Sebastian Ruder recently wrote an article on The Gradient and asserted that the oracle of natural language processing is emerging. While I am not sure such confident statement is overstated, I do look forward to the moment that we will download pre-trained embedded language models and transfer to our use cases, just like we are using pre-trained word-embedding models such as Word2Vec and FastText.

I do not think one can really draw a parallelism between computer vision and natural language processing. Computer vision is challenging, but natural language processing is even more difficult because the tasks regarding linguistics are not limited to object or meaning recognition, but also human psychology, cultures, and linguistic diversities. The objectives are far from being identical.

However, the transferrable use of embedded language models is definitely a big step forward. Ruder quoted three articles, which I would summarize below in a few words.

• Embeddings from Language Models (ELMo, arXiv:1802.05365): based on the successful bidirectional LSTM language models, the authors developed a deep contextualized embedded models by collapses all layers in the neural network architecture.
• Universal Language Model Fine-Tuning for Text Classification (ULMFiT, arXiv:1801.06146): the authors proposed a type of architectures that learn representations for specific tasks, which involve three steps in training: a) LM pre-training: learning through unlabeled corpus with abundant data; b) LM fine-tuning: learning through labeled corpus; and c) classifier fine-tuning: transferred training for specific classification tasks.
• OpenAI Transformer (article still in progress): the author proposed a simple generative language model with the three similar steps in ULMFit: a) unsupervised pre-training: training a language model that maximizes the likelihood of a sequence of tokens within a context window; b) supervised fine-tuning: a supervised classification training that maximizes the likelihood using the Bayesian approach; c) task-specific input transformations: training the classifiers on a specific task.

These three articles are intricately related to each other. Without abundant data and good hardware, it is almost impossible to produce the language models. As Ruder suggested, we will probably have a pre-trained model up to the second step of the ULMFit and OpenAI Transformer papers, but we train our own specific model for our use. We have been doing this for word-embedding models, and this approach has been common in computer vision too.

There are many tasks in natural language processing that are challenging. This blog entry is on text summarization, which briefly summarizes the survey article on this topic. (arXiv:1707.02268) The authors of the article defined the task to be

Automatic text summarization is the task of producing a concise and fluent summary while preserving key information content and overall meaning.

There are basically two approaches to this task:

• extractive summarization: identifying important sections of the text, and extracting them; and
• abstractive summarization: producing summary text in a new way.

Most algorithmic methods developed are of the extractive type, while most human writers summarize using abstractive approach. There are many methods in extractive approach, such as identifying given keywords, identifying sentences similar to the title, or wrangling the text at the beginning of the documents.

How do we instruct the machines to perform extractive summarization? The authors mentioned about two representations: topic and indicator. In topic representations, frequencies, tf-idf, latent semantic indexing (LSI), or topic models (such as latent Dirichlet allocation, LDA) are used. However, simply extracting these sentences out with these algorithms may not generate a readable summary. Employment of knowledge bases or considering contexts (from web search, e-mail conversation threads, scientific articles, author styles etc.) are useful.

In indicator representation, the authors mentioned the graph methods, inspired by PageRank. (see this) “Sentences form vertices of the graph and edges between the sentences indicate how similar the two sentences are.” And the key sentences are identified with ranking algorithms. Of course, machine learning methods can be used too.

Evaluation on the performance on text summarization is difficult. Human evaluation is unavoidable, but with manual approaches, some statistics can be calculated, such as ROUGE.

When it was proposed that GAN uses Wasserstein distance as the training metric, GAN is usually seen as a transportation problem. Previously, it was mentioned in a previous post that GAN can be seen as a transportation problem, and because of that, some computation can be simplified by relating a kernel in the discriminator and the generator.

GAN can be used in word translation problem too. In a recent preprint in arXiv (refer to arXiv:1710.04087), Wasserstein GAN has been used to train a machine translation machine, given that there are no parallel data between the word embeddings between two languages. The translation mapping is seen as a generator, and the mapping is described using Wasserstein distance. The training objective is cross-domain similarity local scaling (CSLS). Their work has been performed in English-Russian and English-Chinese mappings.

It seems to work. Given GAN sometimes does not work for unknown reasons, it is an excitement that it works.

The Python package for text mining shorttext has a new release: 0.5.4. It can be installed by typing in the command line:

pip install -U shorttext

For some people, you may need to install it from “root”, i.e., adding sudo in front of the command. Since the version 0.5 (including releases 0.5.1 and 0.5.4), there have been substantial addition of functionality, mostly about comparisons between short phrases without running a supervised or unsupervised machine learning algorithm, but calculating the “similarity” with various metrics, including:

• soft Jaccard score (the same kind of fuzzy scores based on edit distance in SOCcer),
• Word Mover’s distance (WMD, detailedly described in a previous post), and
• Jaccard index due to word-embedding model.

For the soft Jaccard score due to edit distance, we can call it by:

>>> from shorttext.metrics.dynprog import soft_jaccard_score
>>> soft_jaccard_score(['book', 'seller'], ['blok', 'sellers'])     # gives 0.6716417910447762
>>> soft_jaccard_score(['police', 'station'], ['policeman'])        # gives 0.2857142857142858

The core of this code was written in C, and interfaced to Python using SWIG.

For the Word Mover’s Distance (WMD), while the source codes are the same as my previous post, it can now be called directly. First, load the modules and the word-embedding model:

>>> from shorttext.metrics.wasserstein import word_mover_distance
>>> wvmodel = load_word2vec_model('/path/to/model_file.bin')

And compute the WMD with a single function:

>>> word_mover_distance(['police', 'station'], ['policeman'], wvmodel)                      # gives 3.060708999633789
>>> word_mover_distance(['physician', 'assistant'], ['doctor', 'assistants'], wvmodel)      # gives 2.276337146759033

And the Jaccard index due to cosine distance in Word-embedding model can be called like this:

>>> from shorttext.metrics.embedfuzzy import jaccardscore_sents
>>> jaccardscore_sents('doctor', 'physician', wvmodel)   # gives 0.6401538990056869
>>> jaccardscore_sents('chief executive', 'computer cluster', wvmodel)   # gives 0.0022515450768836143
>>> jaccardscore_sents('topological data', 'data of topology', wvmodel)   # gives 0.67588977344632573

Most new functions can be found in this tutorial.

And there are some minor bugs fixed.

Much about the use of word-embedding models such as Word2Vec and GloVe have been covered. However, how to measure the similarity between phrases or documents? One natural choice is the cosine similarity, as I have toyed with in a previous post. However, it smoothed out the influence of each word. Two years ago, a group in Washington University in St. Louis proposed the Word Mover’s Distance (WMD) in a PMLR paper that captures the relations between words, not simply by distance, but also the “transportation” from one phrase to another conveyed by each word. This Word Mover’s Distance (WMD) can be seen as a special case of Earth Mover’s Distance (EMD), or Wasserstein distance, the one people talked about in Wasserstein GAN. This is better than bag-of-words (BOW) model in a way that the word vectors capture the semantic similarities between words.

# Word Mover’s Distance (WMD)

The formulation of WMD is beautiful. Consider the embedded word vectors $\mathbf{X} \in R^{d \times n}$, where $d$ is the dimension of the embeddings, and $n$ is the number of words. For each phrase, there is a normalized BOW vector $d \in R^n$, and $d_i = \frac{c_i}{\sum_i c_i}$, where $i$‘s denote the word tokens. The distance between words are the Euclidean distance of their embedded word vectors, denoted by $c(i, j) = || \mathbf{x}_i - \mathbf{x}_j ||_2$, where $i$ and $j$ denote word tokens. The document distance, which is WMD here, is defined by $\sum_{i, j} \mathbf{T}_{i j} c(i, j)$, where $\mathbf{T}$ is a $n \times n$ matrix. Each element $\mathbf{T}_{ij} \geq 0$ denote how nuch of word $i$ in the first document (denoted by $\mathbf{d}$) travels to word $j$ in the new document (denoted by $\mathbf{d}'$).

Then the problem becomes the minimization of the document distance, or the WMD, and is formulated as:

$\text{min}_{\mathbf{T} \geq 0} \sum_{i, j=1}^n \mathbf{T}_{ij} c(i, j)$,

given the constraints:

$\sum_{j=1}^n \mathbf{T}_{ij} = d_i$, and

$\sum_{i=1}^n \mathbf{T}_{ij} = d_j'$.

This is essentially a simplified case of the Earth Mover’s distance (EMD), or the Wasserstein distance. (See the review by Gibbs and Su.)

# Using PuLP

The WMD is essentially a linear optimization problem. There are many optimization packages on the market, and my stance is that, for those common ones, there are no packages that are superior than others. In my job, I happened to handle a missing data problem, in turn becoming a non-linear optimization problem with linear constraints, and I chose limSolve, after I shop around. But I actually like a lot of other packages too. For WMD problem, I first tried out cvxopt first, which should actually solve the exact same problem, but the indexing is hard to maintain. Because I am dealing with words, it is good to have a direct hash map, or a dictionary. I can use the Dictionary class in gensim. But I later found out I should use PuLP, as it allows indices with words as a hash map (dict in Python), and WMD is a linear programming problem, making PuLP is a perfect choice, considering code efficiency.

An example of using PuLP can be demonstrated by the British 1997 UG Exam, as in the first problem of this link, with the Jupyter Notebook demonstrating this.

# Implementation of WMD using PuLP

The demonstration can be found in the Jupyter Notebook.

from itertools import product
from collections import defaultdict

import numpy as np
from scipy.spatial.distance import euclidean
import pulp
import gensim

Then define the functions the gives the BOW document vectors:

def tokens_to_fracdict(tokens):
cntdict = defaultdict(lambda : 0)
for token in tokens:
cntdict[token] += 1
totalcnt = sum(cntdict.values())
return {token: float(cnt)/totalcnt for token, cnt in cntdict.items()}

Then implement the core calculation. Note that PuLP is actually a symbolic computing package. This function return a pulp.LpProblem class:

def word_mover_distance_probspec(first_sent_tokens, second_sent_tokens, wvmodel, lpFile=None):
all_tokens = list(set(first_sent_tokens+second_sent_tokens))
wordvecs = {token: wvmodel[token] for token in all_tokens}

first_sent_buckets = tokens_to_fracdict(first_sent_tokens)
second_sent_buckets = tokens_to_fracdict(second_sent_tokens)

T = pulp.LpVariable.dicts('T_matrix', list(product(all_tokens, all_tokens)), lowBound=0)

prob = pulp.LpProblem('WMD', sense=pulp.LpMinimize)
prob += pulp.lpSum([T[token1, token2]*euclidean(wordvecs[token1], wordvecs[token2])
for token1, token2 in product(all_tokens, all_tokens)])
for token2 in second_sent_buckets:
prob += pulp.lpSum([T[token1, token2] for token1 in first_sent_buckets])==second_sent_buckets[token2]
for token1 in first_sent_buckets:
prob += pulp.lpSum([T[token1, token2] for token2 in second_sent_buckets])==first_sent_buckets[token1]

if lpFile!=None:
prob.writeLP(lpFile)

prob.solve()

return prob

To extract the value, just run pulp.value(prob.objective)

We use Google Word2Vec. Refer the $\mathbf{T}$ matrices in the Jupyter Notebook. Running this by a few examples:

1. document1 = President, talk, Chicago
document2 = President, speech, Illinois
WMD = 2.88587622936
2. document1 = physician, assistant
document2 = doctor
WMD = 2.8760048151
3. document1 = physician, assistant
document2 = doctor, assistant
WMD = 1.00465738773
(compare with example 2!)
4. document1 = doctors, assistant
document2 = doctor, assistant
WMD = 1.02825379372
(compare with example 3!)
5. document1 = doctor, assistant
document2 = doctor, assistant
WMD = 0.0
(totally identical; compare with example 3!)

There are more examples in the notebook.

# Conclusion

WMD is a good metric comparing two documents or sentences, by capturing the semantic meanings of the words. It is more powerful than BOW model as it captures the meaning similarities; it is more powerful than the cosine distance between average word vectors, as the transfer of meaning using words from one document to another is considered. But it is not immune to the problem of misspelling.

This algorithm works well for short texts. However, when the documents become large, this formulation will be computationally expensive. The author actually suggested a few modifications, such as the removal of constraints, and word centroid distances.

Example codes can be found in my Github repository: stephenhky/PyWMD.

In these few days, Facebook published a new research paper, regarding the use of sequence to sequence (seq2seq) model for machine translation. What is special about this seq2seq model is that it uses convolutional neural networks (ConvNet, or CNN), instead of recurrent neural networks (RNN).

The original seq2seq model is implemented with Long Short-Term Memory (LSTM) model, published by Google.(see their paper) It is basically a character-based model that generates texts according to a sequence of input characters. And the same author constructed a neural conversational model, (see their paper) as mentioned in a previous blog post. Daewoo Chong, from Booz Allen Hamilton, presented its implementation using Tensorflow in DC Data Education Meetup on April 13, 2017. Johns Hopkins also published a spell correction algorithm implemented in seq2seq. (see their paper) The real advantage of RNN over CNN is that there is no limit about the size of the tokens input or output.

While the fixing of the size of vectors for CNN is obvious, using CNN serves the purpose of limiting the size of input vectors, and thus limiting the size of contexts. This limits the contents, and speeds up the training process. RNN is known to be trained slow. Facebook uses this CNN seq2seq model for their machine translation model. For more details, take a look at their paper and their Github repository.

On November 21, 2016, the Python package shorttext’ was published. Until today, more than seven versions have been published. There have been a drastic architecture change, but the overall purpose is still the same, as summarized in the first introduction entry:

This package shorttext‘ was designed to tackle all these problems… It contains the following features:

• example data provided (including subject keywords and NIH RePORT);
• text preprocessing;
• pre-trained word-embedding support;
• gensim topic models (LDA, LSI, Random Projections) and autoencoder;
• topic model representation supported for supervised learning using scikit-learn;
• cosine distance classification; and
• neural network classification (including ConvNet, and C-LSTM).

And since the first version, there have been updates, as summarized in the documention (News):

## Version 0.3.3 (Apr 19, 2017)

• Deleted CNNEmbedVecClassifier.

## Version 0.3.2 (Mar 28, 2017)

• Bug fixed for gensim model I/O;
• Console scripts update;
• Neural networks up to Keras 2 standard (refer to this).

## Version 0.3.1 (Mar 14, 2017)

• Compact model I/O: all models are in single files;
• Implementation of stacked generalization using logistic regression.

## Version 0.2.1 (Feb 23, 2017)

• Removal attempts of loading GloVe model, as it can be run using gensim script;
• Confirmed compatibility of the package with tensorflow;
• Use of spacy for tokenization, instead of nltk;
• Use of stemming for Porter stemmer, instead of nltk;
• Removal of nltk dependencies;
• Simplifying the directory and module structures;
• Module packages updated.

Although there are still additions that I would love to add, but it would not change the overall architecture. I may add some more supervised learning algorithms, but under the same network. The upcoming big additions will be generative models or seq2seq models, but I do not see them coming in the short term. I will add corpuses.

I may add tutorials if I have time.

I am thankful that there is probably some external collaboration with other Python packages. Some people have already made some useful contributions. It will be updated if more things are confirmed.

I have worked a lot on text categorization in the past few months, and I started to get bored. I started to become more interested in generative models, and generating texts.

Generative models are not new. Topic models such as LDA, or STM are generative models. However, I have been using the topic vectors or other topic models such as LDA2Vec as the feature of another supervised algorithm. And it is basically the design of my shorttext package.

I attended a meetup event held by DC Data Science and Data Education DC. The speaker, Daewoo Chong, is a senior Data Scientist at Booz Allen Hamilton. He talked about chatbot, building on RNN models on characters. His talk was not exactly about generative models, but it is indeed about generating texts. With the sophistication of GANs (see my entry on GAN and WGAN), it will surely be my next focus of my toy projects.

Ran Chen wrote a blog on his company homepage about natural language generation in his system, Trulia.

And there are a few GAN applications on text:

• “Generating Text via Adversarial Learning” [PDF]
• Lantao Yu, Weinan Zhang, Jun Wang, Yong Yu, “SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient,” arXiv:1609.05473 [arXiv]
• Jiwei Li, Will Monroe, Tianlin Shi, Sébastien Jean, Alan Ritter, Dan Jurafsky, “Adversarial Learning for Neural Dialogue Generation,” arXiv:1701.06547 [arXiv]
• Matt J. Kusner, José Miguel Hernández-Lobato, “GANs for sequence of discrete elements with the Gumbel-softmax distribution,” arXiv:1611.04051 [arXiv]
• David Pfau, Oriol Vinyals, “Connecting generative adversarial network and actor-critic methods,” arXiv:1610.01945 [arXiv]
• Xuerong Xiao, “Text Generation usingGenerative Adversarial Training” [PDF]

The package shorttext has received attention for the past two months. A new release is released yesterday for the following updates:

1. Removal attempts of loading GloVe model, as it can be run using gensim script;
2. Confirmed compatibility of the package with Tensorflow;
3. Use of spacy for tokenization, instead of nltk;
4. Use of stemming for Porter stemmer, instead of nltk;
5. Removal of nltk dependencies;
6. Simplifying the directory and module structures;
7. Module packages updated.

For #1, it actually removed a bug in the previous release. Instead, the users should convert the GloVe models into Word2Vec using the script provided by gensim.

For #3, #4, and #5, it is basically removing any nltk dependencies, because very few functionalities of nltk was used, and it is slow. For Porter stemmer, there is a light-weighted library stemming that performs the task perfectly. For tokenization, the tokenizer in spaCy is significantly faster than nltk, as shown in this Jupyter Notebook. We can do a simple test here, by first importing:

import time
import shorttext


nihdata = shorttext.data.nihreports()
nihtext = ' '.join(map(lambda item: ' '.join(item[1]), nihdata.items()))


Then find the time of using the tokenizer in nltk:

from nltk import word_tokenize

nltkt0 = time.time()
tokens = word_tokenize(nihtext)
nltkt1 = time.time()
print nltkt1-nltkt0, ' sec'   # output: 0.0224239826202 sec


On the other hand, using spaCy gives:

import spacy