CaliGraph/caliGraph.py

798 lines
25 KiB
Python
Executable File

#!./.venv/bin/python3.9
import os
import json
import math
import copy
import random
import numpy as np
from scipy.stats import norm
import matplotlib.pyplot as plt
import networkx as nx
from pyvis.network import Network
def getAllAuthors(books):
authors = set()
for book in books:
for author in getAuthors(book):
authors.add(author)
return list(authors)
def getAuthors(book):
return book['authors'].split(' & ')
def getRecommenders(book):
for tag in book['tags']:
if tag.find(" Recommendation") != -1:
yield tag.replace(" Recommendation", "")
elif tag.find("s Literature Club") != -1:
yield tag.replace("s Literature Club", "")
def getTags(book):
for tag in book['tags']:
if tag.find(" Recommendation") == -1 and tag.find("s Literature Club") == -1 and tag.find(" Top ") == -1:
yield tag
def getAllRecommenders(books):
recs = set()
for book in books:
for rec in getRecommenders(book):
recs.add(rec)
return list(recs)
def getTopLists(book):
lists = set()
for tag in book['tags']:
if tag.find(" Top ") != -1:
lists.add(tag.split(" Top ")[0])
return list(lists)
def getAllTopLists(books):
tops = set()
for book in books:
for top in getTopLists(book):
tops.add(top)
return list(tops)
def getAllSeries(books):
series = set()
for book in books:
if 'series' in book:
series.add(book['series'])
return list(series)
def getAllTags(books):
tags = set()
for book in books:
for tag in getTags(book):
tags.add(tag)
return list(tags)
def getTopListWeight(book, topList):
minScope = 100000
for tag in book['tags']:
if tag.find(topList+" Top ") != -1:
scope = int(tag.split(" Top ")[1])
minScope = min(minScope, scope)
if minScope == 100000:
raise Exception("You stupid?")
if minScope == 10:
return 1
elif minScope == 25:
return 0.85
elif minScope == 100:
return 0.5
return 50 / minScope
def removeRead(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
if node['rating'] != None:
G.remove_node(n)
def removeUnread(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
if node['rating'] == None:
G.remove_node(n)
def removePriv(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
if 'priv' in node['tags']:
G.remove_node(n)
def removeDangling(G, alsoBooks=False):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] != 'book' or alsoBooks:
if not len(G.adj[n]):
G.remove_node(n)
def removeEdge(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] != 'book':
if len(G.adj[n]) < 2:
G.remove_node(n)
def removeBad(G, threshold, groups=['book', 'topList', 'recommender', 'author', 'series', 'tag']):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] in groups:
if 'score' in node and (node['score'] == None or node['score'] < threshold):
G.remove_node(n)
def removeKeepBest(G, num, maxDistForRead=1):
bestlist = []
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
if 'score' in node and node['score'] != None:
bestlist.append(node)
bestlist.sort(key=lambda node: node['score'], reverse=True)
bestlist = bestlist[:num]
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book' and node not in bestlist or 'score' in node and node['score'] == None:
if not 'rating' in node or node['rating'] == None or node['rating'] < bestlist[-1]['score']-maxDistForRead:
G.remove_node(n)
def removeTags(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'tag':
G.remove_node(n)
def pruneTags(G, minCons=2):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'tag':
foundCon = 0
for book in G.adj[n]:
for con in G.adj[book]:
conType = G.nodes[con]['t']
if conType not in ['topList']:
if conType in ['recommender']:
foundCon += 0.5
elif conType in ['tag', 'series']:
foundCon += 0.25
else:
foundCon += 1
if foundCon > minCons:
G.remove_node(n)
def pruneRecommenderCons(G, maxCons=5):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'recommender':
if len(G.adj[n]) > maxCons:
bestlist = []
for m in list(G.adj[n]):
book = G.nodes[m]
if book['t'] == 'book':
if 'score' in book and book['score'] != None:
bestlist.append(book)
bestlist.sort(key=lambda node: node['score'], reverse=True)
bestlist = bestlist[:maxCons]
for m in list(G.adj[n]):
book = G.nodes[m]
if book['t'] == 'book' and book not in bestlist or 'score' in book and book['score'] == None:
if not 'rating' in book or book['rating'] == None:
foundCon = 0
for con in G.adj[m]:
if G.nodes[con]['t'] not in ['topList']:
foundCon += 1
if foundCon < 2:
G.remove_node(m)
def pruneAuthorCons(G, maxCons=3):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'author':
if len(G.adj[n]) > maxCons:
bestlist = []
for m in list(G.adj[n]):
book = G.nodes[m]
if book['t'] == 'book':
if 'score' in book and book['score'] != None:
bestlist.append(book)
bestlist.sort(key=lambda node: node['score'], reverse=True)
bestlist = bestlist[:maxCons]
for m in list(G.adj[n]):
book = G.nodes[m]
if book['t'] == 'book' and book not in bestlist or 'score' in book and book['score'] == None:
if not 'rating' in book or book['rating'] == None:
foundCon = 0
for con in G.adj[m]:
if G.nodes[con]['t'] not in ['topList']:
foundCon += 1
if foundCon < 2:
G.remove_node(m)
def removeHighSpanTags(G, maxCons=5):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'tag':
if len(G.adj[n]) > maxCons:
G.remove_node(n)
def removeTopLists(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'topList':
G.remove_node(n)
def removeRestOfSeries(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'series':
seriesState = 0
for adj in G.adj[n]:
adjNode = G.nodes[adj]
if adjNode['rating'] != None:
seriesState = max(seriesState, int(
adjNode['series_index']))
for adj in list(G.adj[n]):
adjNode = G.nodes[adj]
if adjNode['series_index'] > seriesState + 1.0001:
G.remove_node(adj)
def removeUnusedRecommenders(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'recommender':
for adj in G.adj[n]:
adjNode = G.nodes[adj]
if adjNode['t']=='book' and 'score' in adjNode:
break
else: # No unrated recommendation
G.remove_node(n)
def removeUselessReadBooks(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
for adj in G.adj[n]:
foundUnread = True
adjNode = G.nodes[adj]
for cousin in G.adj[adj]:
cousinNode = G.nodes[cousin]
if cousinNode['t']=='book' and 'score' in cousinNode:
break
else: # No unrated book here
foundUnread = False
if foundUnread:
break
else: # No unrated book in cousins
G.remove_node(n)
def removeUselessTags(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'tag':
for adj in G.adj[n]:
foundUnread = True
adjNode = G.nodes[adj]
if adjNode['t']=='book' and 'score' in adjNode:
break
else: # No unrated book here
foundUnread = False
if foundUnread:
break
else: # No unrated book in cousins
G.remove_node(n)
def scoreOpinions(G, globMu, globStd, errorFac=0):
for n in list(G.nodes):
node = G.nodes[n]
feedbacks = []
if node['t'] in ['topList', 'recommender', 'author', 'series', 'tag']:
adjacens = list(G.adj[n].keys())
for adj in adjacens:
adjNode = G.nodes[adj]
if adjNode['rating'] != None:
feedbacks.append(adjNode['rating'])
if len(feedbacks):
node['mean'], node['std'] = norm.fit(feedbacks)
node['se'] = globStd / math.sqrt(len(feedbacks))
ratio = len(feedbacks) / len(adjacens)
node['score'] = node['mean'] + errorFac * \
node['se']*(6/7 + (1-ratio)/7) + 0.01 * \
(node['t'] == 'recommender') \
- 0.5 / len(feedbacks)**2
node['feedbacks'] = feedbacks
else:
node['score'] = None
def scoreUnread(G, globMu, globStd, errorFac=-0.6):
for n in list(G.nodes):
feedbacks = [globMu]
weights = [getWeightForType('mu')]
node = G.nodes[n]
if node['t'] == 'book':
if node['rating'] == None:
adjacens = list(G.adj[n].keys())
for adj in adjacens:
adjNode = G.nodes[adj]
if 'score' in adjNode and adjNode['score'] != None:
w = getWeightForType(adjNode['t'], G[n][adj]['weight'] if 'weight' in G[n][adj] else 1)
for fb in adjNode['feedbacks']:
feedbacks.append(fb)
weights.append(w)
if len(feedbacks):
node['meanUnweighted'], node['std'] = norm.fit(feedbacks)
feedbacks.append(node['std'])
weights.append(getWeightForType('sigma'))
node['mean'] = sum([fb*w for fb, w in zip(feedbacks, weights)])/len(feedbacks)
node['se'] = globStd / math.sqrt(len(feedbacks))
node['score'] = node['mean'] + errorFac*node['se']
else:
node['score'] = globMu + errorFac*globStd + len(feedbacks)*0.0000000001
if 'series' in node:
if node['series_index'] == 1.0:
node['score'] += 0.000000001
# TODO: Make this neural and train it
def getWeightForType(nodeType, edgeWeight=1):
global weights
w = weights[nodeType]
if nodeType == 'topList':
return edgeWeight*w
else:
return w
def printBestList(G, num=-1):
bestlist = []
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] == 'book':
if 'score' in node and node['score'] != None:
bestlist.append(node)
bestlist.sort(key=lambda node: node['score'], reverse=True)
for i, book in enumerate(bestlist):
print("["+str(i+1).zfill(int((math.log10(num) if num!=-1 else 3)+1))+"] "+book['title'] +
" ("+" & ".join(book['authors'])+"): {:.5f}".format(book['score']))
if num!=-1 and i == num-1:
break
def readColor(book):
if 'rating' in book:
return 'green'
else:
return 'gray'
def loadBooksFromDB():
return json.loads(os.popen("calibredb list --for-machine -f all").read())
def buildBookGraph(books):
G = nx.Graph()
# Books
for book in books:
if 'rating' in book:
rating = book['rating']
else:
rating = None
if 'comments' in book:
desc = '' # book['comments']
else:
desc = ''
if 'series' in book:
series = book['series']
series_index = book['series_index']
else:
series = None
series_index = None
G.add_node(book['id'], t='book', label=book['title'], title=book['title'], shape='image', image=book['cover'], rating=rating,
tags=book['tags'], desc=desc, isbn=book['isbn'], files=book['formats'], authors=getAuthors(book), series=series, series_index=series_index)
return G
def graphAddAuthors(G, books):
for author in getAllAuthors(books):
G.add_node('a/'+author, color='green', t='author', label=author)
for book in books:
for author in getAuthors(book):
G.add_edge('a/'+author, book['id'], color=readColor(book))
return G
def graphAddRecommenders(G, books):
for rec in getAllRecommenders(books):
G.add_node('r/'+rec, color='orange', t='recommender', label=rec)
for book in books:
for rec in getRecommenders(book):
G.add_edge('r/'+rec, book['id'], color=readColor(book))
return G
def graphAddTopLists(G, books):
for tl in getAllTopLists(books):
G.add_node('t/'+tl, color='yellow', t='topList', label=tl)
for book in books:
for top in getTopLists(book):
G.add_edge('t/'+top, book['id'], weight=getTopListWeight(
book, top), color=readColor(book))
return G
def graphAddSeries(G, books):
for series in getAllSeries(books):
G.add_node('s/'+series, color='red', t='series', label=series, shape='triangle')
for book in books:
if 'series' in book:
G.add_edge('s/'+book['series'], book['id'], color=readColor(book))
return G
def graphAddTags(G, books):
for tag in getAllTags(books):
G.add_node('t/'+tag, color='lightGray', t='tag', label=tag, shape='box')
for book in books:
for tag in getTags(book):
G.add_edge('t/'+tag, book['id'], color=readColor(book))
return G
def calcRecDist(G, books):
globRatings = []
for book in books:
if G.nodes[book['id']]['rating'] != None:
globRatings.append(G.nodes[book['id']]['rating'])
return norm.fit(globRatings)
def scaleBooksByRating(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] not in []:
if 'rating' in node and node['rating'] != None:
node['value'] = 20 + 5 * int(node['rating'])
else:
if 'score' in node and node['score'] != None:
node['value'] = 20 + 5 * int(node['score'])
else:
node['value'] = 15
def scaleOpinionsByRating(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] in ['topList', 'recommender', 'author', 'series']:
if 'score' in node and node['score'] != None:
node['value'] = 20 + 5 * int(node['score'])
else:
node['value'] = 20
def addScoreToLabels(G):
for n in list(G.nodes):
node = G.nodes[n]
if node['t'] not in ['tag']:
if 'rating' in node and node['rating'] != None:
node['label'] += " ("+str(node['rating'])+")"
else:
if 'score' in node and node['score'] != None:
node['label'] += " (~{:.2f}".format(node['score'])+")"
else:
node['label'] += " (~0)"
def genAndShowHTML(G, showButtons=False):
net = Network('1080px', '1920px')
if showButtons:
net.show_buttons(filter_=['configure', 'layout',
'interaction', 'physics', 'edges'])
net.from_nx(G)
net.show('nx.html')
def buildFullGraph():
books = loadBooksFromDB()
G = buildBookGraph(books)
graphAddAuthors(G, books)
graphAddRecommenders(G, books)
graphAddTopLists(G, books)
graphAddSeries(G, books)
graphAddTags(G, books)
return G, books
def genScores(G, books):
globMu, globStd = calcRecDist(G, books)
scoreOpinions(G, globMu, globStd)
scoreUnread(G, globMu, globStd)
return globMu, globStd
def recommendNBooks(G, mu, std, n, removeTopListsB=True, removeUselessRecommenders=True):
removeRestOfSeries(G)
removeBad(G, mu-std*2-1)
removeKeepBest(G, int(n*2) + 5, maxDistForRead=1.5*2)
removeEdge(G)
removeHighSpanTags(G, 9)
removeDangling(G, alsoBooks=False)
pruneTags(G, 6*2)
removeBad(G, mu, groups=['book'])
removeUselessReadBooks(G)
pruneTags(G, 4.25*2)
pruneRecommenderCons(G, int(n/7)+1)
pruneAuthorCons(G, int(n/15))
removeUselessTags(G)
if removeTopListsB:
removeTopLists(G)
removeDangling(G, alsoBooks=True)
removeKeepBest(G, n, maxDistForRead=0.75*2)
removeEdge(G)
removeDangling(G, alsoBooks=True)
if removeUselessRecommenders:
removeUnusedRecommenders(G)
removeDangling(G, alsoBooks=True)
scaleBooksByRating(G)
scaleOpinionsByRating(G)
addScoreToLabels(G)
def fullGraph(G, removeTopLists=True):
removeEdge(G)
removeHighSpanTags(G, 7)
removeDangling(G, alsoBooks=False)
if removeTopLists:
removeTopLists(G)
pruneTags(G, 3)
removeDangling(G, alsoBooks=True)
scaleBooksByRating(G)
scaleOpinionsByRating(G)
addScoreToLabels(G)
def readBooksAnalysis(G, minRating=0, showAllTags=True, removeUnconnected=False, removeTopListsB=True):
removeUnread(G)
removeBad(G, minRating)
if not showAllTags:
removeEdge(G)
removeHighSpanTags(G, 15)
removeDangling(G, alsoBooks=removeUnconnected)
if removeTopListsB:
removeTopLists(G)
pruneTags(G, 8)
scaleBooksByRating(G)
scaleOpinionsByRating(G)
addScoreToLabels(G)
def analyze(G, type_name, name, dist=2.7):
from fuzzywuzzy import fuzz
type_ident = type_name[0]
full_name = type_ident + "/" + name
bestRatio, match, n = 0, None, 0
for ni in list(G.nodes):
node = G.nodes[ni]
if node['t'] == type_name or type_name=="any":
if name==node['label'] or full_name==node['label']:
match, n = node, ni
break
ratio = fuzz.ratio(node['label'], name)
if ratio > bestRatio:
bestRatio, match, n = ratio, node, ni
if bestRatio < 70:
print("Best Match: "+match['label'])
menge = set()
waveFlow(G, match, n, dist, menge)
for n in list(G.nodes):
if n not in menge:
G.remove_node(n)
removeHighSpanTags(G, 12)
if dist > 1:
removeDangling(G, True)
scaleBooksByRating(G)
scaleOpinionsByRating(G)
#match['value'] = 100
if not 'shape' in match:
match['shape'] = 'star'
addScoreToLabels(G)
match['label'] = "*"+match['label']+"*"
def waveFlow(G, node, n, dist, menge, firstEdge=False):
if dist <= 0:
return
dist -= 1
if menge==set():
firstEdge=True
if node['t'] in ['topList']:
if firstEdge:
menge.add(n)
return
menge.add(n)
if node['t'] in ['tag']:
if firstEdge:
dist-=0.1
else:
return
bestlist = []
keeplist = []
for m in list(G.adj[n]):
book = G.nodes[m]
if book['t'] not in ['NOTHING']:
if 'score' in book and book['score'] != None:
bestlist.append(book)
elif 'rating' in book and book['rating'] != None:
keeplist.append(book)
else:
book['score'] = 0
bestlist.append(book)
bestlist.sort(key=lambda node: node['score'], reverse=True)
toKeep = min(int(dist*10), math.ceil(len(bestlist) * dist - len(keeplist)*0.5))
if toKeep <= 0:
keeplist.sort(key=lambda node: node['rating'], reverse=True)
keeplist = keeplist[:min(int(dist*10), int(len(keeplist) * dist))]
bestlist = []
else:
bestlist = bestlist[:toKeep]
for m in list(G.adj[n]):
node = G.nodes[m]
if node in bestlist or node in keeplist:
waveFlow(G, node, m, dist, menge, firstEdge=firstEdge)
def evaluateFitness():
G, books = buildFullGraph()
ratedBooks = [n for n in list(G.nodes) if 'rating' in G.nodes[n] and G.nodes[n]['rating'] != None]
errSq = []
for m in ratedBooks:
rating = G.nodes[m]['rating']
G.nodes[m]['rating'] = None
mu, std = genScores(G, books)
errSq.append((rating - G.nodes[m]['score'])**2)
G.nodes[m]['rating'] = rating
return sum(errSq) / len(errSq)
def train(gamma = 0.1):
global weights
bestWeights = copy.copy(weights)
best_mse = evaluateFitness()
w = list(weights.keys())
attr = random.choice(w)
delta = gamma * (-0.5 + (0.75 + 0.25*random.random()))
while True:
print({'mse': best_mse, 'w': weights, 'gamma': gamma})
weights = copy.copy(bestWeights)
weights[attr] += delta
mse = evaluateFitness()
if mse < best_mse: # got better
saveWeights(weights)
gamma *= 1.1
bestWeights = copy.copy(weights)
best_mse = mse
delta *= 2
if random.random() < 0.10:
attr = random.choice(w)
else:
weights = copy.copy(bestWeights)
gamma *= 0.8
attr = random.choice(w)
delta = gamma * (-0.5 + (0.75 + 0.25*random.random()))
def saveWeights(weights):
with open('neuralWeights.json', 'w') as f:
f.write(json.dumps(weights))
def loadWeights():
with open('neuralWeights.json', 'r') as f:
weights = json.loads(f.read())
return weights
def cliInterface():
import argparse
parser = argparse.ArgumentParser(description='TODO: Write Description.')
parser.add_argument('--keep-priv', action="store_true")
parser.add_argument('--remove-read', action="store_true")
parser.add_argument('--remove-unread', action="store_true")
parser.add_argument('--no-web', action="store_true")
parser.add_argument('--no-list', action="store_true")
parser.add_argument('--remove-edge', action="store_true")
parser.add_argument('--keep-top-lists', action="store_true")
parser.add_argument('--keep-useless-recommenders', action="store_true")
cmds = parser.add_subparsers(required=True, dest='cmd')
p_rec = cmds.add_parser('recommend', description="TODO", aliases=['rec'])
p_rec.add_argument('-n', type=int, default=25, help='number of books to recommend')
p_read = cmds.add_parser('read', description="TODO", aliases=[])
p_read.add_argument('--min-rating', type=int, default=0)
p_read.add_argument('--all-tags', action="store_true")
p_read.add_argument('--only-connected', action="store_true")
p_show = cmds.add_parser('analyze', description="TODO", aliases=[])
p_show.add_argument('type', choices=['any', 'book', 'recommender', 'author', 'series'])
p_show.add_argument('name', type=str)
p_show.add_argument('-d', type=float, default=2.7, help='depth of expansion')
p_train = cmds.add_parser('train', description="TODO", aliases=[])
p_train.add_argument('-g', type=float, default=0.1, help='learning rate gamma')
p_full = cmds.add_parser('full', description="TODO", aliases=[])
args = parser.parse_args()
if args.cmd=="train":
train(args.g)
exit()
G, books = buildFullGraph()
mu, std = genScores(G, books)
if args.cmd=="recommend":
recommendNBooks(G, mu, std, args.n, not args.keep_top_lists, not args.keep_useless_recommenders)
elif args.cmd=="read":
readBooksAnalysis(G, args.min_rating, args.all_tags, args.only_connected, not args.keep_top_lists)
elif args.cmd=="analyze":
analyze(G, args.type, args.name, args.d)
elif args.cmd=="full":
fullGraph(G, not args.keep_top_lists)
else:
raise Exception("Bad")
if not args.keep_priv:
removePriv(G)
if args.remove_read:
removeRead(G)
elif args.remove_unread:
removeUnread(G)
removeDangling(G, alsoBooks=True)
if args.remove_edge:
removeEdge(G)
if not args.no_list:
printBestList(G)
if not args.no_web:
genAndShowHTML(G)
weights = loadWeights()
if __name__ == "__main__":
cliInterface()