def getTagInfo(self, tag): """ DOCUMENT @return: (documents, totalsize, [rtag:(intersect,total)]) """ docs = self.tddb[tag] rel = dict((tag, (len(intersect), len(self.tddb[tag]))) for tag, intersect in invert_multimap((doc, self.dtdb[doc]) for doc in docs).iteritems()) prod = dict((prod, (len(ps), len(self.pddb[prod]))) for prod, ps in invert_multimap((doc, self.dppb[doc]) for doc in docs).iteritems()) return TagInfo(tag, docs, rel, prod, self.totalsize)
def inferRelProds(self, prod): """ Infer a set of related producers for the given producer. This implementation selects producers that hold the documents in the representative documents set of the given source producer. """ rel = invert_multimap((doc, self.dppb[doc]) for doc in prod.rep_d) if rel: del rel[prod.nsid] # if it's not empty, then it must refer back to itself return dict((nsid, float(len(docs))/len(self.pddb[nsid])) for nsid, docs in rel.iteritems())
def generatePTables(self):
"""
DOCUMENT
"""
name = "ptables"
id_u = dict((nsid, vid) for vid, nsid in enumerate(self.socgr.vs[NID]))
base_h = len(id_u)
lab_h, id_h = zip(*((nsid, (nsid, base_h+vid)) for vid, nsid in enumerate(self.gumap.iterkeys()))) if self.gumap else ([], [])
id_h = dict(id_h)
base_g = len(id_h) + base_h
lab_g, id_g = zip(*((nsid, (nsid, base_g+vid)) for vid, nsid in enumerate(self.sprdgr.vs[NID]))) if len(self.sprdgr.vs) > 0 else ([], [])
id_g = dict(id_g)
ptabgr = self.socgr.copy()
edges = set()
# add arcs to self
edges.update((vid, vid) for vid in xrange(0, base_h))
# add arcs to indexes
for nsid, users in self.gumap.iteritems():
hvid = id_h[nsid]
edges.update((id_u[user], hvid) for user in users)
# add arcs to tgraphs
phmap = {}
for i, hvids in enumerate(self.comm):
gvid = base_g + i
for hid in self.prodgr.vs.select(hvids)[NID]:
if hid in id_u:
continue
for user in self.gumap[hid]:
if user in phmap:
phmap[user].add(gvid)
else:
phmap[user] = set([gvid])
# only add some of these, to prevent a user linking to eg. 400 tgraphs
for user, gvids in phmap.iteritems():
pvid = id_u[user]
for gvid in sample(gvids, int(len(gvids)**0.5)):
edges.add((pvid, gvid))
ptabgr.add_vertices(len(id_h) + len(id_g))
eend = len(ptabgr.es)
ptabgr.add_edges(edges)
ptabgr.es[eend:][AAT] = [0.5] * len(edges)
ptabgr.es.select(ptabgr.get_eid(vid, vid) for vid in xrange(0, base_h))[AAT] = [1.0] * base_h
ptabgr.vs[base_h:][NID] = lab_h + lab_g
ptabgr["base_z"] = 0
ptabgr["base_h"] = base_h
ptabgr["base_g"] = base_g
self.ptabgr = ptabgr
# for easy access / human readability
ugmap = dict((nsid, set(gnsid)) for nsid, gnsid in invert_multimap(self.gumap.iteritems(),
dict((nsid, []) for nsid in self.socgr.vs[NID])).iteritems())
for i, pvid in enumerate(self.comm):
spid = self.sprdgr.vs[i][NID]
for nsid in self.prodgr.vs.select(pvid)[NID]:
if nsid in ugmap:
continue
for uid in self.gumap[nsid]:
ugmap[uid].add(spid)
self.ptbmap = ugmap
