How does word2vec create vectors for words? I trained two word2vec models using two different files (from commoncrawl website) but I am getting same word vectors for a given word from both models.
Actually, I have created multiple word2vec models using different text files from the commoncrawl website. Now I want to check which model is better among all. How can select the best model out of all these models and why I am getting same word vectors for different models?
Sorry, If the question is not clear.
If you are getting identical word-vectors from models that you've prepared from different text corpuses, something is likely wrong in your process. You may not be performing any training at all, perhaps because of a problem in how the text iterable is provided to the Word2Vec class. (In that case, word-vectors would remain at their initial, randomly-initialized values.)
You should enable logging, and review the logs carefully to see that sensible counts of words, examples, progress, and incremental-progress are displayed during the process. You should also check that results for some superficial, ad-hoc checks look sensible after training. For example, does model.most_similar('hot') return other words/concepts somewhat like 'hot'?
Once you're sure models are being trained on varied corpuses – in which case their word-vectors should be very different from each other – deciding which model is 'best' depends on your specific goals with word-vectors.
You should devise a repeatable, quantitative way to evaluate a model against your intended end-uses. This might start crudely with a few of your own manual reviews of results, like looking over most_similar() results for important words for better/worse results – but should become more extensive. rigorous, and automated as your project progresses.
An example of such an automated scoring is the accuracy() method on gensim's word-vectors object. See:
https://github.com/RaRe-Technologies/gensim/blob/6d6f5dcfa3af4bc61c47dfdf5cdbd8e1364d0c3a/gensim/models/keyedvectors.py#L652
If supplied with a specifically-formatted file of word-analogies, it will check how well the word-vectors solve those analogies. For example, the questions-words.txt of Google's original word2vec code release includes the analogies they used to report vector quality. Note, though, that the word-vectors that are best for some purposes, like understanding text topics or sentiment, might not also be the best at solving this style of analogy, and vice-versa. If training your own word-vectors, it's best to choose your training corpus/parameters based on your own goal-specific criteria for what 'good' vectors will be.
Related
I have a question about CBOW prediction. Suppose my job is to use 3 surrounding words w(t-3), w(t-2), w(t-1)as input to predict one target word w(t). Once the model is trained and I want to predict a missing word after a sentence. Does this model only work for a sentence with four words which the first three are known and the last is unknown? If I have a sentence in 10 words. The first nine words are known, can I use 9 words as input to predict the last missing word in that sentence?
Word2vec CBOW mode typically uses symmetric windows around a target word. But it simply averages the (current in-training) word-vectors for all words in the window to find the 'inputs' for the prediction neural-network. Thus, it is tolerant of asymmetric windows – if there are fewer words are available on either side, fewer words on that side are used (and perhaps even zero on that side, for words at the front/end of a text).
Additionally, during each training example, it doesn't always use the maximum-window specified, but some random-sized window up-to the specified size. So for window=5, it will sometimes use just 1 on either side, and other times 2, 3, 4, or 5. This is done to effectively overweight closer words.
Finally and most importantly for your question, word2vec doesn't really do a full-prediction during training of "what exact word does the model say should be heat this target location?" In either the 'hierarchical softmax' or 'negative-sampling' variants, such an exact prediction can be expensive, requiring calculations of neural-network output-node activation levels proportionate to the size of the full corpus vocabulary.
Instead, it does the much-smaller number-of-calculations required to see how strongly the neural-network is predicting the actual target word observed in the training data, perhaps in contrast to a few other words. In hierarchical-softmax, this involves calculating output nodes for a short encoding of the one target word – ignoring all other output nodes encoding other words. In negative-sampling, this involves calculating the one distinct output node for the target word, plus a few output nodes for other randomly-chosen words (the 'negative' examples).
In neither case does training know if this target word is being predicted in preference over all other words – because it's not taking the time to evaluate all others words. It just looks at the current strength-of-outputs for a real example's target word, and nudges them (via back-propagation) to be slightly stronger.
The end result of this process is the word-vectors that are usefully-arranged for other purposes, where similar words are close to each other, and even certain relative directions and magnitudes also seem to match human judgements of words' relationships.
But the final word-vectors, and model-state, might still be just mediocre at predicting missing words from texts – because it was only ever nudged to be better on individual examples. You could theoretically compare a model's predictions for every possible target word, and thus force-create a sort of ranked-list of predicted-words – but that's more expensive than anything needed for training, and prediction of words like that isn't the usual downstream application of sets of word-vectors. So indeed most word2vec libraries don't even include any interface methods for doing full target-word prediction. (For example, the original word2vec.c from Google doesn't.)
A few versions ago, the Python gensim library added an experimental method for prediction, [predict_output_word()][1]. It only works for negative-sampling mode, and it doesn't quite handle window-word-weighting the same way as is done in training. You could give it a try, but don't be surprised if the results aren't impressive. As noted above, making actual predictions of words isn't the usual real goal of word2vec-training. (Other more stateful text-analysis, even just large co-occurrence tables, might do better at that. But they might not force word-vectors into interesting constellations like word2vec.)
In the traditional "one-hot" representation of words as vectors you have a vector of the same dimension as the cardinality of your vocabulary. To reduce dimensionality usually stopwords are removed, as well as applying stemming, lemmatizing, etc. to normalize the features you want to perform some NLP task on.
I'm having trouble understanding whether/how to preprocess text to be embedded (e.g. word2vec). My goal is to use these word embeddings as features for a NN to classify texts into topic A, not topic A, and then perform event extraction on them on documents of topic A (using a second NN).
My first instinct is to preprocess removing stopwords, lemmatizing stemming, etc. But as I learn about NN a bit more I realize that applied to natural language, the CBOW and skip-gram models would in fact require the whole set of words to be present --to be able to predict a word from context one would need to know the actual context, not a reduced form of the context after normalizing... right?). The actual sequence of POS tags seems to be key for a human-feeling prediction of words.
I've found some guidance online but I'm still curious to know what the community here thinks:
Are there any recent commonly accepted best practices regarding punctuation, stemming, lemmatizing, stopwords, numbers, lowercase etc?
If so, what are they? Is it better in general to process as little as possible, or more on the heavier side to normalize the text? Is there a trade-off?
My thoughts:
It is better to remove punctuation (but e.g. in Spanish don't remove the accents because the do convey contextual information), change written numbers to numeric, do not lowercase everything (useful for entity extraction), no stemming, no lemmatizing.
Does this sound right?
I've been working on this problem myself for some time. I totally agree with the other answers, that it really depends on your problem and you must match your input to the output that you expect.
I found that for certain tasks like sentiment analysis it's OK to remove lot's of nuances by preprocessing, but e.g. for text generation, it is quite essential to keep everything.
I'm currently working on generating Latin text and therefore I need to keep quite a lot of structure in the data.
I found a very interesting paper doing some analysis on that topic, but it covers only a small area. However, it might give you some more hints:
On the Role of Text Preprocessing in Neural Network Architectures: An Evaluation Study on Text Categorization and Sentiment Analysis
by Jose Camacho-Collados and Mohammad Taher Pilehvar
https://arxiv.org/pdf/1707.01780.pdf
Here is a quote from their conclusion:
"Our evaluation highlights the importance of being consistent in the preprocessing strategy employed across training and evaluation data. In general a simple tokenized corpus works equally or better than more complex preprocessing techniques such as lemmatization or multiword grouping, except for a dataset corresponding to a specialized domain, like health, in which sole tokenization performs poorly. Addi- tionally, word embeddings trained on multiword- grouped corpora perform surprisingly well when applied to simple tokenized datasets."
So many questions. The answer to all of them is probably "depends". It needs to be considered the classes you are trying to predict and the kind of documents you have. It's not the same to try to predict authorship (then you definitely need to keep all kinds of punctuation and case so stylometry will work) than sentiment analysis (where you can get rid of almost everything but have to pay special attention to things like negations).
I would say apply the same preprocessing to both ends. The surface forms are your link so you can't normalise in different ways. I do agree with the point Joseph Valls makes, but my impression is that most embeddings are trained in a generic rather than a specific manner. What I mean is that the Google News embeddings perform quite well on various different tasks and I don't think they had some fancy preprocessing. Getting enough data tends to be more important. All that being said -- it still depends :-)
In common applied learn-to-rank tasks, the inputs are usually semantic and have good syntactic structure, like Question-Answer ranking tasks. In this scenario, CNN or LSTM is a good structure to capture the latent information (local or long dependency) of QA-pairs.
But in reality, sometimes we just have short pair and discrete words. In this occasion, CNN or LSTM is still a fair choice？Or is there some more appropriate method can handle this?
The bigger question is how much training data you have. There's a lot of interesting work, but the reason that the deep neural network approaches tend to use QA ranking tasks is because those tasks typically have hundreds of thousands or millions of training examples.
When you have shorter queries, i.e. title or web queries, you will possibly need even more data to learn, because less of the network will be exercised by each training instance. It is possible, but the method you choose should be based on the training data you have available, rather than the size of your queries, in general.
[0-50 queries] -> Hand-tuned, time-tested model such as Query Likelihood, BM25, (or if you want better results, ngram models such as SDM) (if you want more recall, pseudo-relevance-feedback models such as RM3).
[50-1000 queries] -> Linear or Tree-based learning-to-rank methods
[1000-millions] -> Deep approach, or possibly still learning-to-rank. I'm not sure any of the deep papers have truly dominated a state-of-the-art gradient-boosted-regression-tree setup.
A recent paper by one of my labmates used pseudo-labels from BM25 to bootstrap a DNN. They got good results (better than BM25), but they literally had to be Google (training-time-wise) to pull it off.
I am stuck between a decision to apply classification or clustering on the data set I got. The more I think about it, the more I get confused. Heres what I am confronted with.
I have got news documents (around 3000 and continuously increasing) containing news about companies, investment, stocks, economy, quartly income etc. My goal is to have the news sorted in such a way that I know which news correspond to which company. e.g for the news item "Apple launches new iphone", I need to associate the company Apple with it. A particular news item/document only contains 'title' and 'description' so I have to analyze the text in order to find out which company the news referes to. It could be multiple companies too.
To solve this, I turned to Mahout.
I started with clustering. I was hoping to get 'Apple', 'Google', 'Intel' etc as top terms in my clusters and from there I would know the news in a cluster corresponds to its cluster label, but things were a bit different. I got 'investment', 'stocks', 'correspondence', 'green energy', 'terminal', 'shares', 'street', 'olympics' and lots of other terms as the top ones (which makes sense as clustering algos' look for common terms). Although there were some 'Apple' clusters but the news items associated with it were very few.I thought may be clustering is not for this kind of problem as many of the company news goes into more general clusters(investment, profit) instead of the specific company cluster(Apple).
I started reading about classification which requires training data, The name was convincing too as I actually want to 'classify' my news items into 'company names'. As I read on, I got an impression that the name classification is a bit deceiving and the technique is used more for prediction purposes as compared to classification. The other confusions that I got was how can I prepare training data for news documents? lets assume I have a list of companies that I am interested in. I write a program to produce training data for the classifier. the program will see if the news title or description contains the company name 'Apple' then its a news story about apple. Is this how I can prepare training data?(off course I read that training data is actually a set of predictors and target variables). If so, then why should I use mahout classification in the first place? I should ditch mahout and instead use this little program that I wrote for training data(which actually does the classification)
You can see how confused I am about how to address this issue. Another thing that concerns me is that if its possible to make a system this intelligent, that if the news says 'iphone sales at a record high' without using the word 'Apple', the system can classify it as a news related to apple?
Thank you in advance for pointing me in the right direction.
Copying my reply from the mailing list:
Classifiers are supervised learning algorithms, so you need to provide
a bunch of examples of positive and negative classes. In your example,
it would be fine to label a bunch of articles as "about Apple" or not,
then use feature vectors derived from TF-IDF as input, with these
labels, to train a classifier that can tell when an article is "about
Apple".
I don't think it will quite work to automatically generate the
training set by labeling according to the simple rule, that it is
about Apple if 'Apple' is in the title. Well, if you do that, then
there is no point in training a classifier. You can make a trivial
classifier that achieves 100% accuracy on your test set by just
checking if 'Apple' is in the title! Yes, you are right, this gains
you nothing.
Clearly you want to learn something subtler from the classifier, so
that an article titled "Apple juice shown to reduce risk of dementia"
isn't classified as about the company. You'd really need to feed it
hand-classified documents.
That's the bad news, but, sure you can certainly train N classifiers
for N topics this way.
Classifiers put items into a class or not. They are not the same as
regression techniques which predict a continuous value for an input.
They're related but distinct.
Clustering has the advantage of being unsupervised. You don't need
labels. However the resulting clusters are not guaranteed to match up
to your notion of article topics. You may see a cluster that has a lot
of Apple articles, some about the iPod, but also some about Samsung
and laptops in general. I don't think this is the best tool for your
problem.
First of all, you don't need Mahout. 3000 documents is close to nothing. Revisit Mahout when you hit a million. I've been processing 100.000 images on a single computer, so you really can skip the overhead of Mahout for now.
What you are trying to do sounds like classification to me. Because you have predefined classes.
A clustering algorithm is unsupervised. It will (unless you overfit the parameters) likely break Apple into "iPad/iPhone" and "Macbook". Or on the other hand, it may merge Apple and Google, as they are closely related (much more than, say, Apple and Ford).
Yes, you need training data, that reflects the structure that you want to measure. There is other structure (e.g. iPhones being not the same as Macbooks, and Google, Facebook and Apple being more similar companies than Kellogs, Ford and Apple). If you want a company level of structure, you need training data at this level of detail.
What is Conditional Random Field?
How does exactly Conditional Random Field identify proper names as person, organization, or place in a structured or unstructured text?
For example: This product is ordered by StackOverFlow Inc.
What does Conditional Random Field do to identify StackOverFlow Inc. as an organization?
A CRF is a discriminative, batch, tagging model, in the same general family as a Maximum Entropy Markov model.
A full explanation is book-length.
A short explanation is as follows:
Humans annotate 200-500K words of text, marking the entities.
Humans select a set of features that they hope indicate entities. Things like capitalization, or whether the word was seen in the training set with a tag.
A training procedure counts all the occurrences of the features.
The meat of the CRF algorithm search the space of all possible models that fit the counts to find a pretty good one.
At runtime, a decoder (probably a Viterbi decoder) looks at a sentence and decides what tag to assign to each word.
The hard parts of this are feature selection and the search algorithm in step 4.
Well to understand that you got to study a lot of things.
For start
Understand the basic of markov and bayesian networks.
Online course available in coursera by daphne coller
https://class.coursera.org/pgm/lecture/index
CRF is a special type of markov network where we have observation and hidden states.
The objective is to find the best State Assignment to the unobserved variables also known as MAP problem.
Be Prepared for a lot of probability and Optimization. :-)