from sklearn.manifold import MDS
import urllib.request
from gensim.models import Word2Vec

sonalist=urllib.request.urlopen("https://minitorn.tlu.ee/~jaagup/oma/too/23/05/word2vec/sonaindeksid.txt").read().decode("utf-8").split("\n")
sonad={rida.split(",")[1]:int(rida.split(",")[0]) for rida in sonalist if rida}
ryhmalist=urllib.request.urlopen("https://minitorn.tlu.ee/~jaagup/oma/too/23/05/word2vec/vastused/jaotus1000.txt").read().decode("utf-8").split("\n")
ryhmad=[ryhmalist[i*3+1].split() for i in range(1000)]
mudel=Word2Vec.load("https://minitorn.tlu.ee/~jaagup/oma/too/23/05/word2vec/lemmad1.model")
print("loetud")
sonakohad=mudel.wv.get_normed_vectors()
keskmised=[]
for ryhm in ryhmad:
   kohad=[sonakohad[sonad[sona]].tolist() for sona in ryhm if sona in sonad]
   keskmine=[sum(vaartused)/len(vaartused) for vaartused in zip(*kohad)]
   keskmised.append(keskmine)
ryhmakeskmed=MDS().fit_transform(keskmised) 
print(ryhmakeskmed) 
vastus=[str(rida[0])+","+str(rida[1]) for rida in ryhmakeskmed]
f2=open("paigutus4_ryhmakeskmed.txt", "w", encoding="utf-8")
print("\n".join(vastus), file=f2)
f2.close()