_email='timmahrt@gmail.com',
       package_dir={'pysle':'pysle'},
       packages=['pysle'],
       license='LICENSE',
-      long_description=open('README.rst', 'r').read(),
+      long_description=codecs.open('README.rst', 'r', encoding="utf-8").read(),
 #       install_requires=[], # No requirements! # requires 'from setuptools import setup'
       )

test/basic_examples.py

@@ -1,3 +1,4 @@
+#encoding: utf-8
 '''
 Created on Oct 22, 2014
 
@@ -12,16 +13,18 @@ from pysle import pronunciationtools
 # In this first example we look up the syllabification of a word and get it's 
 # stress information.
 
-searchWord = 'pumpkins'
-isleDict = isletool.LexicalTool('islev2.txt')
+searchWord = 'catatonic'
+isleDict = isletool.LexicalTool('ISLEdict.txt')
 lookupResults = isleDict.lookup(searchWord)
 
 firstEntry = lookupResults[0]
 firstSyllableList = firstEntry[0] 
+firstSyllableList = ".".join([u" ".join(syllable) for syllable in firstSyllableList])
 firstStressList = firstEntry[1]
 
 print(searchWord)
-print(firstSyllableList, firstStressList) # 3rd syllable carries stress
+print(firstSyllableList)
+print(firstStressList) # 3rd syllable carries stress
 
 
 # Here we determine the syllabification of a word, as it was said.
@@ -35,10 +38,14 @@ returnList = pronunciationtools.findBestSyllabification(isleDict,
                                                         searchWord, 
                                                         anotherPhoneList)
 
-stressedSyllable, syllableList, syllabification, stressedIndex = returnList
-
+(stressedSyllable, syllableList, syllabification,
+stressedSyllableIndexList, stressedPhoneIndexList,
+flattenedStressIndexList) = returnList
 print(searchWord)
 print(anotherPhoneList)
-print(syllableList) # We can see the first syllable was elided
-
+print(stressedSyllableIndexList) # We can see the first syllable was elided
+print(stressedPhoneIndexList)
+print(flattenedStressIndexList)
+print(syllableList)
+print(syllabification)
 

test/dictionary_search.py

@@ -0,0 +1,55 @@
+#encoding: utf-8
+'''
+Created on July 08, 2016
+
+@author: tmahrt
+
+Basic examples of common usage.
+'''
+
+import random
+
+from pysle import isletool
+
+tmpPath = r"C:\Users\Tim\Dropbox\workspace\pysle\test\ISLEdict.txt"
+isleDict = isletool.LexicalTool(tmpPath)
+
+def printOutMatches(matchStr, numSyllables=None, wordInitial='ok',
+                    wordFinal='ok', spanSyllable='ok', stressedSyllable='ok',
+                    multiword='ok', numMatches=None, matchList=None):
+
+    if matchList is None:
+        matchList = isleDict.search(matchStr, numSyllables, wordInitial,
+                                    wordFinal, spanSyllable, stressedSyllable,
+                                    multiword)
+    else:
+        matchList = isletool.search(matchList, matchStr, numSyllables, wordInitial,
+                                    wordFinal, spanSyllable, stressedSyllable,
+                                    multiword)
+    
+    if numMatches is not None and len(matchList) > numMatches:
+        random.shuffle(matchList)
+        
+    for i, matchTuple in enumerate(matchList):
+        if numMatches is not None and i > numMatches:
+            break
+        word, pronList = matchTuple
+        print("%s: %s" % (word, ",".join(pronList)))
+    print("")
+    
+    return matchList
+
+# 2-syllable words with a stressed syllable containing 'dV' but not word initially
+printOutMatches("dV", stressedSyllable="only", spanSyllable="no",
+                wordInitial="no", numSyllables=2, numMatches=10)
+ 
+# 3-syllable word with an 'ld' sequence that spans a syllable boundary
+printOutMatches("lBd", wordInitial="no", multiword='no',
+                numSyllables=3, numMatches=10)
+ 
+# words ending in 'inth'
+matchList = printOutMatches(u"ɪnɵ", wordFinal="only", numMatches=10)
+ 
+# that also start with 's'
+matchList = printOutMatches("s", wordInitial="only", numMatches=10,
+                            matchList=matchList, multiword="no")