def sentence_to_word(sentence, lang):
if(lang=='en'):
words = nltk.word_tokenize(sentence)
return [e.lower() for e in words]
else:
tokenizer = SinhalaTokenizer()
return tokenizer.tokenize(sentence)
def get_sn_process_setup():
"""
Returns a tokenizer and stemmer object for sinhala language
"""
tokenizer = SinhalaTokenizer()
stemmer = sinhalaStemmer.stemmer()
return tokenizer, stemmer
class SyntaxAnalysis:
def __init__(self):
self.__tokenizer = SinhalaTokenizer()
self.__tagger = POSTagger()
self.__stemming = Stemming()
def pos_tagger(self, sentence):
# tokenization
tokens = [self.__tokenizer.tokenize(f'{ss}.') for ss in self.__tokenizer.split_sentences(sentence)]
# Stemming
stems = []
for token in tokens[0]:
stems.append(self.__stemming.find_root(token))
# POS Tagging
pos_tags = self.__tagger.predict([stems])
return pos_tags
from elasticsearch import Elasticsearch
from sinling import SinhalaTokenizer
from sinling import word_splitter
es = Elasticsearch()
tokenizer = SinhalaTokenizer()
artist_name_boosters = [
'ගේ', 'කීව', 'කී', 'ගායනා කරන', 'ගයන', 'ගායනා', 'ගේ', 'හඩින්', 'කියනා',
'කිව්ව', 'කිව්', 'කිව', 'ගායනය', 'ගායනා කළා', 'ගායනා කල', 'ගැයූ'
]
writer_name_boosters = [
'ලියා', 'ලියූ', 'ලිව්ව', 'ලිව්', 'රචනා', 'ලියා ඇති', 'රචිත', 'ලියන ලද',
'ලියන', 'හදපු', 'පද', 'රචනය', 'හැදූ', 'හැදුව', 'ලියන', 'ලියන්න', 'ලීව',
'ලියපු', 'ලියා ඇත', 'ලිඛිත'
]
music_boosters = ["වාදනය", "සංගීතය", "නාද"]
movie_boosters = ['චිත්‍රපට', 'සිනමා']
genre_boosters = [
'දේවානුභාවයෙ', 'පැරණි', 'පොප්ස්', 'පොප්', 'පරණ', 'ක්ලැසික්', 'ක්ලැසි',
'ඉල්ලීම', 'චිත්‍රපට', 'නව', 'වර්ගයේ', 'අයත්', 'වර්ගයට', 'කණ්ඩායම්', 'යුගල'
]
views_boosters = [
'සුපිරි', 'නියම', 'ප්‍රමුඛතම', 'පට්ට', 'ඉහළම', 'ගෝල්ඩන්', 'හොඳ', 'හොඳම',
'එලකිරි', 'එළකිරි', 'සුප්පර්', 'සුප්රකට', 'ඉහල', 'වැඩිපුර', 'වැඩිපුරම',
'සුප්රකට', 'ජනප්රිය', 'ජනප්රියම', 'ජනප්රිය', 'ජනප්රියම', 'ප්‍රකට',
'ප්‍රසිද්ධ'
]
boosts_default = {
"title_sinhala": 1,
"artist_name": 1,
"writer_name": 1,
def get_sn_process_setup():
tokenizer = SinhalaTokenizer()
stemmer = sinhalaStemmer.stemmer()
return tokenizer, stemmer
class QueryProcessor:
def __init__(self):
self.tokenizer = SinhalaTokenizer()
self.es = Elasticsearch()
self.index = "160376l-ssb-data-2020-modified-index7"
self.translation_dict = {}
# Translate a word from English to Sinhala
def translate_word(self, word):
translated = translate(word, 'si', 'en')
return translated
# Translate an array from English to Sinhala
def translate_array(self, wordlist):
isAscii = lambda s: len(s) == len(s.encode())
translated_array = []
for i in wordlist:
# Check if word is made of ASCII Letters
if isAscii(i):
if i in self.translation_dict.keys():
translated_array.append(self.translation_dict.get(i))
else:
translated_phrase = self.translate_word(i)
self.translation_dict[i] = translated_phrase
translated_array.append(translated_phrase)
return translated_array
# Genereate ES Query for Advacned Queries
def advancedQuery(self, queryDictionary):
multTermValue = []
for i in queryDictionary:
if (queryDictionary[i] != None and queryDictionary[i] != ""):
queryDictionary[i] = queryDictionary[i].replace("."," ")
tokens = self.tokenizer.tokenize(queryDictionary[i])
tokens.extend(self.translate_array(tokens))
stemmed_tokens = self.stemming(tokens)
act = self.autocorrect(stemmed_tokens)
flat_list_act = []
for sublist in act:
for item in sublist:
flat_list_act.append(item)
flat_list_act.append(queryDictionary[i])
multTermValue.append({"terms": {i: flat_list_act, "boost": 2}})
# Generate a ES Boolean Query
res = self.es.search(
index=self.index,
body=
{
"query":
{
"bool": {
"should": multTermValue,
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field": "views",
"ranges": [
{
"from": 0,
"to": 1000
},
{
"from": 1000,
"to": 2000
},
{
"from": 2000,
"to": 3000
},
{
"from": 3000
}
]
}
}
}
}
)
results = res
return results
# Generate ES Query with Boosted MetaData tags
def generateTermsMultipleQuery(self, flat_list_act, fields, classDict, searchQuery):
multTermValue = []
sorted = False
sortedsimpleQuery = False
addedFields = []
for i in fields:
if (i != "popularity"):
multTermValue.append({"terms": {i: flat_list_act, "boost": classDict[i] + 1}})
addedFields.append(i)
else:
sorted = True
if (len(fields) == 1):
sortedsimpleQuery = True
for i in ["writer", "composer", "artist", "genre", "key", "beat", "movie"]:
if i not in addedFields:
multTermValue.append({"terms": {i: flat_list_act, "boost": 1}})
if (not sorted):
for i in ["title", "songLyricsSearchable"]:
if i not in addedFields:
multTermValue.append({"terms": {i: flat_list_act}})
multTermValue.append({"match_phrase": {i: searchQuery}})
# If no ranking terms (i.e හොඳම) have been provided do not sort results
if (not sorted):
res = self.es.search(
index=self.index,
body=
{
"query":
{
"bool": {
"should": multTermValue,
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field": "views",
"ranges": [
{
"from": 0,
"to": 1000
},
{
"from": 1000,
"to": 2000
},
{
"from": 2000,
"to": 3000
},
{
"from": 3000
}
]
}
}
}
}
)
# If ranking terms (i.e හොඳම) have been provided sort results based on the function score
elif (sorted and not sortedsimpleQuery):
res = self.es.search(
index=self.index,
body=
{
"query": {
"function_score": {
"functions": [
{
"field_value_factor": {
"field": "views",
"factor": 1.0001,
"missing": 1
}
}
],
"query":
{
"bool": {
"should": multTermValue
}
},
# "sort": [
# "_score",
# {"views": { "order": "desc" }}
# ],
"score_mode": "multiply"
},
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field": "views",
"ranges": [
{
"from": 0,
"to": 1000
},
{
"from": 1000,
"to": 2000
},
{
"from": 2000,
"to": 3000
},
{
"from": 3000
}
]
}
}
}
}
)
elif (sorted and sortedsimpleQuery):
res = self.es.search(
index=self.index,
body=
{
"query": {
"match_all" : {}
},
"sort": [
{"views": { "order": "desc" }}
],
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field": "views",
"ranges": [
{
"from": 0,
"to": 1000
},
{
"from": 1000,
"to": 2000
},
{
"from": 2000,
"to": 3000
},
{
"from": 3000
}
]
}
}
}
}
)
results = res
return results
# Generate a Normal Query with no Boosted MetaData tags
def generateNormalQuery(self, flat_list_act, searchQuery):
print("[INFO] Generating Normal Query")
multTermValue = []
for i in ["artist", "writer", "genre", "composer", "title", "songLyricsSearchable", "movie", "beat", "key"]:
multTermValue.append({"terms": {i: flat_list_act, "boost": 1}})
multTermValue.append({"match_phrase": {i: searchQuery}})
print(multTermValue)
res = self.es.search(
index=self.index,
body=
{
"query":
{
"bool": {
"should": multTermValue
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field": "views",
"ranges": [
{
"from": 0,
"to": 1000
},
{
"from": 1000,
"to": 2000
},
{
"from": 2000,
"to": 3000
},
{
"from": 3000
}
]
}
}
}
}
)
results = res
return results
# Analyze the word and generate appropriate keyword for user's search request
def generateQuery(self, searchQuery):
print("[INFO] Generating Query")
searchQuery = searchQuery.replace("."," ")
tokens = self.tokenizer.tokenize(searchQuery)
tokens.extend(self.translate_array(tokens))
stemmed_tokens = self.stemming(tokens)
act = self.autocorrect(stemmed_tokens)
flat_list_act = []
for sublist in act:
for item in sublist:
flat_list_act.append(item)
classDict = self.searchClassification(act)
if (len(classDict) <= 0):
results = self.generateNormalQuery(flat_list_act, searchQuery)
# self.generateMLTQuery(searchQuery, ["artist","songLyricsSearchable","writer","composer","genre"])
else:
rankedlist = []
for i in classDict:
if (i in ["writer", "composer", "artist", "genre", "popularity", "key", "beat", "movie"]):
rankedlist.append(i)
if i == "key":
p = re.compile(r"[A-G,a-g][b,#]{0,1} (major|minor|Major|Major)")
r = p.search(searchQuery)
flat_list_act.append(r[0])
if i == "beat":
p = re.compile(r"\b[0-9]{1,2}\/[0-9]{1,2}")
r = p.findall(searchQuery)
flat_list_act.append(r[0])
if (len(rankedlist) > 0):
results = self.generateTermsMultipleQuery(flat_list_act, rankedlist, classDict, searchQuery)
# self.generateFuzzyQuery()
return results
def getSubsets(self, iterable):
return chain.from_iterable(combinations(iterable, r) for r in range(len(iterable) + 1))
# Observe tokens and return query type
# If special keyword boosting terms exist return the presence of those keywords
def searchClassification(self, tokens):
synonyms = "resource_files/synonyms.txt"
try:
synonymsFile = io.open(synonyms, "r", encoding='utf-8').read()
except UnicodeDecodeError:
synonymsFile = io.open(synonyms, "r", encoding='latin-1').read()
synonymsList = synonymsFile.split("\n")
synonymsDict = {}
for i in synonymsList:
splitSynonymLine = i.split(":")
try:
synonymsDict[splitSynonymLine[0]] = splitSynonymLine[1].split(",")
except:
print()
rankedQuery = {}
for corrected_tokens in tokens:
for token in corrected_tokens:
foundsynonym = False
for key in synonymsDict:
if token in synonymsDict[key]:
if rankedQuery.get(key) == None:
rankedQuery[key] = 1
else:
rankedQuery[key] = rankedQuery[key] + 1
foundsynonym = True
if foundsynonym:
break
return rankedQuery
# Looks at basic error rules within the Sinhala Lanugage and appends likely errors
def autocorrect(self, tokens):
allWords = [[] for i in range(len(tokens))]
missFileDirec = "resource_files/mispellings.txt"
try:
misspellingsFile = io.open(missFileDirec, "r", encoding='utf-8').read()
except UnicodeDecodeError:
misspellingsFile = io.open(missFileDirec, "r", encoding='latin-1').read()
missList = misspellingsFile.split()
missListSet = []
for i in missList:
missListSet.append(i.split(','))
for token_number in range(len(tokens)):
token = tokens[token_number]
missListForWord = []
for misspellPairs in missListSet:
if misspellPairs[0] in token or misspellPairs[1] in token:
missListForWord.append(misspellPairs)
print(list(self.getSubsets(missListForWord)))
for j in list(self.getSubsets(missListForWord)):
k = token
for d in list(j):
if d[0] in k:
k = k.replace(d[0], d[1])
elif d[1] in token:
k = k.replace(d[1], d[0])
allWords[token_number].append(k)
return allWords
def rreplace(self, s, old, new, occurrence):
li = s.rsplit(old, occurrence)
return new.join(li)
# Reduce strings to simple formats based on rules
def stemming(self, doc):
suffFileDirec = "resource_files/suffixes.txt"
try:
suffixFile = io.open(suffFileDirec, "r", encoding='utf-8').read()
except UnicodeDecodeError:
suffixFile = io.open(suffFileDirec, "r", encoding='latin-1').read()
suffixList = suffixFile.split()
doc.sort()
stemmedWordlist = []
stemmedWordlist.extend(doc)
for i in doc:
for j in suffixList:
if i.endswith(j):
stemmedWordlist.append(self.rreplace(i, j, "", 1))
return stemmedWordlist
def __init__(self):
self.tokenizer = SinhalaTokenizer()
self.es = Elasticsearch()
self.index = "160376l-ssb-data-2020-modified-index7"
self.translation_dict = {}
class QueryProcessor:
def __init__(self):
self.tokenizer = SinhalaTokenizer()
self.es = Elasticsearch()
self.index = "160376l-ssb-data-2020-modified-index4"
self.translation_dict = {}
def translate_word(self, word):
translated = translate(word, 'si', 'en')
return translated
def translate_array(self, wordlist):
isascii = lambda s: len(s) == len(s.encode())
translated_array = []
for i in wordlist:
if isascii(i):
if i in self.translation_dict.keys():
translated_array.append(self.translation_dict.get(i))
else:
translated_phrase = self.translate_word(i)
self.translation_dict[i] = translated_phrase
translated_array.append(translated_phrase)
return translated_array
def advancedQuery(self, queryDictionary):
multTermValue = []
for i in queryDictionary:
if (queryDictionary[i] != None and queryDictionary[i] != ""):
tokens = self.tokenizer.tokenize(queryDictionary[i])
tokens.extend(self.translate_array(tokens))
stemmed_tokens = self.stemming(tokens)
act = self.autocorrect(stemmed_tokens)
flat_list_act = []
for sublist in act:
for item in sublist:
flat_list_act.append(item)
flat_list_act.append(queryDictionary[i])
multTermValue.append({"terms": {i: flat_list_act, "boost": 2}})
print(multTermValue)
res = self.es.search(index=self.index,
body={
"query": {
"bool": {
"should": multTermValue,
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field":
"views",
"ranges": [{
"from": 0,
"to": 1000
}, {
"from": 1000,
"to": 2000
}, {
"from": 2000,
"to": 3000
}, {
"from": 3000
}]
}
}
}
})
results = res
return results
def generateMLTQuery(self, searchQuery, rankedlist):
print("[INFO] Generating ranked Query")
res = self.es.search(index=self.index,
body={
"query": {
"more_like_this": {
"fields": rankedlist,
"like": searchQuery,
"min_term_freq": 1,
"max_query_terms": 12
}
}
})
results = res['hits']['hits']
return results
def generateTermsMultipleQuery(self, flat_list_act, fields, classDict):
multTermValue = []
sorted = False
addedFields = []
for i in fields:
if (i != "popularity"):
multTermValue.append(
{"terms": {
i: flat_list_act,
"boost": classDict[i] + 1
}})
addedFields.append(i)
else:
sorted = True
if (len(fields) == 1):
for i in [
"writer", "composer", "artist", "genre", "key",
"beat", "movie"
]:
for b in flat_list_act:
wildcardString = "*" + b + "*"
multTermValue.append(
{"wildcard": {
i: wildcardString
}})
for i in [
"writer", "composer", "artist", "genre", "key", "beat", "movie"
]:
if i not in addedFields:
for b in flat_list_act:
wildcardString = "*" + b + "*"
multTermValue.append({"wildcard": {i: wildcardString}})
if (not sorted):
for i in ["title", "songLyricsSearchable"]:
multTermValue.append({"terms": {i: flat_list_act}})
multTermValue.append({"match_phrase": {i: searchQuery}})
print(multTermValue)
if (not sorted):
res = self.es.search(index=self.index,
body={
"query": {
"bool": {
"should": multTermValue,
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field":
"views",
"ranges": [{
"from": 0,
"to": 1000
}, {
"from": 1000,
"to": 2000
}, {
"from": 2000,
"to": 3000
}, {
"from": 3000
}]
}
}
}
})
else:
print("run query")
res = self.es.search(index=self.index,
body={
"query": {
"bool": {
"should": multTermValue
}
},
"sort":
["_score", {
"views": {
"order": "desc"
}
}],
"size":
100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field":
"views",
"ranges": [{
"from": 0,
"to": 1000
}, {
"from": 1000,
"to": 2000
}, {
"from": 2000,
"to": 3000
}, {
"from": 3000
}]
}
}
}
})
results = res
return results
def generateTermsSingleQuery(self, flat_list_act, field):
res = self.es.search(index=self.index,
body={"query": {
"terms": {
field: flat_list_act
}
}})
results = res['hits']['hits']
return results
def generateNormalQuery(self, flat_list_act, searchQuery):
print("[INFO] Generating Normal Query")
multTermValue = []
for i in [
"artist", "writer", "genre", "composer", "title",
"songLyricsSearchable", "movie", "beat", "key"
]:
for b in flat_list_act:
wildcardString = "*" + b + "*"
multTermValue.append({"wildcard": {i: wildcardString}})
multTermValue.append({"match_phrase": {i: searchQuery}})
print(multTermValue)
res = self.es.search(index=self.index,
body={
"query": {
"bool": {
"should": multTermValue
}
},
"size": 100,
"aggs": {
"Artist Filter": {
"terms": {
"field": "artist.keyword",
"size": 10
}
},
"Composer Filter": {
"terms": {
"field": "composer.keyword",
"size": 10
}
},
"Genre Filter": {
"terms": {
"field": "genre.keyword",
"size": 10
}
},
"Movie Filter": {
"terms": {
"field": "movie.keyword",
"size": 10
}
},
"Writer Filter": {
"terms": {
"field": "writer.keyword",
"size": 10
}
},
"Key Filter": {
"terms": {
"field": "key.keyword",
"size": 10
}
},
"Beat Filter": {
"terms": {
"field": "beat.keyword",
"size": 10
}
},
"View Filter": {
"range": {
"field":
"views",
"ranges": [{
"from": 0,
"to": 1000
}, {
"from": 1000,
"to": 2000
}, {
"from": 2000,
"to": 3000
}, {
"from": 3000
}]
}
}
}
})
results = res
return results
def generateQuery(self, searchQuery):
print("[INFO] Generating Query")
tokens = self.tokenizer.tokenize(searchQuery)
tokens.extend(self.translate_array(tokens))
stemmed_tokens = self.stemming(tokens)
#act = self.autocorrect(stemmed_tokens)
act = [[i] for i in stemmed_tokens]
flat_list_act = []
for sublist in act:
for item in sublist:
flat_list_act.append(item)
classDict = self.searchClassification(act)
if (len(classDict) <= 0):
results = self.generateNormalQuery(flat_list_act, searchQuery)
# self.generateMLTQuery(searchQuery, ["artist","songLyricsSearchable","writer","composer","genre"])
else:
rankedlist = []
for i in classDict:
if (i in [
"writer", "composer", "artist", "genre", "popularity",
"key", "beat", "movie"
]):
rankedlist.append(i)
if i == "key":
p = re.compile(
r"[A-G,a-g][b,#]{0,1} (major|minor|Major|Major)")
r = p.search(searchQuery)
flat_list_act.append(r[0])
if i == "beat":
p = re.compile(r"\b[0-9]{1,2}\/[0-9]{1,2}")
r = p.findall(searchQuery)
flat_list_act.append(r[0])
if (len(rankedlist) > 0):
results = self.generateTermsMultipleQuery(
flat_list_act, rankedlist, classDict)
# self.generateFuzzyQuery()
return results
def getSubsets(self, iterable):
return chain.from_iterable(
combinations(iterable, r) for r in range(len(iterable) + 1))
# Observe tokens and return query type as 'Normal Lyric Search', 'Feature Search', 'Ranked Feature Search'
def searchClassification(self, tokens):
synonyms = "synonyms.txt"
try:
synonymsFile = io.open(synonyms, "r", encoding='utf-8').read()
except UnicodeDecodeError:
synonymsFile = io.open(synonyms, "r", encoding='latin-1').read()
synonymsList = synonymsFile.split("\n")
synonymsDict = {}
for i in synonymsList:
splitSynonymLine = i.split(":")
try:
synonymsDict[splitSynonymLine[0]] = splitSynonymLine[1].split(
",")
except:
print()
rankedQuery = {}
for corrected_tokens in tokens:
for token in corrected_tokens:
foundsynonym = False
for key in synonymsDict:
if token in synonymsDict[key]:
if rankedQuery.get(key) == None:
rankedQuery[key] = 1
else:
rankedQuery[key] = rankedQuery[key] + 1
foundsynonym = True
if foundsynonym:
break
return rankedQuery
# Looks at basic error rules within the Sinhala Lanugage and appends likely errors
def autocorrect(self, tokens):
allWords = [[] for i in range(len(tokens))]
missFileDirec = "mispellings.txt"
try:
misspellingsFile = io.open(missFileDirec, "r",
encoding='utf-8').read()
except UnicodeDecodeError:
misspellingsFile = io.open(missFileDirec, "r",
encoding='latin-1').read()
missList = misspellingsFile.split()
missListSet = []
for i in missList:
missListSet.append(i.split(','))
for token_number in range(len(tokens)):
token = tokens[token_number]
missListForWord = []
for misspellPairs in missListSet:
if misspellPairs[0] in token or misspellPairs[1] in token:
missListForWord.append(misspellPairs)
for j in list(self.getSubsets(missListForWord)):
for d in list(j):
if d[0] in token:
token = token.replace(d[0], d[1])
elif d[1] in token:
token = token.replace(d[1], d[0])
allWords[token_number].append(token)
return allWords
def rreplace(self, s, old, new, occurrence):
li = s.rsplit(old, occurrence)
return new.join(li)
# Reduce strings to simple formats based on rules
def stemming(self, doc):
suffFileDirec = "suffixes.txt"
try:
suffixFile = io.open(suffFileDirec, "r", encoding='utf-8').read()
except UnicodeDecodeError:
suffixFile = io.open(suffFileDirec, "r", encoding='latin-1').read()
suffixList = suffixFile.split()
doc.sort()
stemmedWordlist = []
stemmedWordlist.extend(doc)
for i in doc:
for j in suffixList:
if i.endswith(j):
stemmedWordlist.append(self.rreplace(i, j, "", 1))
return stemmedWordlist
features['BOS'] = True
if i < len(sent) - 1:
word_next = sent[i + 1]
features.update({
f'+1:word': word_next,
'+1:word.isdigit()': word_next.isdigit(),
})
else:
features['EOS'] = True
return features
if __name__ == '__main__':
from sinling import SinhalaTokenizer
tokenizer = SinhalaTokenizer()
document = 'මනුෂ්‍යයා අවුරුදු ලක්ෂ ගණනක සිට වෛරස් වසංගත නිසා එළිපිටම පීඩා විඳි සත්වයෙකි. ' \
'ඇතැම් වෛරස් රෝග වලට වැක්සීන හෙවත් එන්නත් ද වෛරස් නාශක ඖෂධ ද තිබුනද සියලූ වෛරස් ' \
'සම්බන්ධයෙන් ඒ න්‍යාය වැඩ කරන්නේ නැත. වසූරිය වෛරසය මිනිසා විසින් මිහිමතින් තුරන් කර තිබේ.'
tokenized_sentences = [
tokenizer.tokenize(f'{ss}.')
for ss in tokenizer.split_sentences(document)
]
tagger = POSTagger()
pos_tags = tagger.predict(tokenized_sentences)
for sent in pos_tags:
from sinling import SinhalaTokenizer tokenizer = SinhalaTokenizer() sentence = 'එච්.ආර්.ජෝතිපාල' s = tokenizer.tokenize(sentence) print(s) from SinhalaStemming import sinhalaStemmer from SinhalaStemming import sinhalaStemmer testx = sinhalaStemmer.stemmer() print(testx)
def __init__(self):
self.__tokenizer = SinhalaTokenizer()
self.__tagger = POSTagger()
self.__stemming = Stemming()
def search(term: str, count: int) -> List[SearchResult]:
client = Elasticsearch()
# Elasticsearch 6 requires the content-type header to be set, and this is
# not included by default in the current version of elasticsearch-py
client.transport.connection_pool.connection.headers.update(HEADERS)
tokenizer = SinhalaTokenizer()
terms = tokenizer.tokenize(term)
print(terms)
if (term == "songs" or terms == []):
s = Search(using=client, index=INDEX_NAME)
docs = s.query({"bool": {
"must": [{
"match_all": {}
}]
}})[:count].execute()
return [SearchResult.from_doc(d) for d in docs]
if ('top' in term and ('songs' in term or 'artist' in term)):
if ('songs' in term):
bool_query = {
'bool': {
'must': {
'range': {
'track_rating.sort': {
'gte': 0
}
}
},
}
}
s = Search(using=client, index="tokenized")
docs = s.query(bool_query)[:count].sort(
'-track_rating.sort').execute()
return [SearchResult.from_doc(d) for d in docs]
if ('artis' in term):
bool_query = {
'bool': {
'must': {
'range': {
'artist_rating': {
'gte': 0
}
}
},
}
}
s = Search(using=client, index="tokenized")
docs = s.query(bool_query)[:count].sort(
'-artist_rating.sort').execute()
return [SearchResult.from_doc(d) for d in docs]
elif ('artist' in terms and ':' in terms):
terms.remove('artist')
terms.remove(':')
term = " ".join(terms)
print('artist got here ' + term)
bool_query = {
'bool': {
'must': {
'match': {
'artist_name': {
'query': term,
'operator': 'and',
'fuzziness': 'AUTO'
}
}
},
'should': {
'multi_match': {
'query': term,
'fields': ['title^2', 'lyrics'],
'type': 'best_fields',
'operator': 'or'
}
}
}
}
s = Search(using=client, index="tokenized")
docs = s.query(bool_query)[:count].execute()
return [SearchResult.from_doc(d) for d in docs]
elif ('lyrics' in terms and ':' in terms):
terms.remove('lyrics')
terms.remove(':')
term = " ".join(terms)
print('lyrics got here ' + term)
bool_query = {
'bool': {
'must': {
'match': {
'lyrics': {
'query': term,
'operator': 'and',
'fuzziness': '2'
}
}
},
'should': {
'multi_match': {
'query': term,
'fields': ['title^3', 'artist_name'],
'type': 'best_fields',
'operator': 'and'
}
}
}
}
s = Search(using=client, index=INDEX_NAME)
docs = s.query(bool_query)[:count].execute()
return [SearchResult.from_doc(d) for d in docs]
elif ('album' in terms and ':' in terms):
terms.remove('album')
terms.remove(':')
term = " ".join(terms)
print('albuns got here ' + term)
bool_query = {
'bool': {
'must': {
'match': {
'album_name': {
'query': term,
'operator': 'and',
'fuzziness': 'AUTO'
}
}
},
'should': {
'multi_match': {
'query': term,
'fields': ['title^3', 'artist_name'],
'type': 'best_fields',
'operator': 'and'
}
}
}
}
s = Search(using=client, index=INDEX_NAME)
docs = s.query(bool_query)[:count].execute()
return [SearchResult.from_doc(d) for d in docs]
else:
term = " ".join(terms)
print('else got here ' + term)
s = Search(using=client, index=INDEX_NAME)
title_query = {
'match': {
'title': {
'query': term,
'operator': 'and',
'fuzziness': 'AUTO'
}
}
}
lyrics_query = {
'match': {
'lyrics': {
'query': term,
'operator': 'and',
'fuzziness': 'AUTO'
}
}
}
artist_query = {
'match': {
'artist_name': {
'query': term,
'operator': 'and',
'fuzziness': 'AUTO'
}
}
}
dis_max_query = {
'dis_max': {
'queries': [title_query, artist_query]
},
"tie-breaker": 0.5
}
docs = s.query(dis_max_query)[:count].execute()
#print(docs[0].title)
return [SearchResult.from_doc(d) for d in docs]