lda-prepass.py

from neo4j.v1 import GraphDatabase
from gensim import corpora, models

driver = GraphDatabase.driver("bolt://XXXXX:7687", auth=("neo4j", "XXXX"))
session = driver.session()

q = "MATCH (m:Memory) WHERE id(m) = 34452 \
MATCH (m)-[:HAS_CONTENT]->(c)-[:HAS_ANNOTATED_TEXT]->(at)-[:CONTAINS_SENTENCE]->(s)-[:HAS_TAG]->(tag) RETURN collect(tag.value) AS tags"

result = list(session.run(q))

t = result[0]["tags"]
tags = []
for i in t:
    tags.append(t)


# print(tags)

dictionary = corpora.Dictionary(tags)
corpus = [dictionary.doc2bow(text) for text in tags]
ldamodel = models.ldamodel.LdaModel(corpus, num_topics=10, id2word = dictionary, passes=10)

topics = ldamodel.show_topics(num_words=10, formatted=False)

ldaTopics = []


for topic in topics[0][1]:
    ldaTopics.append({'tag': topic[0], 'weight': topic[1]})

ldaQ = "MATCH (m:Memory) WHERE id(m) = 34452 " \
       "UNWIND {topics} AS topic " \
       "MATCH (t:Tag {id: topic.tag + '_en'}) " \
       "MERGE (m)-[r:LDA_TOPIC_PREPASS]->(t) SET r.weight = topic.weight"

session.run(ldaQ, {'topics': ldaTopics}).consume()