def visual_page(lang):
g.lang = lang
s = set()
s.add(lang)
other_lang = list(set(language_dicts.keys()) - s)[0] # works only for two languages
g.strings = language_dicts[lang]
if request.method == 'POST':
list_data = request.form.getlist('list_query')
if list_data:
model_value = request.form.getlist('model')
if len(model_value) < 1:
model_value = [defaultmodel]
groups = []
for inputform in list_data[:10]:
group = set([process_query(w) for w in inputform.split(',') if len(w) > 1
and w.replace('_', '').replace('-', '').replace('::', '').replace(' ', '').isalnum()][:30])
groups.append(group)
querywords = [word for group in groups for word in group]
if len(set(querywords)) != len(querywords):
error_value = "Words must be unique!"
return render_template("visual.html", error=error_value, models=our_models, other_lang=other_lang,
languages=languages, url=url, usermodels=model_value)
if len(querywords) < 7:
error_value = "Too few words!"
return render_template("visual.html", error=error_value, models=our_models, other_lang=other_lang,
languages=languages, url=url, usermodels=model_value)
if "Incorrect tag!" in querywords:
error_value = "Incorrect tag!"
return render_template('visual.html', error=error_value, models=our_models, other_lang=other_lang,
languages=languages, url=url, usermodels=model_value)
classes = []
for word in querywords:
for group in groups:
if word in group:
classes.append(groups.index(group))
unknown = {}
models_row = {}
links_row = {}
frequencies = {}
for model in model_value:
if not model.strip() in our_models:
return render_template('home.html', other_lang=other_lang, languages=languages, url=url,
usermodels=model_value)
frequencies[model] = {}
unknown[model] = set()
words2vis = querywords
m = hashlib.md5()
name = ':::'.join(['__'.join(group) for group in groups])
name = name.encode('ascii', 'backslashreplace')
m.update(name)
fname = m.hexdigest()
plotfile = "%s_%s.png" % (model, fname)
identifier = plotfile[:-4]
models_row[model] = plotfile
labels = []
if not os.path.exists(root + 'data/images/tsneplots'):
os.makedirs(root + 'data/images/tsneplots')
if not os.access(root + 'data/images/tsneplots/' + plotfile, os.F_OK):
print('No previous image found', root + 'data/images/tsneplots/' + plotfile, file=sys.stderr)
vectors = []
for w in words2vis:
if model_props[model]['tags'] == 'False':
message = {'operation': '4', 'query': w.split('_')[0], 'model': model}
else:
message = {'operation': '4', 'query': w, 'model': model}
result = json.loads(serverquery(message).decode('utf-8'))
frequencies[model].update(result['frequencies'])
if w.split('_')[0] in frequencies[model] and w not in frequencies[model]:
frequencies[model][w] = frequencies[model][w.split('_')[0]]
if w + " is unknown to the model" in result:
unknown[model].add(w)
continue
vector = np.array(result['vector'])
vectors.append(vector)
labels.append(w)
if len(vectors) > 5:
if len(list_data) == 1 and model_props[model]['tags'] == 'True':
classes = [word.split('_')[-1] for word in labels]
print('Embedding...', file=sys.stderr)
matrix2vis = np.vstack(([v for v in vectors]))
embed(labels, matrix2vis.astype('float64'), classes, model, fname)
models_row[model] = plotfile
if tensorflow_integration:
l2c = word2vec2tensor(identifier, vectors, labels, classes)
else:
l2c = None
links_row[model] = l2c
else:
models_row[model] = "Too few words!"
else:
if tensorflow_integration:
links_row[model] = open(root + 'data/images/tsneplots/' + identifier + '.url', 'r').read()
else:
links_row[model] = None
return render_template('visual.html', languages=languages, visual=models_row, words=groups,
number=len(model_value), models=our_models, unknown=unknown, url=url,
usermodels=model_value, l2c=links_row, qwords=querywords, frequencies=frequencies)
else:
error_value = "Incorrect query!"
return render_template("visual.html", error=error_value, models=our_models, other_lang=other_lang,
languages=languages, url=url, usermodels=[defaultmodel])
return render_template('visual.html', models=our_models, other_lang=other_lang, languages=languages,
url=url, usermodels=[defaultmodel])
def visual_page(lang):
g.lang = lang
s = set()
s.add(lang)
other_lang = list(set(language_dicts.keys()) -
s)[0] # works only for two languages
g.strings = language_dicts[lang]
if request.method == 'POST':
list_data = 'dummy'
try:
list_data = request.form['list_query']
except:
pass
if list_data != 'dummy':
model_value = request.form.getlist('model')
if len(model_value) < 1:
model_value = [defaultmodel]
querywords = set([
process_query(w) for w in list_data.split()
if len(w) > 1 and w.replace('_', '').replace('-', '').replace(
'::', '').isalnum()
][:30])
if len(querywords) < 7:
error_value = "Too few words!"
return render_template("visual.html",
error=error_value,
models=our_models,
other_lang=other_lang,
languages=languages,
url=url,
usermodels=model_value)
if "Incorrect tag!" in querywords:
return render_template('visual.html',
word=list_data,
models=our_models,
other_lang=other_lang,
languages=languages,
url=url,
usermodels=model_value)
unknown = {}
models_row = {}
for model in model_value:
if not model.strip() in our_models:
return render_template('home.html',
other_lang=other_lang,
languages=languages,
url=url,
usermodels=model_value)
unknown[model] = set()
words2vis = querywords
m = hashlib.md5()
name = '_'.join(words2vis).encode('ascii', 'backslashreplace')
m.update(name)
fname = m.hexdigest()
plotfile = "%s_%s.png" % (model, fname)
models_row[model] = plotfile
labels = []
if not os.path.exists(root + 'data/images/tsneplots'):
os.makedirs(root + 'data/images/tsneplots')
if not os.access(root + 'data/images/tsneplots/' + plotfile,
os.F_OK):
print >> sys.stderr, 'No previous image found'
vectors = []
for w in words2vis:
message = "4;" + w + ";" + model
result = serverquery(message)
if 'is unknown' in result:
unknown[model].add(w)
continue
vector = np.array(result.split(','))
vectors.append(vector)
labels.append(w)
if len(vectors) > 1:
matrix2vis = np.vstack(([v for v in vectors]))
embed(labels, matrix2vis.astype('float64'), model)
m = hashlib.md5()
name = '_'.join(labels).encode('ascii',
'backslashreplace')
m.update(name)
fname = m.hexdigest()
plotfile = "%s_%s.png" % (model, fname)
models_row[model] = plotfile
else:
models_row[model] = "Too few words!"
return render_template('visual.html',
visual=models_row,
words=querywords,
number=len(model_value),
models=our_models,
unknown=unknown,
url=url,
usermodels=model_value)
else:
error_value = "Incorrect query!"
return render_template("visual.html",
error=error_value,
models=our_models,
other_lang=other_lang,
languages=languages,
url=url,
usermodels=[defaultmodel])
return render_template('visual.html',
models=our_models,
other_lang=other_lang,
languages=languages,
url=url,
usermodels=[defaultmodel])