def add_pub_concepts(year):
#this is all to slow, used neo4j csv import instead
#using periodic commit 10000 load CSV from "file:///pub_concept_2008.txt" as row FIELDTERMINATOR '\t' match (p:Publication{pub_id:toInt(row[0])}) match (c:Concept{name:row[1]}) merge (p)-[:CONCEPT{year:2008}]-(c) return count(p);
print "Adding pub-concept data"
session = neo4j_functions.connect()
#read data
pubConceptDic = readPubConcepts(year)
counter = 0
for p in pubConceptDic:
for c in pubConceptDic[p]:
c = c.replace("'", "\\'")
if counter % 10000 == 0:
t = strftime("%H:%M:%S", gmtime())
print t, counter
session.close()
session = neo4j_functions.connect()
com = "match (p:Publication {pub_id:" + str(
p
) + "}) match (c:Concept {name:'" + c + "'}) merge (p)-[:CONCEPT{year:" + str(
year) + "}]-(c);"
session.run(com)
counter += 1
def pubs_per_org():
session = neo4j_functions.connect()
com = 'match (o:Org)-[:MEMBER_OF]-(s:Staff)-[:PUBLISHED]-(p:Publication) return o.code as o,count(distinct(p)) as c;'
oDic = {}
for res in session.run(com):
oDic[res['o']] = res['c']
return oDic
def staff_per_org():
session = neo4j_functions.connect()
com = ' match (o:Org)-[:MEMBER_OF]-(s:Staff) return o.code as o,count(s) as c;'
sDic = {}
for res in session.run(com):
sDic[res['o']] = res['c']
return sDic
def pubs_per_person():
session = neo4j_functions.connect()
com = 'match (s:Staff)-[:PUBLISHED]-(p:Publication) return s.person_id as p,count(p) as c;'
pDic = {}
for res in session.run(com):
pDic[res['p']] = res['c']
return pDic
def add_metrics_to_graph():
session = neo4j_functions.connect()
with open('output/pubMetrics.txt', 'rb') as p:
for line in p:
line = line.rstrip()
pid, cf, ld, ts_flesch_reading_ease, ts_smog_index, ts_flesch_kincaid_grade, ts_coleman_liau_index, ts_automated_readability_index, ts_dale_chall_readability_score, ts_difficult_words, ts_linsear_write_formula, ts_gunning_fog, ts_text_standard = line.split(
'\t')
com = "match (p:Publication) where p.pub_id = "+str(pid)+" set p.cf = "+str("%.4f" % cf)+",p.ld ="+str("%.4f" % ld)+"" \
",p.ts_flesch_reading_ease="+str("%.4f" % ts_flesch_reading_ease)+",p.ts_smog_index="+str("%.4f" % ts_smog_index)+",p.ts_flesch_kincaid_grade="+str("%.4f" % ts_flesch_kincaid_grade)+"" \
",p.ts_coleman_liau_index="+str("%.4f" % ts_coleman_liau_index)+",p.ts_automated_readability_index="+str("%.4f" % ts_automated_readability_index)+"" \
",p.ts_dale_chall_readability_score="+str("%.4f" % ts_dale_chall_readability_score)+",p.ts_difficult_words="+str("%.4f" % ts_difficult_words)+"" \
",p.ts_linsear_write_formula="+str("%.4f" % ts_linsear_write_formula)+",p.ts_gunning_fog="+str("%.4f" % ts_gunning_fog)+",p.ts_text_standard='"+str(ts_text_standard)+"';"
#print com
session.run(com)
session.close()
def get_metrics():
#check for previous
pubDic = {}
if os.path.exists('output/pubMetrics.txt'):
with open('output/pubMetrics.txt', 'rb') as p:
for line in p:
if len(line.split('\t')) == 13:
pid, cf, ld, ts_flesch_reading_ease, ts_smog_index, ts_flesch_kincaid_grade, ts_coleman_liau_index, ts_automated_readability_index, ts_dale_chall_readability_score, ts_difficult_words, ts_linsear_write_formula, ts_gunning_fog, ts_text_standard = line.split(
'\t')
pubDic[int(pid)] = ''
#print pubDic
print "Found " + str(len(pubDic)) + " entries already processed"
session = neo4j_functions.connect()
com = "match (p:Publication) return distinct p.abstract as a, p.pub_id as pid;"
counter = 0
o = open('output/pubMetrics.txt', 'a')
for res in session.run(com):
if counter % 1000 == 0:
print counter, strftime("%Y-%m-%d %H:%M:%S", gmtime())
#session.close()
#session = neo4j_functions.connect()
counter += 1
abs = res['a']
pid = res['pid']
#print pid
if pid not in pubDic:
#content fraction
if len(abs) > 0:
cf = nltk_functions.content_fraction(abs)
ld = nltk_functions.lexical_diversity(abs)
ts = nltk_functions.run_textstat(abs)
ts_flesch_reading_ease, ts_smog_index, ts_flesch_kincaid_grade, ts_coleman_liau_index, ts_automated_readability_index, ts_dale_chall_readability_score, ts_difficult_words, ts_linsear_write_formula, ts_gunning_fog, ts_text_standard = ts
#print pid, cf, ld, ts_flesch_reading_ease
o.write(
str(pid) + "\t" + str("%.4f" % cf) + "\t" +
str("%.4f" % ld) + "\t" +
str("%.4f" % ts_flesch_reading_ease) + ""
"\t" + str("%.4f" % ts_smog_index) + "\t" +
str("%.4f" % ts_flesch_kincaid_grade) + "\t" +
str("%.4f" % ts_coleman_liau_index) + ""
"\t" + str("%.4f" % ts_automated_readability_index) +
"\t" + str("%.4f" % ts_dale_chall_readability_score) + ""
"\t" + str("%.4f" % ts_difficult_words) + "\t" +
str("%.4f" % ts_linsear_write_formula) + "\t" +
str("%.4f" % ts_gunning_fog) + ""
"\t" + str(ts_text_standard) + "\n")
def distance_metrics(year):
session = neo4j_functions.connect()
#staff
print "Creating staff distance data"
com = "match (p1:Staff)-[e1:ENRICHED]->(s), (p2:Staff)-[e2:ENRICHED]->(s) where id(p1) < id(p2) and e1.year = " + str(
year
) + " and e2.year = " + str(
year
) + " with sqrt(sum((-log10(e1.cpval) - -log10(e2.cpval))^2)) as euc,p1,p2 merge (p1)-[d:DISTANCE]-(p2) set d.euclidean_" + str(
year) + " = euc return count(euc);"
print com
for res in session.run(com):
print res
#orgs
print "Create org distance data"
com = "match (p1:Org)-[e1:ENRICHED]->(s), (p2:Org)-[e2:ENRICHED]->(s) where id(p1) < id(p2) and e1.year = " + str(
year
) + " and e2.year = " + str(
year
) + " with sqrt(sum((-log10(e1.cpval) - -log10(e2.cpval))^2)) as euc,p1,p2 merge (p1)-[d:DISTANCE]-(p2) set d.euclidean_" + str(
year) + " = euc return count(euc);"
print com
for res in session.run(com):
print res
def copy_graph_to_mysql():
session = neo4j_functions.connect()
cnx = connect()
curA = cnx.cursor(buffered=True)
#add orgs
print "Adding orgs"
neo4j_com = "match (o:Org) return o.code as c, o.short_name as s, o.full_name as f, o.url as u, o.type as t order by o.code;"
mysql_com = (
"INSERT IGNORE INTO browser_org (code, short_name, full_name, url, type) "
"VALUES (%s, %s, %s, %s, %s)")
for res in session.run(neo4j_com):
code = res['c']
short = res['s']
full = res['f']
url = res['u']
if len(url) == 0:
url = 'n/a'
type = res['t']
#print code,short,full,url,type
try:
curA.execute(mysql_com, (code, short, full, url, type))
except mysql.connector.Error as err:
print("failed to insert values %s, %s")
#person table
#name,user_name,institute,position,sex
print "Adding people"
neo4j_com = "match (o:Org)--(s:Staff) return s.published_name as n, s.person_id as pid, o.code as s;"
mysql_com = (
"INSERT IGNORE INTO browser_person (org_id, user_name, name, position, sex) "
"VALUES ((SELECT id from browser_org where code = %s), %s, %s, %s, %s)"
)
for res in session.run(neo4j_com):
name = res['n']
#print name
user_name = res['pid']
org = res['s']
pos = 'n/a'
sex = 'n/a'
curA.execute(mysql_com, (org, user_name, name, pos, sex))
#concepts
#name,type
print "Adding concepts"
neo4j_com = "match (c:Concept) return c.name as n, c.type as t"
mysql_com = ("INSERT IGNORE INTO browser_concept (name, type) "
"VALUES (%s, %s)")
for res in session.run(neo4j_com):
name = res['n']
#print name
type = res['t']
curA.execute(mysql_com, (name, type))
#enrichments
#person_id,globalCount,cpval,concept_id,globalTotal,localCount,localTotal,year
print "Adding people-concepts"
neo4j_com = "match (s:Staff)-[e]-(c:Concept) return s.person_id as pid,c.name, e.localCount,e.localTotal,e.globalCount,e.globalTotal,e.year,e.cpval;"
mysql_com = (
"INSERT IGNORE INTO browser_enrichedp (person_id,concept_id,localCount,localTotal,globalCount,globalTotal,year,cpval) "
"VALUES ((SELECT id from browser_person where user_name = %s), (SELECT id from browser_concept where name = %s), %s, %s, %s, %s, %s, %s )"
)
for res in session.run(neo4j_com):
pid = res['pid']
cName = res['c.name']
localCount = res['e.localCount']
localTotal = res['e.localTotal']
globalCount = res['e.globalCount']
globalTotal = res['e.globalTotal']
year = res['e.year']
cpval = res['e.cpval']
#print pid,cName
curA.execute(mysql_com, (pid, cName, localCount, localTotal,
globalCount, globalTotal, year, cpval))
print "Adding org-concepts"
neo4j_com = "match (o:Org)-[e]-(c:Concept) return o.code as c,c.name, e.localCount,e.localTotal,e.globalCount,e.globalTotal,e.year,e.cpval;"
mysql_com = (
"INSERT IGNORE INTO browser_enrichedo (org_id,concept_id,localCount,localTotal,globalCount,globalTotal,year,cpval) "
"VALUES ((SELECT id from browser_org where code = %s), (SELECT id from browser_concept where name = %s), %s, %s, %s, %s, %s, %s )"
)
for res in session.run(neo4j_com):
pid = res['c']
cName = res['c.name']
localCount = res['e.localCount']
localTotal = res['e.localTotal']
globalCount = res['e.globalCount']
globalTotal = res['e.globalTotal']
year = res['e.year']
cpval = res['e.cpval']
#print pid, cName
curA.execute(mysql_com, (pid, cName, localCount, localTotal,
globalCount, globalTotal, year, cpval))
cnx.close()
def add_enriched_to_graph(year):
print "Adding enrichment data to graph..."
session = neo4j_functions.connect()
for concept_type in ['type', 'bigram', 'trigram']:
print "Adding data for data type '" + concept_type + "'"
counter = 0
num_lines = sum(1 for line in open(outDir + '/person_' + concept_type +
'_enriched_' + str(year) + '.txt'))
#create concept nodes and index
with open(
outDir + '/person_' + concept_type + '_enriched_' + str(year) +
'.txt', 'rb') as p:
next(p)
for line in p:
line = line.rstrip()
#print line
if counter % 10000 == 0:
print str(counter) + '/' + str(num_lines)
session.close()
session = neo4j_functions.connect()
counter += 1
name, type, a1, a2, b1, b2, odds, pval, cor_pval = line.split(
'\t')
type = type.replace("'", "\\'")
person_id = name.split(':')[1]
com = "MERGE (c:Concept {name:'" + type + "', type:'" + concept_type + "'})"
session.run(com)
i = "CREATE index on :Concept(name);"
session.run(i)
#i="CREATE index on :Concept(type);"
#session.run(i)
counter = 0
#create person-concept relationships
print "Creating person-concept relationships..."
with open(
outDir + '/person_' + concept_type + '_enriched_' + str(year) +
'.txt', 'rb') as p:
next(p)
for line in p:
line = line.rstrip()
#print line
if counter % 10000 == 0:
print str(counter) + '/' + str(num_lines)
session.close()
session = neo4j_functions.connect()
counter += 1
name, type, a1, a2, b1, b2, odds, pval, cor_pval = line.split(
'\t')
if float(cor_pval) > 0:
cor_pval = ("%03.02e" % float(cor_pval))
else:
cor_pval = 0.0
type = type.replace("'", "\\'")
person_id = name.split(':')[1]
com = "MATCH (s:Staff {person_id: "+person_id+"}) " \
"MATCH (c:Concept {name:'"+type+"',type:'"+concept_type+"'}) " \
"MERGE (s)-[:ENRICHED{type:'pure-comp',year:"+str(year)+",localCount:"+str(a1)+"," \
"localTotal:"+str(a2)+",globalCount:"+str(b1)+",globalTotal:"+str(b2)+",cpval:"+str(cor_pval)+"}]-(c);"
session.run(com)
counter = 0
#create org-concept relationships
print "Creating org-concept relationships..."
with open(
outDir + '/org_' + concept_type + '_enriched_' + str(year) +
'.txt', 'rb') as p:
next(p)
for line in p:
if counter % 10000 == 0:
print counter
session.close()
session = neo4j_functions.connect()
counter += 1
line = line.rstrip()
name, type, a1, a2, b1, b2, odds, pval, cor_pval = line.split(
'\t')
if float(cor_pval) > 0:
cor_pval = ("%03.02e" % float(cor_pval))
else:
cor_pval = 0.0
type = type.replace("'", "\\'")
code = name.split(':')[1]
com = "MATCH (o:Org {code: '"+code+"'}) " \
"MATCH (c:Concept {name:'"+type+"',type:'"+concept_type+"'}) " \
"MERGE (o)-[:ENRICHED{type:'pure-comp',year:"+str(year)+",localCount:"+str(a1)+"," \
"localTotal:"+str(a2)+",globalCount:"+str(b1)+",globalTotal:"+str(b2)+",cpval:"+str(cor_pval)+"}]-(c);"
#print com
session.run(com)
session.close()
def background_frequencies(year):
print "Getting background frequencies"
session = neo4j_functions.connect()
o1 = open(outDir + '/background_type_frequencies_' + str(year) + '.txt',
'w')
o1.write('type\tcount\n')
o2 = open(outDir + '/background_bigram_frequencies_' + str(year) + '.txt',
'w')
o2.write('bigram\tcount\n')
o3 = open(outDir + '/background_trigram_frequencies_' + str(year) + '.txt',
'w')
o3.write('trigram\tcount\n')
com = "match (s:Staff)--(p:Publication) where p.pub_year <= " + str(
year
) + " return distinct p.abstract as a, p.title as t, p.pub_id as pid;"
print com
typeDic = {}
bigramDic = {}
trigramDic = {}
pubConceptDic = defaultdict(dict)
counter = 0
ignoreList = ignore_tokens()
for res in session.run(com):
if counter % 1000 == 0:
print counter
counter += 1
if len(res['a']) > config.minAbsLength:
#combine title and abstract
title_abs = res['t'] + res['a']
#abs = res['a']
pid = res['pid']
types = nltk_functions.tokenise_and_lemm(title_abs)
#bigrams,trigrams = nltk_functions.bigrams_and_trigrams(abs)
bigrams, trigrams = nltk_functions.bigrams_and_trigrams(
" ".join(types))
for s in set(types):
if s in ignoreList:
pass
else:
#add to pub-concept dic
pubConceptDic[pid][s] = 'type'
if s in typeDic:
typeDic[s] += 1
else:
typeDic[s] = 1
for s in set(bigrams):
if len(set(s).intersection(ignoreList)) == 0:
ss = s[0] + ':' + s[1]
#add to pub-concept dic
pubConceptDic[pid][ss] = 'bigram'
if ss in bigramDic:
bigramDic[ss] += 1
else:
bigramDic[ss] = 1
for s in set(trigrams):
if len(set(s).intersection(ignoreList)) == 0:
ss = s[0] + ':' + s[1] + ':' + s[2]
#add to pub-concept dic
pubConceptDic[pid][ss] = 'trigram'
if ss in trigramDic:
trigramDic[ss] += 1
else:
trigramDic[ss] = 1
for t in sorted(typeDic, key=typeDic.get, reverse=True):
o1.write(t + '\t' + str(typeDic[t]) + '\n')
for t in sorted(bigramDic, key=bigramDic.get, reverse=True):
o2.write(t + '\t' + str(bigramDic[t]) + '\n')
for t in sorted(trigramDic, key=trigramDic.get, reverse=True):
o3.write(t + '\t' + str(trigramDic[t]) + '\n')
o4 = open(outDir + '/pub_concept_' + str(year) + '.txt', 'w')
l = len(pubConceptDic)
counter = 0
for p in pubConceptDic:
if counter % 1000 == 0:
print counter, l
counter += 1
for c in pubConceptDic[p]:
o4.write(str(p) + '\t' + c + "\t" + pubConceptDic[p][c] + '\n')
session.close()
def org_frequencies(year):
print "Getting org frequencies"
session = neo4j_functions.connect()
o1 = open(outDir + '/org_type_frequencies_' + str(year) + '.txt', 'w')
o1.write('org\ttype\tcount\n')
o2 = open(outDir + '/org_bigram_frequencies_' + str(year) + '.txt', 'w')
o2.write('org\tbigram\tcount\n')
o3 = open(outDir + '/org_trigram_frequencies_' + str(year) + '.txt', 'w')
o3.write('org\ttrigram\tcount\n')
com = "match (o:Org)--(s:Staff)--(p:Publication) where p.pub_year <= " + str(
year
) + " return distinct o.short_name as o1, o.code as o2, p.pub_id as pid;"
typeDic = defaultdict(dict)
bigramDic = defaultdict(dict)
trigramDic = defaultdict(dict)
counter = 0
ignoreList = ignore_tokens()
#read in type freqs for filtering
typeConceptDic = typeFreqs(year)
#get pubConceptDic
pubConceptDic = readPubConcepts(year)
for res in session.run(com):
if counter % 1000 == 0:
print counter
counter += 1
org = res['o1'] + ':' + res['o2']
pid = str(res['pid'])
for s in pubConceptDic[pid]:
if s in ignoreList:
pass
else:
if pubConceptDic[pid][s] == 'type':
if int(typeConceptDic[s]) > freqCutoff:
#print pid,s,pubConceptDic[pid][s],typeConceptDic[s]
pass
else:
if s in typeDic[org]:
typeDic[org][s] += 1
else:
typeDic[org][s] = 1
elif pubConceptDic[pid][s] == 'bigram':
if s in bigramDic[org]:
bigramDic[org][s] += 1
else:
bigramDic[org][s] = 1
elif pubConceptDic[pid][s] == 'trigram':
if s in trigramDic[org]:
trigramDic[org][s] += 1
else:
trigramDic[org][s] = 1
for p in sorted(typeDic, key=typeDic.get, reverse=True):
for t in sorted(typeDic[p], key=typeDic[p].get, reverse=True):
o1.write(str(p) + '\t' + t + '\t' + str(typeDic[p][t]) + '\n')
for p in sorted(bigramDic, key=bigramDic.get, reverse=True):
for t in sorted(bigramDic[p], key=bigramDic[p].get, reverse=True):
o2.write(str(p) + '\t' + t + '\t' + str(bigramDic[p][t]) + '\n')
for p in sorted(trigramDic, key=trigramDic.get, reverse=True):
for t in sorted(trigramDic[p], key=trigramDic[p].get, reverse=True):
o3.write(str(p) + '\t' + t + '\t' + str(trigramDic[p][t]) + '\n')
session.close()
def person_frequencies(year):
print "Getting person frequencies"
session = neo4j_functions.connect()
o1 = open(outDir + '/person_type_frequencies_' + str(year) + '.txt', 'w')
o1.write('person\ttype\tcount\n')
o2 = open(outDir + '/person_bigram_frequencies_' + str(year) + '.txt', 'w')
o2.write('person\tbigram\tcount\n')
o3 = open(outDir + '/person_trigram_frequencies_' + str(year) + '.txt',
'w')
o3.write('person\ttrigram\tcount\n')
#com = "match (s:Staff)--(p:Publication) where p.pub_id = 2942913 return distinct s.published_name as p1, s.person_id as p2, p.pub_id as pid;"
com = "match (s:Staff)--(p:Publication) where p.pub_year <= " + str(
year
) + " return distinct s.published_name as p1, s.person_id as p2, p.pub_id as pid;"
typeDic = defaultdict(dict)
bigramDic = defaultdict(dict)
trigramDic = defaultdict(dict)
ignoreList = ignore_tokens()
counter = 0
#read in type freqs for filtering
typeConceptDic = typeFreqs(year)
#get pubConceptDic
pubConceptDic = readPubConcepts(year)
for res in session.run(com):
if counter % 1000 == 0:
print counter
counter += 1
person_name = res['p1']
person_id = res['p2']
pid = str(res['pid'])
person = person_name + ':' + str(person_id)
#print res['pid']
for s in pubConceptDic[pid]:
if s in ignoreList:
pass
else:
if pubConceptDic[pid][s] == 'type':
if int(typeConceptDic[s]) > freqCutoff:
#print pid,s,pubConceptDic[pid][s],typeConceptDic[s]
pass
else:
if s in typeDic[person]:
typeDic[person][s] += 1
else:
typeDic[person][s] = 1
elif pubConceptDic[pid][s] == 'bigram':
if s in bigramDic[person]:
bigramDic[person][s] += 1
else:
bigramDic[person][s] = 1
elif pubConceptDic[pid][s] == 'trigram':
if s in trigramDic[person]:
trigramDic[person][s] += 1
else:
trigramDic[person][s] = 1
for p in sorted(typeDic, key=typeDic.get, reverse=True):
for t in sorted(typeDic[p], key=typeDic[p].get, reverse=True):
o1.write(str(p) + '\t' + t + '\t' + str(typeDic[p][t]) + '\n')
for p in sorted(bigramDic, key=bigramDic.get, reverse=True):
for t in sorted(bigramDic[p], key=bigramDic[p].get, reverse=True):
o2.write(str(p) + '\t' + t + '\t' + str(bigramDic[p][t]) + '\n')
for p in sorted(trigramDic, key=trigramDic.get, reverse=True):
for t in sorted(trigramDic[p], key=trigramDic[p].get, reverse=True):
o3.write(str(p) + '\t' + t + '\t' + str(trigramDic[p][t]) + '\n')
session.close()
#import matplotlib
#matplotlib.style.use('ggplot')
from csv import reader
import pandas as pd
import numpy as np
from nltk.corpus import wordnet as wordnet
from textstat.textstat import textstat
from nltk.stem.snowball import SnowballStemmer
stemmer = SnowballStemmer("english")
from nltk.stem import WordNetLemmatizer
wordnet_lemmatizer = WordNetLemmatizer()
bigram_measures = nltk.collocations.BigramAssocMeasures()
trigram_measures = nltk.collocations.TrigramAssocMeasures()
session = neo4j_functions.connect()
#analyse output
# awk -F '\t' '$2>50' output/nltk_counts.txt | sort -t$'\t' -k4 -nr | less
#http://www.nltk.org/book/ch01.html
def content_fraction(text):
stopwords = nltk.corpus.stopwords.words('english')
content = [w for w in text if w.lower() not in stopwords]
return len(content) / len(text)
def filter_stopwords_and_length(text, wordLength):
stopwords = nltk.corpus.stopwords.words('english')
