def get_doc_page(request, alg_db, doc_title, docid, docloc, doc_cutoff=10, topic_cutoff=10, alg=''):
"""
return the document page to the user with related terms and topics and the document text
TODO limit the length of the document returned to first XY bytes
"""
myrelations = relations(alg_db)
doc = Document(docid, doc_title)
topics = myrelations.get_top_related_topics(doc, topic_cutoff)
piearray = get_js_doc_topic_pie_array(topics)
# related topics
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
leftcol = {'piearray':piearray, 'data':topic_keys[:topic_cutoff], 'webname':'topics'}
# related documents
docs = myrelations.get_top_related_docs(doc, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
rightcol = {'data':doc_keys[:topic_cutoff], 'webname':'documents'}
try:
doc_text_file = open(os.path.join(docloc, slugify(unicode(doc_title))),'r')
except IOError:
doc_text_file = open(os.path.join(docloc, slugify(unicode(doc_title)) + '.txt'),'r') # TODO fix hack
midcol = {'doc':gen_clean_text(doc_text_file)}
return render_to_response("three-column-vis.html", {'leftcol':leftcol,
'rightcol':rightcol, 'midcol':midcol, 'title':doc.title}, context_instance=RequestContext(request))
def get_term_page(request, alg_db, term_title, termid, term_cutoff=NUM_TERMS, doc_cutoff=10, topic_cutoff=10, alg=''):
"""
returns the term page to the user with related terms, documents, and topics
"""
# init
myrelations = relations(alg_db)
term = Term(termid, term_title, count=-1)
# related topics
# pie array
topics = myrelations.get_top_related_topics(term, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
piearray = get_js_term_topics_pie_array(myrelations, term, topic_keys)
topics = myrelations.get_top_related_topics(term, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
leftcol = {'piearray':piearray, 'data': topic_keys[:topic_cutoff], 'webname':'topics'}
# related docs
docs = myrelations.get_top_related_docs(term, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
midcol = {'data': doc_keys[:doc_cutoff], 'webname':'documents'}
# related terms
top_related_terms = myrelations.get_top_related_terms(term, term_cutoff)
rightcol = {'data': top_related_terms[:term_cutoff], 'webname':'terms'}
return render_to_response("three-column-vis.html", {'leftcol':leftcol, 'midcol':midcol, 'rightcol':rightcol, 'title':term.title}, context_instance=RequestContext(request))
def get_topic_page(request, alg_db, topic_title, topicid, term_cutoff=NUM_TERMS, doc_cutoff=10, topic_cutoff=10, alg=''):
"""
returns the topic page to the user with related terms, documents, and topics
"""
myrelations = relations(alg_db)
if not topic_title[0] == '{':
topic_title = '{' + ', '.join(topic_title.strip().split()) + '}'
topic = Topic(myrelations, topicid, topic_title)
# related terms
terms = topic.get_terms(term_cutoff)
# add an interactive pie chart
piearray = get_js_topic_term_pie_array(topic, terms) # TODO replace this in template
leftcol = {'piearray':piearray, 'data':terms, 'webname':'terms'}
# # related docs
docs = myrelations.get_top_related_docs(topic, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
midcol = {'data': doc_keys[0:doc_cutoff], 'webname':'documents'}
# # related topics
topics = myrelations.get_top_related_topics(topic, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
rightcol = {'data': topic_keys[0:topic_cutoff], 'webname':'topics'}
return render_to_response("three-column-vis.html", {'leftcol':leftcol, 'midcol':midcol, 'rightcol':rightcol, 'title':topic.title}, context_instance=RequestContext(request))
def get_rel_page(request, alg_db, dataloc, alg=''):
"""
returns the interactive (d3) relationship page
"""
myrelations = relations(alg_db)
# build the doc-doc
ddata = build_graph_json(myrelations.get_docs(), myrelations,
myrelations.get_top_related_docs)
# and term-term graphs
tdata = {}
#tdata = build_graph_json(myrelations.get_terms(), myrelations, myrelations.get_top_related_terms)
# acquire the topics
topdata = build_topic_json(myrelations)
# feed the results into the heavily customized template
ret_val = render_to_response("model-relationships.html", {
'alg': alg,
'doc_data': ddata,
'term_data': tdata,
'top_data': topdata
},
context_instance=RequestContext(request))
return ret_val
def presence_graph(request, item, alg_db, alg='', template='table-graph.html', extra_context=None, RPP=49):
"""
Returns a graph of the relative presence of each item (bar graph)
"""
if request.GET.has_key('page'):
page = int(request.GET['page'])
start_val = (page-1)*RPP
end_val = page*RPP
else:
start_val = 0
end_val = RPP
myrelations = relations(alg_db)
if item == "topics":
mobjs = myrelations.get_topics()
score_fnct = lambda topic: topic.score
max_score = Topic.max_score
title_fnct = lambda score,width: str(width) + '% of max'
elif item == "terms":
mobjs = myrelations.get_terms(start_val=start_val, end_val=end_val)
score_fnct = lambda term: term.count
max_score = Term.max_occ
title_fnct = lambda score, width: str(score) + ' occurences, ' + str(width) + '% of max'
context = {'input':get_bar_chart(mobjs, score_fnct, max_score, title_fnct), 'alg':alg, 'group_data_type': item, 'RPP':RPP}
if extra_context:
context.update(extra_context)
return render_to_response(template, context, context_instance=RequestContext(request))
def get_summary_page(request, alg_db, numterms = NUM_TERMS, numcolumns = 3, alg=''):
"""
Returns the Analyzer's summary page
@param request: the django request object
@param alg_db: the algorithm database
@param numterms: the number of terms to display for the topics
@param numcolumns: the number of columns to use to display the topics
@param alg: the name of the algorithm e.g. 'LDA' or "HDP'
"""
myrelations = relations(alg_db)
topics= myrelations.get_topics()
ncol = min(len(topics), numcolumns) # number of columns for the data table
disp_tops = []
topic_row = []
ct = 1
for i in xrange(len(topics)):
topic_trms = map( lambda x: {'title':x.title, 'id':x.id}, topics[i].get_terms(numterms)) # get the top numterms terms
topic_row.append({'title':topics[i].title, 'id':topics[i].topic_id, 'terms':topic_trms})
if ct == ncol or i==len(topics)-1: # group nicely into rows for django
ct = 1
disp_tops.append(topic_row)
topic_row = []
else:
ct += 1
template = 'summary.html'
return render_to_response(template, {'disp_topics':disp_tops, 'alg':alg}, context_instance=RequestContext(request))
def get_doc_page(request,
alg_db,
doc_title,
docid,
docloc,
doc_cutoff=10,
topic_cutoff=10,
alg=''):
"""
return the document page to the user with related terms and topics and the document text
TODO limit the length of the document returned to first XY bytes
"""
myrelations = relations(alg_db)
doc = Document(docid, doc_title)
topics = myrelations.get_top_related_topics(doc, topic_cutoff)
piearray = get_js_doc_topic_pie_array(topics)
# related topics
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
leftcol = {
'piearray': piearray,
'data': topic_keys[:topic_cutoff],
'webname': 'topics'
}
# related documents
docs = myrelations.get_top_related_docs(doc, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
rightcol = {'data': doc_keys[:topic_cutoff], 'webname': 'documents'}
try:
doc_text_file = open(os.path.join(docloc, slugify(unicode(doc_title))),
'r')
except IOError:
doc_text_file = open(
os.path.join(docloc,
slugify(unicode(doc_title)) + '.txt'),
'r') # TODO fix hack
midcol = {'doc': gen_clean_text(doc_text_file)}
return render_to_response("three-column-vis.html", {
'leftcol': leftcol,
'rightcol': rightcol,
'midcol': midcol,
'title': doc.title
},
context_instance=RequestContext(request))
def get_term_page(request,
alg_db,
term_title,
termid,
term_cutoff=NUM_TERMS,
doc_cutoff=10,
topic_cutoff=10,
alg=''):
"""
returns the term page to the user with related terms, documents, and topics
"""
# init
myrelations = relations(alg_db)
term = Term(termid, term_title, count=-1)
# related topics
# pie array
topics = myrelations.get_top_related_topics(term, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
piearray = get_js_term_topics_pie_array(myrelations, term, topic_keys)
topics = myrelations.get_top_related_topics(term, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
leftcol = {
'piearray': piearray,
'data': topic_keys[:topic_cutoff],
'webname': 'topics'
}
# related docs
docs = myrelations.get_top_related_docs(term, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
midcol = {'data': doc_keys[:doc_cutoff], 'webname': 'documents'}
# related terms
top_related_terms = myrelations.get_top_related_terms(term, term_cutoff)
rightcol = {'data': top_related_terms[:term_cutoff], 'webname': 'terms'}
return render_to_response("three-column-vis.html", {
'leftcol': leftcol,
'midcol': midcol,
'rightcol': rightcol,
'title': term.title
},
context_instance=RequestContext(request))
def get_summary_page(request,
alg_db,
numterms=NUM_TERMS,
numcolumns=3,
alg=''):
"""
Returns the Analyzer's summary page
@param request: the django request object
@param alg_db: the algorithm database
@param numterms: the number of terms to display for the topics
@param numcolumns: the number of columns to use to display the topics
@param alg: the name of the algorithm e.g. 'LDA' or "HDP'
"""
myrelations = relations(alg_db)
topics = myrelations.get_topics()
ncol = min(len(topics), numcolumns) # number of columns for the data table
disp_tops = []
topic_row = []
ct = 1
for i in xrange(len(topics)):
topic_trms = map(lambda x: {
'title': x.title,
'id': x.id
}, topics[i].get_terms(numterms)) # get the top numterms terms
topic_row.append({
'title': topics[i].title,
'id': topics[i].topic_id,
'terms': topic_trms
})
if ct == ncol or i == len(
topics) - 1: # group nicely into rows for django
ct = 1
disp_tops.append(topic_row)
topic_row = []
else:
ct += 1
template = 'summary.html'
return render_to_response(template, {
'disp_topics': disp_tops,
'alg': alg
},
context_instance=RequestContext(request))
def get_rel_page(request, alg_db, dataloc, alg=''):
"""
returns the interactive (d3) relationship page
"""
myrelations = relations(alg_db)
# build the doc-doc
ddata = build_graph_json(myrelations.get_docs(), myrelations, myrelations.get_top_related_docs)
# and term-term graphs
tdata = {}
#tdata = build_graph_json(myrelations.get_terms(), myrelations, myrelations.get_top_related_terms)
# acquire the topics
topdata = build_topic_json(myrelations)
# feed the results into the heavily customized template
ret_val = render_to_response("model-relationships.html", {'alg':alg, 'doc_data':ddata, 'term_data':tdata, 'top_data':topdata}, context_instance=RequestContext(request))
return ret_val
def table_graph_rel(request, type, alg_db, alg='', template='table-graph.html', extra_context=None, RPP=49):
"""
constructs a table-graph to display relationships in the data
"""
# check for ajax pagination
if request.GET.has_key('page'):
page = int(request.GET['page'])
start_val = (page-1)*RPP
end_val = page*RPP
else:
start_val = 0
end_val = RPP
myrelations = relations(alg_db)
num_topics = myrelations.get_num_topics()
if type == 'doc-graph':
main_objs = myrelations.get_docs(start_val=start_val, end_val=end_val)
rel_pct_fct = lambda top, doc, tops: 100*tops[top]/sum(tops.values())
get_top_related_fnct = myrelations.get_top_related_topics
group_data_type = "documents"
data_type = "topics"
color_bars = True
elif type == 'term-graph':
main_objs = myrelations.get_terms(start_val=start_val, end_val=end_val)
rel_pct_fct = myrelations.get_top_in_term_rel_pct
get_top_related_fnct = myrelations.get_top_related_topics
group_data_type = "terms"
data_type = "topics"
color_bars = True
elif type == 'topic-graph':
main_objs = myrelations.get_topics(start_val=start_val, end_val=end_val)
rel_pct_fct = myrelations.get_term_in_top_rel_pct
get_top_related_fnct = myrelations.get_topic_terms
group_data_type = "topics"
data_type = "terms"
color_bars = False
context = {'input':table_object_gen(main_objs, get_top_related_fnct, rel_pct_fct), "num_topics":num_topics, "color_bars":color_bars, 'alg':alg, 'group_data_type': group_data_type, 'data_type':data_type, 'RPP':RPP}
if extra_context:
context.update(extra_context)
return render_to_response(template, context, context_instance=RequestContext(request))
def presence_graph(request,
item,
alg_db,
alg='',
template='table-graph.html',
extra_context=None,
RPP=49):
"""
Returns a graph of the relative presence of each item (bar graph)
"""
if request.GET.has_key('page'):
page = int(request.GET['page'])
start_val = (page - 1) * RPP
end_val = page * RPP
else:
start_val = 0
end_val = RPP
myrelations = relations(alg_db)
if item == "topics":
mobjs = myrelations.get_topics()
score_fnct = lambda topic: topic.score
max_score = Topic.max_score
title_fnct = lambda score, width: str(width) + '% of max'
elif item == "terms":
mobjs = myrelations.get_terms(start_val=start_val, end_val=end_val)
score_fnct = lambda term: term.count
max_score = Term.max_occ
title_fnct = lambda score, width: str(score) + ' occurences, ' + str(
width) + '% of max'
context = {
'input': get_bar_chart(mobjs, score_fnct, max_score, title_fnct),
'alg': alg,
'group_data_type': item,
'RPP': RPP
}
if extra_context:
context.update(extra_context)
return render_to_response(template,
context,
context_instance=RequestContext(request))
def get_topic_page(request,
alg_db,
topic_title,
topicid,
term_cutoff=NUM_TERMS,
doc_cutoff=10,
topic_cutoff=10,
alg=''):
"""
returns the topic page to the user with related terms, documents, and topics
"""
myrelations = relations(alg_db)
if not topic_title[0] == '{':
topic_title = '{' + ', '.join(topic_title.strip().split()) + '}'
topic = Topic(myrelations, topicid, topic_title)
# related terms
terms = topic.get_terms(term_cutoff)
# add an interactive pie chart
piearray = get_js_topic_term_pie_array(
topic, terms) # TODO replace this in template
leftcol = {'piearray': piearray, 'data': terms, 'webname': 'terms'}
# # related docs
docs = myrelations.get_top_related_docs(topic, doc_cutoff)
doc_keys = docs.keys()
doc_keys.sort(lambda x, y: -cmp(docs[x], docs[y]))
midcol = {'data': doc_keys[0:doc_cutoff], 'webname': 'documents'}
# # related topics
topics = myrelations.get_top_related_topics(topic, topic_cutoff)
topic_keys = topics.keys()
topic_keys.sort(lambda x, y: -cmp(topics[x], topics[y]))
rightcol = {'data': topic_keys[0:topic_cutoff], 'webname': 'topics'}
return render_to_response("three-column-vis.html", {
'leftcol': leftcol,
'midcol': midcol,
'rightcol': rightcol,
'title': topic.title
},
context_instance=RequestContext(request))
def get_model_page(request, alg_db, corpus_dbloc, dataloc, alg='', num_terms=NUM_TERMS, bing_dict={}, form=None): # TODO stem the terms once for bing and wiki calculations
"""
Obtain/save the appropriate data for the model-description page
@param request: the django request object
@param alg_db: the algorithm database
@param corpus_dbloc: the location of the corpus specific database (contains cooccurence information)
@param dataloc: the location of the data for this analysis
@param alg: the algorithm used for this analysis
@param num_terms: the number of terms to display for the topics
"""
alg_loc = "%s/%s" % (dataloc, alg)
# prep the model likelihood results
lin = open(os.path.join(alg_loc,'js_likelihood.dat'), 'r')
ldata = lin.readline().strip()
lin.close()
# likelihood
myrelations = relations(alg_db)
topics = myrelations.get_topics()
log_like = {}
for tpc in topics:
terms = myrelations.get_topic_terms(tpc, num_terms)
log_like[tpc.id] = round( sum(terms.values()), 2)
# topic coherence
tc_scores = get_topic_coherence_scores(topics, corpus_dbloc)
for tid in tc_scores:
tc_scores[tid] = round(sum(tc_scores[tid]),2)
# see if we already acquired the bing scores, if not, save the topic_terms for AJAX queries
search_title_scores = {}
if bing_dict != {}:
search_title_scores= bing_dict
else:
save_topic_terms(topics, alg_loc, num_terms)
# wikipedia scores
wiki_abs_scores = get_wiki_pmi_coherence(topics)
for tid in wiki_abs_scores:
if len(wiki_abs_scores[tid]) > 0:
wiki_abs_scores[tid] = round(median(wiki_abs_scores[tid]), 2)
else:
wiki_abs_scores[tid] = -1 # TODO verify why this happens --- it's usually with strange words
# build "topic dictionaries" for faster django access
top_dicts = []
srt_tc_scores = sorted(tc_scores.values(), reverse=False)
srt_search_title_scores = sorted(search_title_scores.values(), reverse=False)
srt_wiki_abs_scores = sorted(wiki_abs_scores.values(), reverse=False)
srt_log_like = sorted(log_like.values(), reverse=False)
ntopics = len(topics)
for i in xrange(ntopics):
top_dicts.append({})
topics[i].get_terms(num_terms)
top_dicts[i]['id'] = topics[i].id
top_dicts[i]['terms'] = ', '.join([topics[i].get_term(x).title for x in range(num_terms)])
if tc_scores != {}:
top_dicts[i]['tc_score'] = tc_scores[topics[i].id]
top_dicts[i]['tc_score_alpha'] = round(srt_tc_scores.index(tc_scores[topics[i].id])/float(ntopics-1),3) # TODO remove code repetion
if search_title_scores != {}:
top_dicts[i]['search_title_score'] = search_title_scores[topics[i].id]
top_dicts[i]['search_title_score_alpha'] = round(srt_search_title_scores.index(search_title_scores[topics[i].id])/float(ntopics-1),3)
if wiki_abs_scores != {}:
top_dicts[i]['wiki_abs_score'] = wiki_abs_scores[topics[i].id]
top_dicts[i]['wiki_abs_score_alpha'] = round(srt_wiki_abs_scores.index(wiki_abs_scores[topics[i].id])/float(ntopics-1),3)
if log_like != {}:
top_dicts[i]['topic_likelihood'] = log_like[topics[i].id]
top_dicts[i]['topic_likelihood_alpha'] = round(srt_log_like.index(log_like[topics[i].id])/float(ntopics-1),3)
rgb = {"r":255,"g":171,"b":115}
ret_val = render_to_response("model-analysis.html", {'form': form,'max_ntopics': MAX_NUM_TOPICS, 'rgb':rgb, 'like_data':ldata, 'topics':top_dicts, "query_bing":search_title_scores=={}, 'do_ppx':ALLOW_PERPLEX}, context_instance=RequestContext(request))
return ret_val
def table_graph_rel(request,
type,
alg_db,
alg='',
template='table-graph.html',
extra_context=None,
RPP=49):
"""
constructs a table-graph to display relationships in the data
"""
# check for ajax pagination
if request.GET.has_key('page'):
page = int(request.GET['page'])
start_val = (page - 1) * RPP
end_val = page * RPP
else:
start_val = 0
end_val = RPP
myrelations = relations(alg_db)
num_topics = myrelations.get_num_topics()
if type == 'doc-graph':
main_objs = myrelations.get_docs(start_val=start_val, end_val=end_val)
rel_pct_fct = lambda top, doc, tops: 100 * tops[top] / sum(tops.values(
))
get_top_related_fnct = myrelations.get_top_related_topics
group_data_type = "documents"
data_type = "topics"
color_bars = True
elif type == 'term-graph':
main_objs = myrelations.get_terms(start_val=start_val, end_val=end_val)
rel_pct_fct = myrelations.get_top_in_term_rel_pct
get_top_related_fnct = myrelations.get_top_related_topics
group_data_type = "terms"
data_type = "topics"
color_bars = True
elif type == 'topic-graph':
main_objs = myrelations.get_topics(start_val=start_val,
end_val=end_val)
rel_pct_fct = myrelations.get_term_in_top_rel_pct
get_top_related_fnct = myrelations.get_topic_terms
group_data_type = "topics"
data_type = "terms"
color_bars = False
context = {
'input': table_object_gen(main_objs, get_top_related_fnct,
rel_pct_fct),
"num_topics": num_topics,
"color_bars": color_bars,
'alg': alg,
'group_data_type': group_data_type,
'data_type': data_type,
'RPP': RPP
}
if extra_context:
context.update(extra_context)
return render_to_response(template,
context,
context_instance=RequestContext(request))
def get_model_page(
request,
alg_db,
corpus_dbloc,
dataloc,
alg='',
num_terms=NUM_TERMS,
bing_dict={},
form=None): # TODO stem the terms once for bing and wiki calculations
"""
Obtain/save the appropriate data for the model-description page
@param request: the django request object
@param alg_db: the algorithm database
@param corpus_dbloc: the location of the corpus specific database (contains cooccurence information)
@param dataloc: the location of the data for this analysis
@param alg: the algorithm used for this analysis
@param num_terms: the number of terms to display for the topics
"""
alg_loc = "%s/%s" % (dataloc, alg)
# prep the model likelihood results
lin = open(os.path.join(alg_loc, 'js_likelihood.dat'), 'r')
ldata = lin.readline().strip()
lin.close()
# likelihood
myrelations = relations(alg_db)
topics = myrelations.get_topics()
log_like = {}
for tpc in topics:
terms = myrelations.get_topic_terms(tpc, num_terms)
log_like[tpc.id] = round(sum(terms.values()), 2)
# topic coherence
tc_scores = get_topic_coherence_scores(topics, corpus_dbloc)
for tid in tc_scores:
tc_scores[tid] = round(sum(tc_scores[tid]), 2)
# see if we already acquired the bing scores, if not, save the topic_terms for AJAX queries
search_title_scores = {}
if bing_dict != {}:
search_title_scores = bing_dict
else:
save_topic_terms(topics, alg_loc, num_terms)
# wikipedia scores
wiki_abs_scores = get_wiki_pmi_coherence(topics)
for tid in wiki_abs_scores:
if len(wiki_abs_scores[tid]) > 0:
wiki_abs_scores[tid] = round(median(wiki_abs_scores[tid]), 2)
else:
wiki_abs_scores[
tid] = -1 # TODO verify why this happens --- it's usually with strange words
# build "topic dictionaries" for faster django access
top_dicts = []
srt_tc_scores = sorted(tc_scores.values(), reverse=False)
srt_search_title_scores = sorted(search_title_scores.values(),
reverse=False)
srt_wiki_abs_scores = sorted(wiki_abs_scores.values(), reverse=False)
srt_log_like = sorted(log_like.values(), reverse=False)
ntopics = len(topics)
for i in xrange(ntopics):
top_dicts.append({})
topics[i].get_terms(num_terms)
top_dicts[i]['id'] = topics[i].id
top_dicts[i]['terms'] = ', '.join(
[topics[i].get_term(x).title for x in range(num_terms)])
if tc_scores != {}:
top_dicts[i]['tc_score'] = tc_scores[topics[i].id]
top_dicts[i]['tc_score_alpha'] = round(
srt_tc_scores.index(tc_scores[topics[i].id]) /
float(ntopics - 1), 3) # TODO remove code repetion
if search_title_scores != {}:
top_dicts[i]['search_title_score'] = search_title_scores[
topics[i].id]
top_dicts[i]['search_title_score_alpha'] = round(
srt_search_title_scores.index(
search_title_scores[topics[i].id]) / float(ntopics - 1), 3)
if wiki_abs_scores != {}:
top_dicts[i]['wiki_abs_score'] = wiki_abs_scores[topics[i].id]
top_dicts[i]['wiki_abs_score_alpha'] = round(
srt_wiki_abs_scores.index(wiki_abs_scores[topics[i].id]) /
float(ntopics - 1), 3)
if log_like != {}:
top_dicts[i]['topic_likelihood'] = log_like[topics[i].id]
top_dicts[i]['topic_likelihood_alpha'] = round(
srt_log_like.index(log_like[topics[i].id]) /
float(ntopics - 1), 3)
rgb = {"r": 255, "g": 171, "b": 115}
ret_val = render_to_response("model-analysis.html", {
'form': form,
'max_ntopics': MAX_NUM_TOPICS,
'rgb': rgb,
'like_data': ldata,
'topics': top_dicts,
"query_bing": search_title_scores == {},
'do_ppx': ALLOW_PERPLEX
},
context_instance=RequestContext(request))
return ret_val