refactoring.

jetzt kommt der umbau cleanedcoprus --> doctermmatrix --LDA & labaled_lines.txt --> LLDA

cleaning.py

@@ -30,6 +30,16 @@ with open(config_ini) as f:
 
 
 def clean(stringstream):#, NOUNS):
+    """
+    fix bad unicode
+    seperate_words_on_regex `\=~%^&*()_+\[\]{};\'"|</>
+    normalize whitespace
+    remove linebreaks
+    replaceRockDöts
+
+    :param stringstream: str-gen
+    :return: string-gen
+    """
 
     #NOUNS = [n.lower() for n in NOUNS]
 
@@ -90,19 +100,22 @@ corpus_de_path = FILEPATH + config.get("de_corpus", "path")
 def cleanCorpus(corpus):
     logprint("Clean {0}_corpus at {1}".format(corpus.lang, datetime.now()))
 
-
+    """
-
     ressources_path = FILEPATH + "ressources/"
+ 
+    
     path2nouns_list = ressources_path + config.get("nouns", "pickle_file")
+
+    
     #NOUNS = load_obj(path2nouns_list)
     #noun_disjunction = '|'.join(NOUNS)
     #nouns_tuples = []
     #for n in NOUNS:
     #    nouns_tuples.append((n.lower(),n))
-
+    """
-    cleanCorpus_name = corpus.lang + "_clean"
 
 
+    # load Corpus
     raw_corpus = corpus
     parser = corpus.spacy_lang
 
@@ -115,13 +128,14 @@ def cleanCorpus(corpus):
     )
 
 
-    # leere docs aus corpi kicken
+    # leere docs aus corpus kicken
     cleaned_corpus.remove(lambda doc: len(doc) == 0)
 
 
 
 
     #save corpus
+    cleanCorpus_name = corpus.lang + "_clean"
     save_corpus(corpus=cleaned_corpus, corpus_path=corpus_de_path, corpus_name=cleanCorpus_name)
 
 

corporization.py

@@ -90,7 +90,16 @@ corpus_de_path = FILEPATH + config.get("de_corpus", "path")
 
 
 def ticketcsv2Corpus(path2_csv, corpus_path, content_collumn_name, lang, printrandom=0):
+    """
+    Use textacy to create a Corpus out of the ITMC-Ticket.csv
 
+    :param path2_csv: str
+    :param corpus_path: str
+    :param content_collumn_name: str    the Collumn which is used as the Docs text
+    :param lang: str       standard 2-letter language
+    :param printrandom: print n random Documents
+    :return: textacy.Corpus
+    """
 
     # print paths
     path_csv_split = path2_csv.split("/")

init.py

@@ -28,20 +28,20 @@ with open(config_ini) as f:
 
 def create_lemma_dict(path2lemmalist):
     """
-    Creates a dict out of a file a la:
+    Creates a dict out of a txt file a la:
 
     l1 w1
     l1 w2
     l2 w1
     l2 w2
 
-    Result will be used as lemma_dict["word"] --> lemma
+    Result will be used as lemma_dict[word] --> lemma
 
     :param path2lemmalist: str
     :return: dictionary
     """
-    lemmalist = list(map(textacy.preprocess.normalize_whitespace, list(
+    file_gen = textacy.fileio.read_file_lines(path2lemmalist)
-        textacy.fileio.read_file_lines(path2lemmalist))))
+    lemmalist = list(map(textacy.preprocess.normalize_whitespace, list(file_gen)))
 
     lemma_dict = {}
 
@@ -63,7 +63,7 @@ def build_thesaurus_dict(path2wordnet,returnall=False):
     Creates a dict out of the deWordNet
     https://raw.githubusercontent.com/hdaSprachtechnologie/odenet/master/deWordNet.xml
 
-    Result will be used as lemma_dict["word"] --> lemma
+    Result will be used as thesaurus[word] --> main_synonym
 
     :param path2wordnet: str
     :param returnall: bool    if True, also return , word2synsets, synset2Words
@@ -73,6 +73,7 @@ def build_thesaurus_dict(path2wordnet,returnall=False):
 
     lexroot = lextree.getroot()
 
+    # Build word2synsets
     word2synsets = {}
     template = {"w1": ["s1", "s2"]}
 
@@ -82,7 +83,6 @@ def build_thesaurus_dict(path2wordnet,returnall=False):
                 lex_dictlist = [subentry.attrib for subentry in elem]
 
                 # idee technischer thesaurus
-                # idee hauptsynonmy muss einzelnes wort sein
 
                 synlist = []
                 string = "WORD"
@@ -96,55 +96,92 @@ def build_thesaurus_dict(path2wordnet,returnall=False):
                     if 'writtenForm' in lex_dict.keys():
                             string = (lex_dict["writtenForm"])
 
-                            if string == "Kennwort":
-                                pass
 
                             # replaceRockDots
                             string = re.sub(r'[ß]', "ss", string)
                             string = re.sub(r'[ö]', "oe", string)
+                            string = re.sub(r'[Ö]', "Oe", string)
+
                             string = re.sub(r'[ü]', "ue", string)
+                            string = re.sub(r'[Ü]', "Ue", string)
+
                             string = re.sub(r'[ä]', "ae", string)
+                            string = re.sub(r'[Ä]', "ae", string)
 
                             # alle punkte raus
                             string = re.sub(r'[.]', "", string)
 
                             # alles in klammern raus
-                            string = re.sub(r"\((.*)\)", " ", string)
+                            if "auptform" in string:
+                                string = re.sub(r"\((.*)\)", " ", string)
+                                string = string + " (hauptform)"            # evtl. als hauptform merken
+                            else:
+                                string = re.sub(r"\((.*)\)", " ", string)
 
                             # längeres leerzeichen normalisieren
                             string = textacy.preprocess.normalize_whitespace(string)
 
-                            string = string.lower().strip()
+                            string = string.strip()#.lower()
 
-                word2synsets[string] = synlist
+                if string != '':
+                    word2synsets[string] = synlist
 
 
+
+    # Build synset2Words
     synset2Words = {}
     template = {"s1": ["w1","w2"]}
 
     for word,synset in word2synsets.items():
         if word != '':
+
+
             for syn in synset:
                 if syn not in synset2Words.keys():
                     synset2Words[syn] = [word]
                 else:
                     synset2Words[syn].append(word)
 
-    # nach anzhal der wörter in den strings sortieren
+
-    for synset in word2synsets.values():
+    # Sortieren
-        synset.sort(key=lambda x: len(x.split()))
+    for words in synset2Words.values():
+        words.sort(key=lambda w: len(w.split())) # nach anzhal der wörter in den strings (weniger nach vorne)
+        for w in words:
+            if "(hauptform)" in w:
+                to_insert = re.sub(r"\((.*)\)", " ", w).strip()
+
+                words.remove(w)
+                words.insert(0, to_insert)  # Hauptform evtl. nach vorne
+
+
+
 
     thesaurus = {}
     thesaurus_template = {"w1" : "mainsyn"}
+    # word --> [synset1, synset2, .. ] --> synset1 --> [syn1, syn2, ... ] --> syn1 / mainsyn
 
 
-    for word,synset in word2synsets.items():
+    for word,synsets in word2synsets.items():   #word , [synset1, synset2, .. ]
         try:
-            thesaurus[word] = synset2Words[synset[0]][0]    #Ann.: erstes synonym ist das Hauptsynonym #todo nach (hauptform) suchen?
+            if "Passwort" in word:
+                x=2
+
+            first_synset = synsets[0]           #erstes synset wählen . praktischer Grund
+
+            syns = synset2Words[first_synset]   # [syn1, syn2, ... ]
+
+            first_syn = syns[0] # erstes synonym (evtl. Hauptform)  wählen
+
+            word = re.sub(r"\((.*)\)", " ", word).strip() #(hautpform weg)
+
+
+
+            thesaurus[word] = first_syn  #Ann.: erstes synonym ist das Hauptsynonym
         except:
             pass
 
 
+
     if returnall:
         return thesaurus, word2synsets, synset2Words
     else:
@@ -237,39 +274,8 @@ def build_words_for_spellchecking(path2words):
 
 ##################################################################################################
 
-# THESAURUS
-ressources_path = FILEPATH +  "ressources/"
-path2wordnet = ressources_path + config.get("thesaurus","input")
-path2thesaurus_dict = ressources_path + config.get("thesaurus","pickle_file")
 
 
-# SPELLCHECKING
-path2words_file = ressources_path + config.get("spellchecking","input")
-path2wordlist = ressources_path + config.get("spellchecking","pickle_file")
-
-
-# LEMMA
-path2lemma_file = ressources_path + config.get("lemmatization","input")
-path2lemmadict = ressources_path + config.get("lemmatization","pickle_file")
-
-# NOMEN
-nouns0  = ressources_path + config.get("nouns","input")
-nouns1 = ressources_path + config.get("nouns","input1")
-nouns2 = ressources_path + config.get("nouns","input2")
-path2nouns_list = ressources_path + config.get("nouns","pickle_file")
-
-
-# VORNAMEN
-firstnames_txt = ressources_path + config.get("firstnames","input")
-path2firstnameslist = ressources_path + config.get("firstnames","pickle_file")
-
-# STOPWORDS
-stop1 = ressources_path + config.get("de_stopwords","input1")
-stop2 = ressources_path + config.get("de_stopwords","input2")
-stop3 = ressources_path + config.get("de_stopwords","input3")
-path2stopwordlist_de = ressources_path + config.get("de_stopwords","pickle_file")
-
-path2stopwordlist_en = ressources_path + config.get("en_stopwords","pickle_file")
 
 
 
@@ -277,44 +283,135 @@ def main():
     start = time.time()
     logprint("Init: {0}".format(datetime.now()))
 
-
+    ressources_path = FILEPATH + "ressources/"
-    """"""
-    logprint("create and save lemma_dict")
-    lemma_dict = create_lemma_dict(path2lemma_file)
-    save_obj(lemma_dict, path2lemmadict)
-
-
-    logprint("Build and save Wordlist for Spellchecking")
-    words = build_words_for_spellchecking(path2words_file)
-    save_obj(words, path2wordlist)
 
 
 
+
+
+
+
+    # THESAURUS
     logprint("Build and save Thesaurus")
+
+    path2wordnet = ressources_path + config.get("thesaurus", "input")
     thesaurus = build_thesaurus_dict(path2wordnet)
+
+    path2thesaurus_dict = ressources_path + config.get("thesaurus", "pickle_file")
     save_obj(thesaurus, path2thesaurus_dict)
 
 
 
 
+
+
+
+
+
+
+    # LEMMA
+    logprint("create and save lemma_dict")
+
+    path2lemma_file = ressources_path + config.get("lemmatization", "input")
+    lemma_dict = create_lemma_dict(path2lemma_file)
+
+    path2lemmadict = ressources_path + config.get("lemmatization", "pickle_file")
+    save_obj(lemma_dict, path2lemmadict)
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+    # SPELLCHECKING
+    logprint("Build and save Wordlist for Spellchecking")
+
+    path2words_file = ressources_path + config.get("spellchecking", "input")
+    words = build_words_for_spellchecking(path2words_file)
+
+    path2words_counter = ressources_path + config.get("spellchecking", "pickle_file")
+    save_obj(words, path2words_counter)
+
+
+
+
+
+
+
+
+
+
+
+    # STOPWORDS
     logprint("Build and save stoppwortliste")
+
+    stop1 = ressources_path + config.get("de_stopwords", "input1")
+    stop2 = ressources_path + config.get("de_stopwords", "input2")
+    stop3 = ressources_path + config.get("de_stopwords", "input3")
     de_stop_words, en_stop_words = create_stopword_lists(stop1, stop2, stop3)
+
+    path2stopwordlist_de = ressources_path + config.get("de_stopwords", "pickle_file")
     save_obj(de_stop_words, path2stopwordlist_de)
+
+    path2stopwordlist_en = ressources_path + config.get("en_stopwords", "pickle_file")
     save_obj(en_stop_words, path2stopwordlist_en)
 
 
 
+
+
+
+
+
+
+
+    # NOMEN
     logprint("Build and save nomenliste")
-    #nouns = list_from_files(nouns1,nouns2)
+
-    nouns = list_from_files(nouns0)
+    nouns0 = ressources_path + config.get("nouns", "input")
+    nouns1 = ressources_path + config.get("nouns", "input1")
+    nouns2 = ressources_path + config.get("nouns", "input2")
+    nouns = list_from_files(nouns0,nouns1,nouns2)
+
+    path2nouns_list = ressources_path + config.get("nouns", "pickle_file")
     save_obj(nouns, path2nouns_list)
 
 
+
+
+
+
+
+
+
+    # VORNAMEN
     logprint("Build and save firstnameslist")
+
+    firstnames_txt = ressources_path + config.get("firstnames", "input")
     vornamen = list_from_files(firstnames_txt)
+
+    path2firstnameslist = ressources_path + config.get("firstnames", "pickle_file")
     save_obj(vornamen, path2firstnameslist)
 
 
+
+
+
+
+
+
+
+
+
+
     end = time.time()
     logprint("Time Elapsed Initialization:{0} min".format((end - start) / 60))
 

main.py

@@ -30,15 +30,13 @@ start = time.time()
 
 # todo modelle testen
 
-
+# todo ticket2kbkeys, subj, cats in init.py
-
 
 
 logprint("main.py started at {}".format(datetime.now()))
 
 
-
+init.main()
-#init.main()
 logprint("")
 
 raw_corpus = corporization.main()

miscellaneous.py

@@ -217,7 +217,6 @@ def save_corpus(corpus, corpus_path, corpus_name):
     :param corpus_path: str
     :param corpus_name: str (should content the language like "_de_")
     """
-    #todo pos und ner tagging speichern
 
     # save parser
     parser = corpus.spacy_lang

preprocessing.py

@@ -126,7 +126,7 @@ def remove_first_names():
 def remove_addresses(string):
     pass  # todo remove_addresses  idee postal.parser und zu metadaten hinzufügen
 
-def lemmatizeWord(word,lemma_dict=LEMMAS,n=3):
+def lemmatizeWord(word,lemma_dict=LEMMAS,n=5):
     for i in range(n):
         try:
             word = lemma_dict[word.lower()] if word.lower() in lemma_dict.keys() else word.lower()
@@ -134,26 +134,29 @@ def lemmatizeWord(word,lemma_dict=LEMMAS,n=3):
             print(word)
     return word
 
-def getFirstSynonym(word, thesaurus=THESAURUS,n=3):
+
+def getFirstSynonym(word, thesaurus=THESAURUS, n=3):
 
     for i in range(n):
+
         try:
-            word = thesaurus[word.lower()] if word.lower() in thesaurus.keys() else word.lower()
+            if word in thesaurus.keys():
+                return thesaurus[word]
+
+            elif word.title() in thesaurus.keys():
+                return thesaurus[word.title()]
+
+            elif word.lower() in thesaurus.keys():
+                return thesaurus[word.lower()]
+
+            else:
+                return word
+
+
+
         except:
-            print(word)
+            print("THESAURUSFEHLER BEI: {}".format(word))
-    return word
+            return word
-
-
-    """
-    if not isinstance(word, str):
-        return str(word)
-    
-    word = word.lower()
-    if word in thesaurus.keys():
-        return thesaurus[word]
-    else:
-        return str(word)
-    """
 
 
 ##########################    Spellchecking        ##########################################
@@ -328,6 +331,15 @@ corpus_en_path = FILEPATH + config.get("en_corpus", "path")
 
 
 def extract_from_corpus(corpus):
+    """
+    Extract from each doc from a corpus a string containing disired token_texts
+
+
+    :param corpus: textacy.Corpus
+    :return: string-gen
+    """
+
+    # WHITELIST erstellen. Enthält zumindest die evtuellen Topics
 
     WHITELIST = ["boss", "sap", "firefox"] #todo autogenerierung relv. techn. begriffe
 
@@ -337,6 +349,7 @@ def extract_from_corpus(corpus):
 
     WHITELIST = WHITELIST + kb_cats + kb_keys + kb_subjs
 
+
     THESAURUS = load_obj(path2thesaurus_dict)
     #WORDS = load_obj(path2wordsdict)
     LEMMAS = load_obj(path2lemmadict)
@@ -344,6 +357,9 @@ def extract_from_corpus(corpus):
     #EN_STOP_WORDS = load_obj(path2ENstopwordlist)
     VORNAMEN = load_obj(path2firstnameslist)
 
+    ents_boss = []
+    ents_sap = []
+
     for doc in corpus:
         result = []
 
@@ -353,10 +369,16 @@ def extract_from_corpus(corpus):
 
         for tok in doc:
 
-            if tok.lower_ =="boss" or tok.lower_ =="sap":
-                print(tok.lower_+": "+tok.ent_type_)
 
+            """
+            if tok.lower_ =="boss":
+                ents_boss.append(tok.ent_type_)
 
+            if tok.lower_ =="sap":
+                ents_sap.append(tok.ent_type_)
+            """
+
+            # wenn in whitelist, direkt übernehmen
             if tok.lower_ in WHITELIST:
                 result.append(tok.lower_)
 
@@ -372,25 +394,27 @@ def extract_from_corpus(corpus):
             or tok.lower_ in VORNAMEN:
                 continue
 
-            # cut after footer
-            if replaceRockDots(tok.lower_) in ["gruss", "grusse", "gruesse", "gruessen", "grusses"]:    # fehler schneidet bei INC40506 das meiste weg
-                break
 
-            # boss/SAP ent_type = 'ORG' oder '' (ein-weimal LOC oder PERSON)
+
+            # cut after footer
+            if replaceRockDots(tok.lower_) in ["gruss", "grusse", "gruesse", "gruessen", "grusses"]:    # fehler schneidet bei zB INC40506 das meiste weg
+                break
 
 
 
             if tok.pos_ in ["NOUN"] \
-            or tok.ent_type_ in ["NORP","FACILITY","ORG","PRODUCT","WORK_OF_ART"]:
+            or tok.ent_type_ in ["NORP","FACILITY","ORG","PRODUCT","WORK_OF_ART","LOC"]:
             #or tok.dep_ == "ROOT":
             # or tok.lower_ in NOUNS \ #,"PERSON"] \
                 toktext = tok.lower_
 
 
                 toktext = lemmatized_word
-                """
+
+                # hauptsynonym bilden idee zwar das Huaptsyn bilden und zählen aber die originalen wörter in den llda algo geben
+                """     
                 first_synonym = getFirstSynonym(toktext, thesaurus=THESAURUS)
-                if first_synonym is not None:
+                if first_synonym is not None or first_synonym != '':
                     toktext = first_synonym if len(first_synonym.split()) == 1 else toktext
                 """
 
@@ -402,6 +426,14 @@ def extract_from_corpus(corpus):
         yield " ".join(result)
 
 
+    """
+    print(list(set(ents_sap)))
+        ['', 'ORG', 'PERSON', 'LOC']
+
+    print(list(set(ents_boss)))
+        ['', 'ORG', 'PERSON', 'LOC']
+    """
+
 
 
 
@@ -433,6 +465,9 @@ def preprocessCorpus(corpus, clean_in_meta):
 
     )
 
+    # idee labeled_lines.txt enthählt bigramme mit unterstrich
+    # todo preCorpus weg. llda bekommt labaled_lines.txt und lda doctermamtrix
+
 
     # leere docs aus corpi kicken
     pre_corpus.remove(lambda doc: len(doc) == 0)

ressources/deWordNet.xml

@@ -71439,7 +71439,7 @@
 	</Sense>
 </LexicalEntry>
 <LexicalEntry id="w10531">
-	<Lemma writtenForm="Passwort" partOfSpeech="n"/>
+	<Lemma writtenForm="Passwort (Hauptform)" partOfSpeech="n"/>
 	<Sense id="w10531_2177-n" synset="de-2177-n">
 	</Sense>
 </LexicalEntry>
@@ -750689,4 +750689,4 @@
 </Synset>
 			 
 </Lexicon>
-</LexicalResource>
+</LexicalResource>

ressources/lemmas.txt

@@ -1,3 +1,5 @@
+kennwort kennworts
+kennwort kennwortes
 a as
 aachen aachens
 aal aale
@@ -358471,4 +358473,4 @@ zynisch zynischstes
 zynische zynischen
 zynischere zynischeren
 zynischste zynischsten
-zyste zysten
+zyste zysten

topicModeling.py

@@ -571,14 +571,14 @@ def jgibbsLLDA_KB_v2(corpus, path2save_results, top_topic_words = 7):
                 count_dict[kb] = 1
 
     sorted_dict = sorted(count_dict.items(), key=operator.itemgetter(1))
-
+    """
     for k,v in sorted_dict:
         subs = kb2subjects_dict[k]
         keys = kb2keywords_dict[k]
         print(subs, keys , v) # frage wieviele tickets pro topic?
 
     print("kb_entrys used: {}".format(len(sorted_dict)))     # frage wie viele kb_entry's insg genutzt?: 155
-
+    """
 
 
     labelist = ticket2keywords_dict.values()
@@ -644,7 +644,7 @@ def load_from_labled_lines(path):
 
     #idee plan
     # clean laden, pre laden
-    # unigramme und num/wort-bigramme doc-term # frage wie geht llda mit bigrammen um? idee notfalls bigramme als geklammerte "wörter"
+    # unigramme und num/wort-bigramme doc-term # frage wie geht llda mit bigrammen um? idee bigramme mit _ verbinden
         # nimm nur ngrams wo midn. ein token in pre vorkommt