def instance(self, row, diags, procs, gt):
data = np.zeros(int(self.word2vec_dims / 2), dtype=np.float32)
excludes = []
# sum over all vectors (first vector is the code token)
for diag in diags:
if self.code_type in ['pdx', 'sdx']:
excludes.append(diag)
for t in self.vectors_by_code['ICD_' + diag]:
data += t
data = unitvec(data)
data_procedures = np.zeros(int(self.word2vec_dims / 2), dtype=np.float32)
for proc in procs:
if self.code_type == 'srg':
excludes.append(proc)
for t in self.vectors_by_code['CHOP_' + proc]:
data_procedures += t
data_procedures = unitvec(data_procedures)
data = np.append(data, data_procedures)
data.resize(self.vector_size)
for i, var in enumerate(self.demo_variables_to_use):
data[self.word2vec_dims + i] = self.convert_demographic_variable(row, var)
return [data, gt, excludes]
def instance(self, row, diags, procs, gt):
sequence = []
demographic = np.zeros(self.vector_size, dtype=np.float32)
for i, var in enumerate(self.demo_variables_to_use):
demographic[self.word2vec_dims +
i] = self.convert_demographic_variable(row, var)
sequence.append(demographic)
self.demo_vars.append(demographic)
excludes = []
for diag in diags:
data = np.zeros(self.word2vec_dims, dtype=np.float32)
if self.code_type in ['pdx', 'sdx']:
excludes.append(diag)
for t in self.vectors_by_code['ICD_' + diag]:
data += t
data = unitvec(data)
data.resize(self.vector_size)
sequence.append(data)
for proc in procs:
data = np.zeros(self.word2vec_dims, dtype=np.float32)
if self.code_type == 'srg':
excludes.append(proc)
for t in self.vectors_by_code['CHOP_' + proc]:
data += t
data = unitvec(data)
data.resize(self.vector_size)
sequence.append(data)
return [sequence, gt, excludes]
def instance(self, row, diags, procs, gt):
sequence = []
demographic = np.zeros(self.vector_size, dtype=np.float32)
for i, var in enumerate(self.demo_variables_to_use):
demographic[self.word2vec_dims + i] = self.convert_demographic_variable(row, var)
sequence.append(demographic)
self.demo_vars.append(demographic)
excludes = []
for diag in diags:
data = np.zeros(self.word2vec_dims, dtype=np.float32)
if self.code_type in ["pdx", "sdx"]:
excludes.append(diag)
for t in self.vectors_by_code["ICD_" + diag]:
data += t
data = unitvec(data)
data.resize(self.vector_size)
sequence.append(data)
for proc in procs:
data = np.zeros(self.word2vec_dims, dtype=np.float32)
if self.code_type == "srg":
excludes.append(proc)
for t in self.vectors_by_code["CHOP_" + proc]:
data += t
data = unitvec(data)
data.resize(self.vector_size)
sequence.append(data)
return [sequence, gt, excludes]
if(code.startswith(code_type)):
code_vocab.append(code)
vector_size = vectors_by_codes[code_vocab[0]][0].shape[0]
print("Vector size is " + str(vector_size))
average_vector_by_code = np.zeros((len(code_vocab), vector_size), dtype=np.float32)
for i, code in enumerate(code_vocab):
vectors = vectors_by_codes[code]
data = np.zeros(vector_size, dtype=np.float32)
# sum over all vectors (first vector is the code token)
for v in vectors:
data += v
data = unitvec(data)
average_vector_by_code[i] = data
tokenizer = GermanTokenizer()
load_time = time.clock() - start
print('Loading vectors took ' + str(load_time) + ' seconds')
start = time.clock()
print("Search..")
tokens = tokenizer.tokenize(phrase)
print(tokens)
average_phrase = np.zeros(vector_size, dtype=np.float32)
for token in tokens:
if token in vector_by_token.keys():
code_vocab.append(code)
vector_size = vectors_by_codes[code_vocab[0]][0].shape[0]
print("Vector size is " + str(vector_size))
average_vector_by_code = np.zeros((len(code_vocab), vector_size),
dtype=np.float32)
for i, code in enumerate(code_vocab):
vectors = vectors_by_codes[code]
data = np.zeros(vector_size, dtype=np.float32)
# sum over all vectors (first vector is the code token)
for v in vectors:
data += v
data = unitvec(data)
average_vector_by_code[i] = data
tokenizer = GermanTokenizer()
load_time = time.clock() - start
print('Loading vectors took ' + str(load_time) + ' seconds')
start = time.clock()
print("Search..")
tokens = tokenizer.tokenize(phrase)
print(tokens)
average_phrase = np.zeros(vector_size, dtype=np.float32)
for token in tokens:
if token in vector_by_token.keys():
