Doc2Vec Tutorial on the Lee Dataset


In [1]:
import gensim
import os
import collections
import smart_open
import random

What is it?

Doc2Vec is an NLP tool for representing documents as a vector and is a generalizing of the Word2Vec method. This tutorial will serve as an introduction to Doc2Vec and present ways to train and assess a Doc2Vec model.

Getting Started

To get going, we'll need to have a set of documents to train our doc2vec model. In theory, a document could be anything from a short 140 character tweet, a single paragraph (i.e., journal article abstract), a news article, or a book. In NLP parlance a collection or set of documents is often referred to as a corpus.

For this tutorial, we'll be training our model using the Lee Background Corpus included in gensim. This corpus contains 314 documents selected from the Australian Broadcasting Corporation’s news mail service, which provides text e-mails of headline stories and covers a number of broad topics.

And we'll test our model by eye using the much shorter Lee Corpus which contains 50 documents.


In [2]:
# Set file names for train and test data
test_data_dir = '{}'.format(os.sep).join([gensim.__path__[0], 'test', 'test_data'])
lee_train_file = test_data_dir + os.sep + 'lee_background.cor'
lee_test_file = test_data_dir + os.sep + 'lee.cor'

Define a Function to Read and Preprocess Text

Below, we define a function to open the train/test file (with latin encoding), read the file line-by-line, pre-process each line using a simple gensim pre-processing tool (i.e., tokenize text into individual words, remove punctuation, set to lowercase, etc), and return a list of words. Note that, for a given file (aka corpus), each continuous line constitutes a single document and the length of each line (i.e., document) can vary. Also, to train the model, we'll need to associate a tag/number with each document of the training corpus. In our case, the tag is simply the zero-based line number.


In [3]:
def read_corpus(fname, tokens_only=False):
    with smart_open.smart_open(fname, encoding="iso-8859-1") as f:
        for i, line in enumerate(f):
            if tokens_only:
                yield gensim.utils.simple_preprocess(line)
            else:
                # For training data, add tags
                yield gensim.models.doc2vec.TaggedDocument(gensim.utils.simple_preprocess(line), [i])

In [4]:
train_corpus = list(read_corpus(lee_train_file))
test_corpus = list(read_corpus(lee_test_file, tokens_only=True))

Let's take a look at the training corpus


In [5]:
train_corpus[:2]


Out[5]:
[TaggedDocument(words=[u'hundreds', u'of', u'people', u'have', u'been', u'forced', u'to', u'vacate', u'their', u'homes', u'in', u'the', u'southern', u'highlands', u'of', u'new', u'south', u'wales', u'as', u'strong', u'winds', u'today', u'pushed', u'huge', u'bushfire', u'towards', u'the', u'town', u'of', u'hill', u'top', u'new', u'blaze', u'near', u'goulburn', u'south', u'west', u'of', u'sydney', u'has', u'forced', u'the', u'closure', u'of', u'the', u'hume', u'highway', u'at', u'about', u'pm', u'aedt', u'marked', u'deterioration', u'in', u'the', u'weather', u'as', u'storm', u'cell', u'moved', u'east', u'across', u'the', u'blue', u'mountains', u'forced', u'authorities', u'to', u'make', u'decision', u'to', u'evacuate', u'people', u'from', u'homes', u'in', u'outlying', u'streets', u'at', u'hill', u'top', u'in', u'the', u'new', u'south', u'wales', u'southern', u'highlands', u'an', u'estimated', u'residents', u'have', u'left', u'their', u'homes', u'for', u'nearby', u'mittagong', u'the', u'new', u'south', u'wales', u'rural', u'fire', u'service', u'says', u'the', u'weather', u'conditions', u'which', u'caused', u'the', u'fire', u'to', u'burn', u'in', u'finger', u'formation', u'have', u'now', u'eased', u'and', u'about', u'fire', u'units', u'in', u'and', u'around', u'hill', u'top', u'are', u'optimistic', u'of', u'defending', u'all', u'properties', u'as', u'more', u'than', u'blazes', u'burn', u'on', u'new', u'year', u'eve', u'in', u'new', u'south', u'wales', u'fire', u'crews', u'have', u'been', u'called', u'to', u'new', u'fire', u'at', u'gunning', u'south', u'of', u'goulburn', u'while', u'few', u'details', u'are', u'available', u'at', u'this', u'stage', u'fire', u'authorities', u'says', u'it', u'has', u'closed', u'the', u'hume', u'highway', u'in', u'both', u'directions', u'meanwhile', u'new', u'fire', u'in', u'sydney', u'west', u'is', u'no', u'longer', u'threatening', u'properties', u'in', u'the', u'cranebrook', u'area', u'rain', u'has', u'fallen', u'in', u'some', u'parts', u'of', u'the', u'illawarra', u'sydney', u'the', u'hunter', u'valley', u'and', u'the', u'north', u'coast', u'but', u'the', u'bureau', u'of', u'meteorology', u'claire', u'richards', u'says', u'the', u'rain', u'has', u'done', u'little', u'to', u'ease', u'any', u'of', u'the', u'hundred', u'fires', u'still', u'burning', u'across', u'the', u'state', u'the', u'falls', u'have', u'been', u'quite', u'isolated', u'in', u'those', u'areas', u'and', u'generally', u'the', u'falls', u'have', u'been', u'less', u'than', u'about', u'five', u'millimetres', u'she', u'said', u'in', u'some', u'places', u'really', u'not', u'significant', u'at', u'all', u'less', u'than', u'millimetre', u'so', u'there', u'hasn', u'been', u'much', u'relief', u'as', u'far', u'as', u'rain', u'is', u'concerned', u'in', u'fact', u'they', u've', u'probably', u'hampered', u'the', u'efforts', u'of', u'the', u'firefighters', u'more', u'because', u'of', u'the', u'wind', u'gusts', u'that', u'are', u'associated', u'with', u'those', u'thunderstorms'], tags=[0]),
 TaggedDocument(words=[u'indian', u'security', u'forces', u'have', u'shot', u'dead', u'eight', u'suspected', u'militants', u'in', u'night', u'long', u'encounter', u'in', u'southern', u'kashmir', u'the', u'shootout', u'took', u'place', u'at', u'dora', u'village', u'some', u'kilometers', u'south', u'of', u'the', u'kashmiri', u'summer', u'capital', u'srinagar', u'the', u'deaths', u'came', u'as', u'pakistani', u'police', u'arrested', u'more', u'than', u'two', u'dozen', u'militants', u'from', u'extremist', u'groups', u'accused', u'of', u'staging', u'an', u'attack', u'on', u'india', u'parliament', u'india', u'has', u'accused', u'pakistan', u'based', u'lashkar', u'taiba', u'and', u'jaish', u'mohammad', u'of', u'carrying', u'out', u'the', u'attack', u'on', u'december', u'at', u'the', u'behest', u'of', u'pakistani', u'military', u'intelligence', u'military', u'tensions', u'have', u'soared', u'since', u'the', u'raid', u'with', u'both', u'sides', u'massing', u'troops', u'along', u'their', u'border', u'and', u'trading', u'tit', u'for', u'tat', u'diplomatic', u'sanctions', u'yesterday', u'pakistan', u'announced', u'it', u'had', u'arrested', u'lashkar', u'taiba', u'chief', u'hafiz', u'mohammed', u'saeed', u'police', u'in', u'karachi', u'say', u'it', u'is', u'likely', u'more', u'raids', u'will', u'be', u'launched', u'against', u'the', u'two', u'groups', u'as', u'well', u'as', u'other', u'militant', u'organisations', u'accused', u'of', u'targetting', u'india', u'military', u'tensions', u'between', u'india', u'and', u'pakistan', u'have', u'escalated', u'to', u'level', u'not', u'seen', u'since', u'their', u'war'], tags=[1])]

And the testing corpus looks like this:


In [6]:
print(test_corpus[:2])


[[u'the', u'national', u'executive', u'of', u'the', u'strife', u'torn', u'democrats', u'last', u'night', u'appointed', u'little', u'known', u'west', u'australian', u'senator', u'brian', u'greig', u'as', u'interim', u'leader', u'shock', u'move', u'likely', u'to', u'provoke', u'further', u'conflict', u'between', u'the', u'party', u'senators', u'and', u'its', u'organisation', u'in', u'move', u'to', u'reassert', u'control', u'over', u'the', u'party', u'seven', u'senators', u'the', u'national', u'executive', u'last', u'night', u'rejected', u'aden', u'ridgeway', u'bid', u'to', u'become', u'interim', u'leader', u'in', u'favour', u'of', u'senator', u'greig', u'supporter', u'of', u'deposed', u'leader', u'natasha', u'stott', u'despoja', u'and', u'an', u'outspoken', u'gay', u'rights', u'activist'], [u'cash', u'strapped', u'financial', u'services', u'group', u'amp', u'has', u'shelved', u'million', u'plan', u'to', u'buy', u'shares', u'back', u'from', u'investors', u'and', u'will', u'raise', u'million', u'in', u'fresh', u'capital', u'after', u'profits', u'crashed', u'in', u'the', u'six', u'months', u'to', u'june', u'chief', u'executive', u'paul', u'batchelor', u'said', u'the', u'result', u'was', u'solid', u'in', u'what', u'he', u'described', u'as', u'the', u'worst', u'conditions', u'for', u'stock', u'markets', u'in', u'years', u'amp', u'half', u'year', u'profit', u'sank', u'per', u'cent', u'to', u'million', u'or', u'share', u'as', u'australia', u'largest', u'investor', u'and', u'fund', u'manager', u'failed', u'to', u'hit', u'projected', u'per', u'cent', u'earnings', u'growth', u'targets', u'and', u'was', u'battered', u'by', u'falling', u'returns', u'on', u'share', u'markets']]

Notice that the testing corpus is just a list of lists and does not contain any tags.

Training the Model

Instantiate a Doc2Vec Object

Now, we'll instantiate a Doc2Vec model with a vector size with 50 words and iterating over the training corpus 55 times. We set the minimum word count to 2 in order to give higher frequency words more weighting. Model accuracy can be improved by increasing the number of iterations but this generally increases the training time. Small datasets with short documents, like this one, can benefit from more training passes.


In [7]:
model = gensim.models.doc2vec.Doc2Vec(size=50, min_count=2, iter=55)

Build a Vocabulary


In [8]:
model.build_vocab(train_corpus)

Essentially, the vocabulary is a dictionary (accessible via model.wv.vocab) of all of the unique words extracted from the training corpus along with the count (e.g., model.wv.vocab['penalty'].count for counts for the word penalty).

Time to Train

If the BLAS library is being used, this should take no more than 3 seconds. If the BLAS library is not being used, this should take no more than 2 minutes, so use BLAS if you value your time.


In [9]:
%time model.train(train_corpus, total_examples=model.corpus_count, epochs=model.iter)


CPU times: user 6.34 s, sys: 109 ms, total: 6.45 s
Wall time: 2.58 s
Out[9]:
2347665

Inferring a Vector

One important thing to note is that you can now infer a vector for any piece of text without having to re-train the model by passing a list of words to the model.infer_vector function. This vector can then be compared with other vectors via cosine similarity.


In [10]:
model.infer_vector(['only', 'you', 'can', 'prevent', 'forrest', 'fires'])


Out[10]:
array([ 0.02664499,  0.00475204, -0.03981256,  0.03796276, -0.03206162,
        0.10963056, -0.04897128,  0.00151982, -0.03258783,  0.04711508,
       -0.00667155, -0.08523653, -0.02975186,  0.00166316,  0.01915652,
       -0.03415785, -0.05794788,  0.05110953,  0.01623618, -0.00512495,
       -0.06385455, -0.0151557 ,  0.00365376,  0.03015811,  0.0229462 ,
        0.03176891,  0.01117626, -0.00743352,  0.02030453, -0.05072152,
       -0.00498496,  0.00151227,  0.06122205, -0.01811385, -0.01715777,
        0.04883198,  0.03925886, -0.03568915,  0.00805744,  0.01654406,
       -0.05160677,  0.0119908 , -0.01527433,  0.02209963, -0.10316766,
       -0.01069367, -0.02432527,  0.00761799,  0.02763799, -0.04288232], dtype=float32)

Assessing Model

To assess our new model, we'll first infer new vectors for each document of the training corpus, compare the inferred vectors with the training corpus, and then returning the rank of the document based on self-similarity. Basically, we're pretending as if the training corpus is some new unseen data and then seeing how they compare with the trained model. The expectation is that we've likely overfit our model (i.e., all of the ranks will be less than 2) and so we should be able to find similar documents very easily. Additionally, we'll keep track of the second ranks for a comparison of less similar documents.


In [11]:
ranks = []
second_ranks = []
for doc_id in range(len(train_corpus)):
    inferred_vector = model.infer_vector(train_corpus[doc_id].words)
    sims = model.docvecs.most_similar([inferred_vector], topn=len(model.docvecs))
    rank = [docid for docid, sim in sims].index(doc_id)
    ranks.append(rank)
    
    second_ranks.append(sims[1])

Let's count how each document ranks with respect to the training corpus


In [12]:
collections.Counter(ranks)  # Results vary due to random seeding and very small corpus


Out[12]:
Counter({0: 289, 1: 11})

Basically, greater than 95% of the inferred documents are found to be most similar to itself and about 5% of the time it is mistakenly most similar to another document. the checking of an inferred-vector against a training-vector is a sort of 'sanity check' as to whether the model is behaving in a usefully consistent manner, though not a real 'accuracy' value.

This is great and not entirely surprising. We can take a look at an example:


In [15]:
print('Document ({}): «{}»\n'.format(doc_id, ' '.join(train_corpus[doc_id].words)))
print(u'SIMILAR/DISSIMILAR DOCS PER MODEL %s:\n' % model)
for label, index in [('MOST', 0), ('MEDIAN', len(sims)//2), ('LEAST', len(sims) - 1)]:
    print(u'%s %s: «%s»\n' % (label, sims[index], ' '.join(train_corpus[sims[index][0]].words)))


Document (299): «australia will take on france in the doubles rubber of the davis cup tennis final today with the tie levelled at wayne arthurs and todd woodbridge are scheduled to lead australia in the doubles against cedric pioline and fabrice santoro however changes can be made to the line up up to an hour before the match and australian team captain john fitzgerald suggested he might do just that we ll make team appraisal of the whole situation go over the pros and cons and make decision french team captain guy forget says he will not make changes but does not know what to expect from australia todd is the best doubles player in the world right now so expect him to play he said would probably use wayne arthurs but don know what to expect really pat rafter salvaged australia davis cup campaign yesterday with win in the second singles match rafter overcame an arm injury to defeat french number one sebastien grosjean in three sets the australian says he is happy with his form it not very pretty tennis there isn too many consistent bounces you are playing like said bit of classic old grass court rafter said rafter levelled the score after lleyton hewitt shock five set loss to nicholas escude in the first singles rubber but rafter says he felt no added pressure after hewitt defeat knew had good team to back me up even if we were down he said knew could win on the last day know the boys can win doubles so even if we were down still feel we are good enough team to win and vice versa they are good enough team to beat us as well»

SIMILAR/DISSIMILAR DOCS PER MODEL Doc2Vec(dm/m,d50,n5,w5,mc2,s0.001,t3):

MOST (83, 0.9967287182807922): «the opposition leader simon crean says child abuse scandal in brisbane has damaged the office of the governor general and its incumbent dr peter hollingworth child advocates have called on dr hollingworth to step down as governor general saying he did not do enough to prevent abuse of children in an anglican school when he was archbishop of brisbane mr crean says he is not calling on dr hollingworth to resign but he says there are still unanswered questions think it has tarnished the office of the governor general the fact that it took so long for this statement to come out he said many people have been calling for it me included think if we are to avoid further damage to the office we need to clear it up completely brisbane lord mayor says the governor general explanation of his handling of child sex abuse allegations at queensland school raises more questions than it answers jim soorley who is former catholic priest says the explanation does not wash within the christian tradition bishops are regarded as shepherds he said it very clear that he was not good shepherd and there are serious consequences for that think his actions are not the actions of good shepherd and think there are still questions to be answered»

MEDIAN (252, 0.9926513433456421): «the labor party is set to have wide ranging review of its structures with frontbencher martin ferguson pushing for the process the new labor leader simon crean is taking set of proposals to next thursday national executive meeting mr ferguson wants the meeting to call review he says suggestions for party changes such as the call by frontbencher joel fitzgibbon for the scrapping of new south wales rule forcing labor members to belong to union should be dealt with internally perhaps the time has come for us to actually sign up to federal executive process which actually enables debate to go forward in highly constructive way rather than individual proposals being put out there without an end game in sight mr ferguson said he says he is happy for the process to include looking at abandoning the rule but says scrapping the rule would only be minor factor in the party self examination for long time ve believed there is requirement for the labor party to actually have hard look at whether or not an archaic formula of union representation is the key to our future»

LEAST (250, 0.9795532822608948): «israel launched massive air raids across the west bank and gaza tuesday piling pressure on yasser arafat with rocket strike on police post next to his offices after prime minister ariel sharon branded his administration sponsor of terrorism israeli warplanes launched series of strikes on gaza city while apache helicopters fired rockets on palestinian security offices in khan yunis in the southern gaza strip and on the west bank towns of salfit and tulkarem they also fired missiles on security post just metres from mr arafat offices in ramallah but the palestinian leader who was in his office at the time was unhurt but two policemen were slightly wounded officials said israeli army spokesman brigadier general ron kitrey said mr arafat was not targeted two people were killed in the gaza strikes and around injured half of them schoolboys palestinian hospital officials said the attacks came as israel foreign minister shimon peres said he did not believe israeli forces would take direct action against the palestinian leader the strikes also came day after mr sharon furious that mr arafat had not stopped hardline islamic groups who killed two dozen israelis in devastating suicide attacks at the weekend ordered his forces to blast symbols of mr arafat power gunships destroyed mr arafat three helicopters in gaza city while bulldozers ploughed up the runway at gaza international airport used by mr arafat for his frequent travels abroad palestinian officials called mr sharon campaign an attempt to topple mr arafat and destroy his self rule palestinian authority mr arafat told cnn television that mr sharon was trying to torpedo his own crackdown on terrorism with the airstrikes he doesn want me to succeed and for this he is escalating his military activities against our towns our cities our establishments the palestinian leader said french foreign minister hubert vedrine accused israel of conducting deliberate policy aimed at eliminating mr arafat arafat has been weakened by the harassment of the israeli army and as result people are using his weakness as an argument to say that since he can not re establish order in his own camp he should in some way be eliminated however britain prime minister tony blair and us president george bush expressed sympathy with israel and called on all sides to do anything they can to stabilise the situation mr sharon hard words and air strikes opened major divisions in his cross party government with left wing mr peres denouncing what he called bid during monday emergency cabinet meeting to cause the downfall of the palestinian authority the region had been braced for huge israeli retaliation after three palestinian suicide bombers from the hardline islamic movement hamas killed people on saturday and sunday in the suicide attacks in jerusalem and haifa mr sharon made national address after blasting gaza city and jenin in the west bank on monday accusing mr arafat of having chosen the path of terrorism and being the greatest obstacle to peace and stability in the middle east mr peres said the move by mr sharon dominant right wingers in effect means israeli policy is based purely on force with no political hope public radio said mr peres had called all the ministers from his labour party for special meeting wednesday to discuss the fallout of the strikes and mr sharon accusation that mr arafat was responsible for everything that has happened here chief palestinian negotiator saeb erakat speaking after mr sharon speech monday evening said the words amounted to declaration of war he called on the united states and europe to rein in mr sharon and dispatch international observers to oversee the spiralling conflict»

Notice above that the most similar document is has a similarity score of ~80% (or higher). However, the similarity score for the second ranked documents should be significantly lower (assuming the documents are in fact different) and the reasoning becomes obvious when we examine the text itself


In [14]:
# Pick a random document from the test corpus and infer a vector from the model
doc_id = random.randint(0, len(train_corpus))

# Compare and print the most/median/least similar documents from the train corpus
print('Train Document ({}): «{}»\n'.format(doc_id, ' '.join(train_corpus[doc_id].words)))
sim_id = second_ranks[doc_id]
print('Similar Document {}: «{}»\n'.format(sim_id, ' '.join(train_corpus[sim_id[0]].words)))


Train Document (193): «new study shows that nearly one third of the aboriginal and torres strait islander population in australia have been arrested in the past five years the study conducted by the australian national university for the new south wales bureau of crime statistics is the first to compare the arrest rates of the aboriginal and non aboriginal population it finds that unemployment alcohol and assault rates were the main causes study author boyd hunter says policy both on community and government level must deal with these issues if the arrest rate is to be decreased addressing the supply of alcohol in remote communities is seen as the most likely avenue for reducing rates of abuse alcohol abuse and hence reduce arrest rates in those communities he said»

Similar Document (125, 0.5374129414558411): «the united states space shuttle endeavour has touched down at florida kennedy space centre after day mission bringing home crew that had been on the international space station since august the shuttle carrying outgoing space station commander frank culbertson and russian cosmonauts vladimir dezhurov and mikhail tyurin along with four other astronauts landed at pm local time taking over from the trio are russian commander yuri onufrienko and us astronauts carl walz and dan bursch who travelled to the station aboard endeavour on december earlier monday the seven us and russian astronauts on board endeavour woke up to the tune please come home for christmas by the rock group bon jovi on sunday the endeavour crew deployed small satellite called starshine from canister located in the shuttle payload bay more than students from countries will track the satellite as it orbits earth for the next eight months the students will collect information in order to calculate the density of the upper atmosphere nasa said on saturday endeavour undocked from the space station after making last minute maneuver to dodge piece of soviet era space refuse the endeavour mission the th shuttle trip to the international space station brought some three tonnes of equipment and materials for scientific experiments to the station the trip carried out under extremely tight security was the first since the september attacks on the united states»

Testing the Model

Using the same approach above, we'll infer the vector for a randomly chosen test document, and compare the document to our model by eye.


In [15]:
# Pick a random document from the test corpus and infer a vector from the model
doc_id = random.randint(0, len(test_corpus))
inferred_vector = model.infer_vector(test_corpus[doc_id])
sims = model.docvecs.most_similar([inferred_vector], topn=len(model.docvecs))

# Compare and print the most/median/least similar documents from the train corpus
print('Test Document ({}): «{}»\n'.format(doc_id, ' '.join(test_corpus[doc_id])))
print(u'SIMILAR/DISSIMILAR DOCS PER MODEL %s:\n' % model)
for label, index in [('MOST', 0), ('MEDIAN', len(sims)//2), ('LEAST', len(sims) - 1)]:
    print(u'%s %s: «%s»\n' % (label, sims[index], ' '.join(train_corpus[sims[index][0]].words)))


Test Document (25): «an islamic high court in northern nigeria rejected an appeal today by single mother sentenced to be stoned to death for having sex out of wedlock clutching her baby daughter amina lawal burst into tears as the judge delivered the ruling lawal was first sentenced in march after giving birth to daughter more than nine months after divorcing»

SIMILAR/DISSIMILAR DOCS PER MODEL Doc2Vec(dm/s,d50,hs,w8,mc2):

MOST (6, 0.44577494263648987): «the united states team of monica seles and jan michael gambill scored decisive victory over unseeded france in their first hopman cup match at burswood dome in perth the pair runners up in the million dollar mixed teams event last year both won their singles encounters to give the us an unbeatable lead the year old seles currently ranked eighth recovered from shaky start to overpower virginie razzano who is ranked nd seles had to fight hard to get home in straight sets winning in minutes then the year old gambill ranked st wore down determined arnaud clement th to win in minutes the americans are aiming to go one better than last year when they were beaten by swiss pair martina hingis and roger federer in the final of the eight nation contest gambill said the win was great way to start the tennis year got little tentative at the end but it was great start to my year he said arnaud is great scrapper and am delighted to beat him even though am frankly bit out of shape that is one of the reasons am here will be in shape by the end of the tournament just aim to keep improving in the new year and if do think have chance to beat anyone when am playing well gambill was pressed hard by clement before taking the first set in minutes but the american gained the ascendancy in the second set breaking in the third and fifth games seles said she had expected her clash with razzano to be tough she was top junior player in the world so it was no surprise that she fought so well she said seles said she still had the hunger to strive to regain her position at the top of her sport this is why you play she said but want to try not to peak too early this season seles slow into her stride slipped to in her opening set against razzano but recovered quickly claiming the set after snatching four games in row in the second set seles broke her opponent in the opening game and completed victory with relative ease despite razzano tenacious efforts»

MEDIAN (9, 0.07038116455078125): «some roads are closed because of dangerous conditions caused by bushfire smoke motorists are being asked to avoid the hume highway between picton road and the illawarra highway where police have reduced the speed limit from kilometres an hour to in southern sydney picton road is closed between wilton and bulli appin road is closed from appin to bulli tops and all access roads to royal national park are closed motorists are also asked to avoid the illawarra highway between the hume highway and robertson and the great western highway between penrith and springwood because of reduced visibility in north western sydney only local residents are allowed to use wisemans ferry road and upper color road under police escort»

LEAST (180, -0.4328160285949707): «australia has linked million of aid to new agreement with nauru to accept an extra asylum seekers the deal means nauru will take up to asylum seekers under australia pacific solution foreign minister alexander downer signed the understanding today with nauru president rene harris mr downer inspected the nauru camps and says they are are practical and efficient had good look at the sanitation the ablution blocks and thought they were pretty good he said the asylum seekers have various things to do there are volleyball facilities and soccer facilities television is available they can see different channels on tv the catering is good there are three meals day provided»

Wrapping Up

That's it! Doc2Vec is a great way to explore relationships between documents.