使用gensim进行LDA主题建模的python IndexError

Question

Another thread has a similar question to mine but leaves out reproducible code.

The goal with the script in question is to create a process that is as memory efficient as possible. So I tried to write a the class corpus() to take advantage of gensims' capabilities. However, I am running into an IndexError that I'm not sure how to resolve when creating lda = models.ldamodel.LdaModel(corpus_tfidf, id2word=checker.dictionary, num_topics=int(options.number_of_topics)) .

The documents that I am using are the same as used in the gensim tutorial, which I placed into tutorial_example.txt:

$ cat tutorial_example.txt 
Human machine interface for lab abc computer applications
A survey of user opinion of computer system response time
The EPS user interface management system
System and human system engineering testing of EPS
Relation of user perceived response time to error measurement
The generation of random binary unordered trees
The intersection graph of paths in trees
Graph minors IV Widths of trees and well quasi ordering
Graph minors A survey

Error received

$./gensim_topic_modeling.py -mn2 -w'english' -l1 tutorial_example.txt 
Traceback (most recent call last):
  File "./gensim_topic_modeling.py", line 98, in <module>
    lda = models.ldamodel.LdaModel(corpus_tfidf, id2word=checker.dictionary, num_topics=int(options.number_of_topics))
  File "/Users/me/anaconda/lib/python2.7/site-packages/gensim/models/ldamodel.py", line 306, in __init__
    self.update(corpus)
  File "/Users/me/anaconda/lib/python2.7/site-packages/gensim/models/ldamodel.py", line 543, in update
    self.log_perplexity(chunk, total_docs=lencorpus)
  File "/Users/me/anaconda/lib/python2.7/site-packages/gensim/models/ldamodel.py", line 454, in log_perplexity
    perwordbound = self.bound(chunk, subsample_ratio=subsample_ratio) / (subsample_ratio * corpus_words)
  File "/Users/me/anaconda/lib/python2.7/site-packages/gensim/models/ldamodel.py", line 630, in bound
    gammad, _ = self.inference([doc])
  File "/Users/me/anaconda/lib/python2.7/site-packages/gensim/models/ldamodel.py", line 366, in inference
    expElogbetad = self.expElogbeta[:, ids]
IndexError: index 7 is out of bounds for axis 1 with size 7

Below is the gensim_topic_modeling.py script:

##gensim_topic_modeling.py

#!/usr/bin/env python
# -*- coding: UTF-8 -*-
import sys
import re
import codecs
import logging
import fileinput
from operator import *
from itertools import *
from sklearn.cluster import KMeans
from gensim import corpora, models, similarities, matutils
import argparse
from nltk.corpus import stopwords

reload(sys)
sys.stdout = codecs.getwriter('utf-8')(sys.stdout)
sys.stdin = codecs.getreader('utf-8')(sys.stdin)


##defs

def stop_word_gen():
    nltk_langs=['danish', 'dutch', 'english', 'french', 'german', 'italian','norwegian', 'portuguese', 'russian', 'spanish', 'swedish']
    stoplist = []
    for lang in options.stop_langs.split(","):
        if lang not in nltk_langs:
            sys.stderr.write('\n'+"Language {0} not supported".format(lang)+'\n')
            continue
        stoplist.extend(stopwords.words(lang))
    return stoplist


def clean_texts(texts):
    # remove tokens that appear only once
    all_tokens = sum(texts, [])
    tokens_once = set(word for word in set(all_tokens) if all_tokens.count(word) == 1)
    return [[word for word in text if word not in tokens_once] for text in texts]

##class

class corpus(object):
    """sparse vector matrix and dictionary"""
    def __iter__(self):
        first=True
        for line in fileinput.FileInput(options.input, openhook=fileinput.hook_encoded("utf-8")):
            # assume there's one document per line; tokenizer option determines how to split
            if options.space_tokenizer:
                rl = re.compile('\s+', re.UNICODE).split(unicode(line,'utf-8'))
            else:
                rl = re.compile('\W+', re.UNICODE).split(tagRE.sub(' ',line)) 
            # create dictionary
            tokens=[token.strip().lower() for token in rl if token != '' and token.strip().lower() not in stoplist]
            if first:
                first=False
                self.dictionary=corpora.Dictionary([tokens])
            else:
                self.dictionary.add_documents([tokens])
                self.dictionary.compactify
            yield self.dictionary.doc2bow(tokens)


##main 

if __name__ == '__main__':
    ##parser
    parser = argparse.ArgumentParser(
                description="Topic model from a column of text.  Each line is a document in the corpus")
    parser.add_argument("input", metavar="args")
    parser.add_argument("-l", "--document-frequency-limit", dest="doc_freq_limit", default=1,
                help="Remove all tokens less than or equal to limit (default 1)")
    parser.add_argument("-m", "--create-model", dest="create_model", default=False, action="store_true",
                help="Create and save a model from existing dictionary and input corpus.")
    parser.add_argument("-n", "--number-of-topics", dest="number_of_topics", default=2,
                help="Number of topics (default 2)")
    parser.add_argument("-t", "--space-tokenizer", dest="space_tokenizer", default=False, action="store_true", 
                help="Use alternate whitespace tokenizer")
    parser.add_argument("-w", "--stop-word-languages", dest="stop_langs", default="danish,dutch,english,french,german,italian,norwegian,portuguese,russian,spanish,swedish",
                help="Desired languages for stopword lists")
    options = parser.parse_args()

    ##globals

    stoplist=set(stop_word_gen())  
    tagRE = re.compile(r'<.*?>', re.UNICODE)    # Remove xml/html tags
    logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO, filename="topic-modeling-log")
    logr = logging.getLogger("topic_model")
    logr.info("#"*15 + " started " + "#"*15)

    ##instance of class 

    checker=corpus()
    logr.info("#"*15 + " SPARSE MATRIX (pre-filter)" + "#"*15)

    ##view sparse matrix and dictionary

    for vector in checker: 
        logr.info(vector)
    logr.info("#"*15 + " DICTIONARY (pre-filter)" + "#"*15)
    logr.info(checker.dictionary)
    logr.info(checker.dictionary.token2id)
    #filter
    checker.dictionary.filter_extremes(no_below=int(options.doc_freq_limit)+1)
    logr.info("#"*15 + " DICTIONARY (post-filter)" + "#"*15)
    logr.info(checker.dictionary)
    logr.info(checker.dictionary.token2id)

    ##Create lda model

    if options.create_model:     
        tfidf = models.TfidfModel(checker,normalize=False)
        print tfidf
        logr.info("#"*15 + " corpus_tfidf " + "#"*15)
        corpus_tfidf = tfidf[checker]
        logr.info("#"*15 + " lda " + "#"*15)
        lda = models.ldamodel.LdaModel(corpus_tfidf, id2word=checker.dictionary, num_topics=int(options.number_of_topics))
        logr.info("#"*15 + " corpus_lda " + "#"*15)
        corpus_lda = lda[corpus_tfidf] 

        ##Evaluate topics based on threshold

        scores = list(chain(*[[score for topic,score in topic] \
                      for topic in [doc for doc in corpus_lda]]))
        threshold = sum(scores)/len(scores)
        print "threshold:",threshold
        print
        cluster1 = [j for i,j in zip(corpus_lda,documents) if i[0][1] > threshold]
        cluster2 = [j for i,j in zip(corpus_lda,documents) if i[1][1] > threshold]
        cluster3 = [j for i,j in zip(corpus_lda,documents) if i[2][1] > threshold]

The resulting topic-modeling-log file is below. Thanks in advance for any help!

topic-modeling-log

2014-05-25 02:58:50,482 : INFO : ############### started ###############
2014-05-25 02:58:50,483 : INFO : ############### SPARSE MATRIX (pre-filter)###############
2014-05-25 02:58:50,483 : INFO : adding document #0 to Dictionary(0 unique tokens: [])
2014-05-25 02:58:50,483 : INFO : built Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...) from 1 documents (total 7 corpus positions)
2014-05-25 02:58:50,483 : INFO : [(0, 1), (1, 1), (2, 1), (3, 1), (4, 1), (5, 1), (6, 1)]
2014-05-25 02:58:50,483 : INFO : adding document #0 to Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...)
2014-05-25 02:58:50,483 : INFO : built Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...) from 2 documents (total 14 corpus positions)
2014-05-25 02:58:50,483 : INFO : [(2, 1), (7, 1), (8, 1), (9, 1), (10, 1), (11, 1), (12, 1)]
2014-05-25 02:58:50,483 : INFO : adding document #0 to Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...)
2014-05-25 02:58:50,484 : INFO : built Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...) from 3 documents (total 19 corpus positions)
2014-05-25 02:58:50,484 : INFO : [(4, 1), (10, 1), (12, 1), (13, 1), (14, 1)]
2014-05-25 02:58:50,484 : INFO : adding document #0 to Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...)
2014-05-25 02:58:50,484 : INFO : built Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...) from 4 documents (total 25 corpus positions)
2014-05-25 02:58:50,484 : INFO : [(3, 1), (10, 2), (13, 1), (15, 1), (16, 1)]
2014-05-25 02:58:50,484 : INFO : adding document #0 to Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...)
2014-05-25 02:58:50,484 : INFO : built Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...) from 5 documents (total 32 corpus positions)
2014-05-25 02:58:50,484 : INFO : [(8, 1), (11, 1), (12, 1), (17, 1), (18, 1), (19, 1), (20, 1)]
2014-05-25 02:58:50,484 : INFO : adding document #0 to Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...)
2014-05-25 02:58:50,484 : INFO : built Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 6 documents (total 37 corpus positions)
2014-05-25 02:58:50,484 : INFO : [(21, 1), (22, 1), (23, 1), (24, 1), (25, 1)]
2014-05-25 02:58:50,485 : INFO : adding document #0 to Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,485 : INFO : built Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 7 documents (total 41 corpus positions)
2014-05-25 02:58:50,485 : INFO : [(24, 1), (26, 1), (27, 1), (28, 1)]
2014-05-25 02:58:50,485 : INFO : adding document #0 to Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,485 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 8 documents (total 49 corpus positions)
2014-05-25 02:58:50,485 : INFO : [(24, 1), (26, 1), (29, 1), (30, 1), (31, 1), (32, 1), (33, 1), (34, 1)]
2014-05-25 02:58:50,485 : INFO : adding document #0 to Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...)
2014-05-25 02:58:50,485 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 9 documents (total 52 corpus positions)
2014-05-25 02:58:50,485 : INFO : [(9, 1), (26, 1), (30, 1)]
2014-05-25 02:58:50,485 : INFO : ############### DICTIONARY (pre-filter)###############
2014-05-25 02:58:50,485 : INFO : Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...)
2014-05-25 02:58:50,485 : INFO : {'minors': 30, 'generation': 22, 'testing': 16, 'iv': 29, 'engineering': 15, 'computer': 2, 'relation': 20, 'human': 3, 'measurement': 18, 'unordered': 25, 'binary': 21, 'abc': 0, 'ordering': 31, 'graph': 26, 'system': 10, 'machine': 6, 'quasi': 32, 'random': 23, 'paths': 28, 'error': 17, 'trees': 24, 'lab': 5, 'applications': 1, 'management': 14, 'user': 12, 'interface': 4, 'intersection': 27, 'response': 8, 'perceived': 19, 'widths': 34, 'well': 33, 'eps': 13, 'survey': 9, 'time': 11, 'opinion': 7}
2014-05-25 02:58:50,486 : INFO : keeping 12 tokens which were in no less than 2 and no more than 4 (=50.0%) documents
2014-05-25 02:58:50,486 : INFO : resulting dictionary: Dictionary(12 unique tokens: ['minors', 'graph', 'system', 'trees', 'eps']...)
2014-05-25 02:58:50,486 : INFO : ############### DICTIONARY (post-filter)###############
2014-05-25 02:58:50,486 : INFO : Dictionary(12 unique tokens: ['minors', 'graph', 'system', 'trees', 'eps']...)
2014-05-25 02:58:50,486 : INFO : {'minors': 0, 'graph': 1, 'system': 2, 'trees': 3, 'eps': 4, 'computer': 5, 'survey': 6, 'user': 7, 'human': 8, 'time': 9, 'interface': 10, 'response': 11}
2014-05-25 02:58:50,486 : INFO : collecting document frequencies
2014-05-25 02:58:50,486 : INFO : adding document #0 to Dictionary(0 unique tokens: [])
2014-05-25 02:58:50,486 : INFO : built Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...) from 1 documents (total 7 corpus positions)
2014-05-25 02:58:50,486 : INFO : PROGRESS: processing document #0
2014-05-25 02:58:50,486 : INFO : adding document #0 to Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...)
2014-05-25 02:58:50,486 : INFO : built Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...) from 2 documents (total 14 corpus positions)
2014-05-25 02:58:50,486 : INFO : adding document #0 to Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...)
2014-05-25 02:58:50,487 : INFO : built Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...) from 3 documents (total 19 corpus positions)
2014-05-25 02:58:50,487 : INFO : adding document #0 to Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...)
2014-05-25 02:58:50,487 : INFO : built Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...) from 4 documents (total 25 corpus positions)
2014-05-25 02:58:50,487 : INFO : adding document #0 to Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...)
2014-05-25 02:58:50,487 : INFO : built Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...) from 5 documents (total 32 corpus positions)
2014-05-25 02:58:50,487 : INFO : adding document #0 to Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...)
2014-05-25 02:58:50,487 : INFO : built Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 6 documents (total 37 corpus positions)
2014-05-25 02:58:50,487 : INFO : adding document #0 to Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,487 : INFO : built Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 7 documents (total 41 corpus positions)
2014-05-25 02:58:50,488 : INFO : adding document #0 to Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,488 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 8 documents (total 49 corpus positions)
2014-05-25 02:58:50,488 : INFO : adding document #0 to Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...)
2014-05-25 02:58:50,488 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 9 documents (total 52 corpus positions)
2014-05-25 02:58:50,488 : INFO : calculating IDF weights for 9 documents and 34 features (51 matrix non-zeros)
2014-05-25 02:58:50,488 : INFO : ############### corpus_tfidf ###############
2014-05-25 02:58:50,488 : INFO : adding document #0 to Dictionary(0 unique tokens: [])
2014-05-25 02:58:50,488 : INFO : built Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...) from 1 documents (total 7 corpus positions)
2014-05-25 02:58:50,489 : INFO : ############### lda ###############
2014-05-25 02:58:50,489 : INFO : using symmetric alpha at 0.5
2014-05-25 02:58:50,489 : INFO : using serial LDA version on this node
2014-05-25 02:58:50,489 : WARNING : input corpus stream has no len(); counting documents
2014-05-25 02:58:50,489 : INFO : adding document #0 to Dictionary(0 unique tokens: [])
2014-05-25 02:58:50,489 : INFO : built Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...) from 1 documents (total 7 corpus positions)
2014-05-25 02:58:50,489 : INFO : adding document #0 to Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...)
2014-05-25 02:58:50,489 : INFO : built Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...) from 2 documents (total 14 corpus positions)
2014-05-25 02:58:50,489 : INFO : adding document #0 to Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...)
2014-05-25 02:58:50,490 : INFO : built Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...) from 3 documents (total 19 corpus positions)
2014-05-25 02:58:50,490 : INFO : adding document #0 to Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...)
2014-05-25 02:58:50,490 : INFO : built Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...) from 4 documents (total 25 corpus positions)
2014-05-25 02:58:50,490 : INFO : adding document #0 to Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...)
2014-05-25 02:58:50,490 : INFO : built Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...) from 5 documents (total 32 corpus positions)
2014-05-25 02:58:50,490 : INFO : adding document #0 to Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...)
2014-05-25 02:58:50,490 : INFO : built Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 6 documents (total 37 corpus positions)
2014-05-25 02:58:50,490 : INFO : adding document #0 to Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,490 : INFO : built Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 7 documents (total 41 corpus positions)
2014-05-25 02:58:50,491 : INFO : adding document #0 to Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,491 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 8 documents (total 49 corpus positions)
2014-05-25 02:58:50,491 : INFO : adding document #0 to Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...)
2014-05-25 02:58:50,491 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 9 documents (total 52 corpus positions)
2014-05-25 02:58:50,491 : INFO : running online LDA training, 2 topics, 1 passes over the supplied corpus of 9 documents, updating model once every 9 documents, evaluating perplexity every 9 documents, iterating 50 with a convergence threshold of 0
2014-05-25 02:58:50,491 : WARNING : too few updates, training might not converge; consider increasing the number of passes or iterations to improve accuracy
2014-05-25 02:58:50,491 : INFO : adding document #0 to Dictionary(0 unique tokens: [])
2014-05-25 02:58:50,491 : INFO : built Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...) from 1 documents (total 7 corpus positions)
2014-05-25 02:58:50,492 : INFO : adding document #0 to Dictionary(7 unique tokens: ['abc', 'lab', 'machine', 'applications', 'computer']...)
2014-05-25 02:58:50,492 : INFO : built Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...) from 2 documents (total 14 corpus positions)
2014-05-25 02:58:50,492 : INFO : adding document #0 to Dictionary(13 unique tokens: ['abc', 'system', 'lab', 'machine', 'applications']...)
2014-05-25 02:58:50,492 : INFO : built Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...) from 3 documents (total 19 corpus positions)
2014-05-25 02:58:50,492 : INFO : adding document #0 to Dictionary(15 unique tokens: ['abc', 'management', 'system', 'lab', 'eps']...)
2014-05-25 02:58:50,492 : INFO : built Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...) from 4 documents (total 25 corpus positions)
2014-05-25 02:58:50,492 : INFO : adding document #0 to Dictionary(17 unique tokens: ['abc', 'testing', 'management', 'system', 'lab']...)
2014-05-25 02:58:50,492 : INFO : built Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...) from 5 documents (total 32 corpus positions)
2014-05-25 02:58:50,493 : INFO : adding document #0 to Dictionary(21 unique tokens: ['measurement', 'perceived', 'abc', 'testing', 'management']...)
2014-05-25 02:58:50,493 : INFO : built Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 6 documents (total 37 corpus positions)
2014-05-25 02:58:50,493 : INFO : adding document #0 to Dictionary(26 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,493 : INFO : built Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...) from 7 documents (total 41 corpus positions)
2014-05-25 02:58:50,493 : INFO : adding document #0 to Dictionary(29 unique tokens: ['generation', 'testing', 'engineering', 'computer', 'relation']...)
2014-05-25 02:58:50,493 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 8 documents (total 49 corpus positions)
2014-05-25 02:58:50,493 : INFO : adding document #0 to Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...)
2014-05-25 02:58:50,493 : INFO : built Dictionary(35 unique tokens: ['minors', 'generation', 'testing', 'iv', 'engineering']...) from 9 documents (total 52 corpus positions)

Answer 1

This is caused by using a corpus and dictionary that don't have the same id-to-word mapping. It can happen if you prune your dictionary and call dictionary.compactify() at the wrong time.

A simple example will make it clear. Let's make a dictionary:

from gensim.corpora.dictionary import Dictionary
documents = [
    ['here', 'is', 'one', 'document'],
    ['here', 'is', 'another', 'document'],
]
dictionary = Dictionary()
dictionary.add_documents(documents)

This dictionary now has entries for these words and maps them to integer id's. It's useful to turn documents into vectors of (id, count) tuples (which we'd want to do before passing them into a model):

vectorized_corpus = [dictionary.doc2bow(doc) for doc in corpus]

Sometimes you'll want to alter your dictionary. For example, you might want to remove very rare, or very common words:

dictionary.filter_extremes(no_below=2, no_above=0.5, keep_n=100000)
dictionary.compactify()

Removing words creates gaps in the dictionary, but calling dictionary.compactify() re-assigns ids to fill in the gaps. But that means our vectorized_corpus from above doesn't use the same id's as the dictionary any more, and if we pass them into a model, we'll get an IndexError .

Solution : make your vector representation using the dictionary after making changes and calling dictionary.compactify() !