def purity_report():
global modeldir, cPathFrame, num2path, path2num, scored_paths, frameScales
print "<table>"
print "<br>%d scored paths in model vocab" % len(scored_paths)
print "<br>showing top-99% scored paths mass per topic"
print "<thead><th>k <th>score <th>num nonzero types <th>n tok <th>paths"
print "<tbody>"
for k in range(cPathFrame.shape[1]):
print "<tr><td>", k
total = sum([cPathFrame[path2num[p],k] for p in scored_paths])
nnz = sum([ cPathFrame[path2num[p],k] > 0 for p in scored_paths] )
sumscore = np.sum([path2score[p]*cPathFrame[path2num[p],k] for p in scored_paths])
meanscore = sumscore / total
print "<td>"
# print "%.3g mean score, over <td>%d non-zero types, <td>%d total tokens" % (meanscore, nnz, total)
print "%.3g <td>%d <td>%d" % (meanscore, nnz, total)
print "<td>"
pathorder = sorted(scored_paths, key=lambda p: -cPathFrame[path2num[p],k])
pathorder = [p for p in pathorder if cPathFrame[path2num[p],k] > 0]
cumsum = 0
for p in pathorder:
count = cPathFrame[path2num[p], k]
cumsum += count
if cumsum >= 0.99*total: break
print util.nicepath(p)
s = path2score[p]
valence = "pos" if s>0 else "neg" if s<0 else "neu" if s==0 else None
codeurl = "http://brenocon.com/tabari_cameo_verbs.html#" + path2code[p]
print "<span class=wordinfo>(%.1f <span class='score %s'>%.1f <a href='%s'>%s</a></span>)</span>" % (
count, valence, s, codeurl, path2code[p])
print ', '
print "</table>"
print "<b>" + prefix + "</b>"
print "only showing paths with count >=", wc_thresh
print "<table class=tablesorter cellpadding=3 border=1 cellspacing=0 width='100%'>"
thead = ['eventtype'] + ['count', 'alpha', 'etaVar'
] + ['d=%d' % dim
for dim in range(Ndim)] + ['top paths']
print "<thead>", ' '.join(["<th>" + x for x in thead]), "</thead>"
print "<tbody>"
print
for k in frameorder:
top_paths = (-path_frames[:, k]).argsort()[:20]
top_paths = top_paths[path_frames[top_paths, k] >= wc_thresh]
pathelts = [util.nicepath(x) for x in path_vocab[top_paths]]
# pathelts = ["%s <span class=wordinfo>(%.0f)</span>" % (util.nicepath(path_vocab[i]), path_frames[i,k]) for i in top_paths]
pathelts = ["%s" % (util.nicepath(path_vocab[i]), ) for i in top_paths]
pathinfo = ', '.join(pathelts)
row = [
'f=%s' % k,
str(framecounts[k]),
'%.3g' % alpha[k],
'%.3g' % etaVar[k],
]
row += ["%.3g" % x for x in frame_scales[k, :]]
row += [pathinfo]
# row += [str(path_vocab[top_paths])]
print '<tr>' + ' '.join('<td>' + str(x) for x in row)
import json, sys
import util
for line in sys.stdin:
parts = line.rstrip('\n').split('\t')
parts[-1] = util.nicepath(parts[-1], html=False)
print '\t'.join(parts)
util.pageheader()
print "<b>" + prefix + "</b>"
print "only showing paths with count >=", wc_thresh
print "<table class=tablesorter cellpadding=3 border=1 cellspacing=0 width='100%'>"
thead = ['eventtype'] + ['count','alpha','etaVar'] + ['d=%d' % dim for dim in range(Ndim)] + ['top paths']
print "<thead>", ' '.join(["<th>"+x for x in thead]), "</thead>"
print "<tbody>"
print
for k in frameorder:
top_paths = (-path_frames[:,k]).argsort()[:20]
top_paths = top_paths[ path_frames[top_paths,k] >= wc_thresh]
pathelts = [util.nicepath(x) for x in path_vocab[top_paths]]
# pathelts = ["%s <span class=wordinfo>(%.0f)</span>" % (util.nicepath(path_vocab[i]), path_frames[i,k]) for i in top_paths]
pathelts = ["%s" % (util.nicepath(path_vocab[i]),) for i in top_paths]
pathinfo = ', '.join(pathelts)
row = ['f=%s' % k, str(framecounts[k]),
'%.3g' % alpha[k], '%.3g' % etaVar[k],
]
row += ["%.3g" % x for x in frame_scales[k,:]]
row += [pathinfo]
# row += [str(path_vocab[top_paths])]
print '<tr>' + ' '.join('<td>'+str(x) for x in row)
print "</tbody>"
print "</table>"