Google launched a new version of the Translate in September 2016. Since then, there have been a few interesting developments in the project, and this post attempts to explain it all in as simple terms as possible.

The earlier version of the Translate used Phrase-based Machine Translation, or PBMT. What PBMT does is break up an input sentence into a set of words/phrases and translate each one individually. This is obviously not an optimal strategy, since it completely misses out on the context of the overall sentence. The new Translate uses what Google calls Google Neural Machine Translation (GNMT), an improvement over a traditional version of NMT. Lets see how GNMT works on a high-level:

The Encoder

Before you understand the encoder, you must understand what an LSTM (Long-Short-Term-Memory) cell is. It is basically a Neural Network with some concept of memory. An LSTM is generally used to ‘learn’ patterns in time-series/temporal data. At any given point, it accepts the latest input vector and produces the intended output using a combination of (the latest input + some ‘context’ regarding what it saw before):

In the above picture, $x_t$ is the input at time $t$. $h_{t-1}$ represents the context from $t-1$. If $x_t$ has a dimensionality of $d$, $h_{t-1}$ of dimensionality $2d$ is a concatenation of two vectors:

1. The intended output by the same LSTM at the last time-step $t-1$ (the Short Term memory), and
2. Another $d$-dimensional vector encoding the Long Term memory – also called the Cell State.

The second part is usually not of use for the next component in the architecture. It is instead used by the same LSTM for the following step. LSTMs are usually trained by providing them with a ton of example input-series with the expected outputs. This enables them to learn what parts of the input to retain/hold, and how to mathematically process $x_t$ and $h_{t-1}$ to come up with $h_t$. If you wish to understand LSTMs better, I recommend this blog post by Christopher Olah.

An LSTM can also be ‘unfolded’, as shown below:

Don’t worry, they are copies of the the same LSTM cell (hence same training), each feeding their output to the next one in line. What this allows us to do is give in the entire set of input vectors (in essence, the whole time-series) all at once, instead of going step-by-step with a single copy of the LSTM.

GNMT’s encoder network is essentially a series of stacked LSTMs:

Each horizontal line of pink/green boxes is an ‘unfolded’ LSTM on its own. The above figure therefore has 8 stacked LSTMs in a series. The input to the whole architecture is the ordered set of tokens in the sentence, each represented in the form of a vector. Mind you, I said tokens – not words. What GNMT does in pre-processing, is break up all words into tokens/pieces, which are then fed as a series to the neural network. This enables the framework to (atleast partially) understand unseen complicated words. For example, suppose I say the word ‘Pteromerhanophobia‘. Even though you may not know exactly what it is, you can tell me that it is some sort of fear based on the token ‘phobia‘. Google calls this approach Wordpiece modeling. The break-up of words into tokens is done based on statistical learning (which group of tokens make most sense?) from a huge vocabulary in the training phase.

When you stack LSTMs, each layer learns a pattern in the time series fed to it by the earlier (lower) layer. As you go higher up the ladder, you see more and more abstract patterns from the data that was fed in to the lowest layer. For example, the lowest layer might see a set of points and deduce a line, the next layer will see a set of lines and deduce a polygon, the next will see a set of polygons and learn an object, and so on… Ofcourse, there is a limit to how many and in what way you should stack LSTMs together – more is not always better, since you will ultimately end up with a model thats too slow and difficult to train.

There are a few interesting things about this architecture shown above, apart from the stacking of LSTMs.

You will see that the second layer from the bottom is green in color. This is because the arrows – the ordering of tokens in the sentence – is reversed for this layer. Which means that the second LSTM sees the entire sentence in reverse order. The reason to do this is simple: When you look at a sentence as a whole, the ‘context’ for any word is not just contained in the words preceding it, but also in the words following it. The two bottom-most layers both see the raw sentence as input, but in opposite order. The third LSTM gets this bidirectional input from the first two layers – basically, a combination of the forward and backward context for any given word. Each layer from this point on learns higher-level patterns in the contextual meanings of words in the sentence.

You might also have noticed the ‘+’ signs that appear before providing inputs to the fifth layer and above. This is a form of Residual Learning. This is what happens from layer 5 onwards: For every layer $N+1$, the input is an addition of the output of layers $N$ and $N-1$. Take a look at my post on Residual Neural Networks to get a better understanding of what this does.

Lastly, you can see the extra <2es> and </s> characters at the end of the input to the encoder. </s> represents ‘end of input’. <2es>, on the other hand, represents the Target Language – in this case, Spanish. GNMT does this unique thing where they provide the Target Language as input to the framework, to improve performance of Translate. More on this later.

Attention Module and the Decoder

The Encoder produces a set of ordered output-vectors (one for each token in the input). These are then fed into the Attention Module & Decoder framework. To a large extent, the Decoder is similar to the Encoder in design- stacked LSTMs and residual connections. Lets discuss the parts that are different.

I have already mentioned that GNMT considers the entire sentence as input, in every sense. However, it is intuitive to think that for every token that the decoder will produce, it should not give equal weightage to all vectors(tokens) in the input sentence. As you write out one part of the story, your focus should slowly drift to the rest of it. This work is done by the Attention Module. What the Attention Module gets as input, is the complete output of the Encoder and the latest vector from the Decoder stack. This lets it ‘understand’ how much/what has already been translated, and it then directs the Decoder to shift attention to the other parts of the Encoder output.

The Decoder LSTM-stack keeps outputting vectors based on the input from the Encoder and directions from the Attention module. These vectors are given to the Softmax Layer. You can think of the Softmax Layer as a Probability distribution-generator. Based on the incoming vector from the topmost LSTM, the Softmax Layer assigns a probability to every possible output token (remember the target language was already provided to the Encoder, so that information has already been propagated). The token that gets the maximum probability is written out.

The whole process stops once the Decoder/Softmax decides that the current token is </s> (or end-of-sentence). Note that the Decoder does not have to follow a number of steps equal to the output vectors from the Encoder, since it is paying weighted attention to all of those at every step of computation.

Overall, this is how you can visualize the  complete translation process:

Training & Zero-Shot Translation

The complete framework (Encoder+Attention+Decoder) is trained by providing it a huge collection of (input, translated) pairs of sentences. The architecture ‘knows’ the input language in a sense when it converts tokens from the incoming sentence to the appropriate vector format. The target language is provided as a parameter as well. The brilliance of deep-LSTMs lies in the fact that the neural network learns all of the computational stuff by itself, using a class of algorithms called Backpropagation/Gradient Descent.

Heres another amazing discovery made by the GNMT team: Simply by providing the target language as an input to the framework, it is able to perform Zero-Shot Translation! What this basically means is: If during training you provide it examples of English->Japanese & English->Korean translations, GNMT automatically does Japanese->Korean reasonably well! In fact, this is the biggest achievement of GNMT as a project. The intuition: what the Encoder essentially produces is a form of interlingua (or universal language). Whenever I say ‘dog‘ in any language, you end up thinking of a friendly canine – essentially, the concept of ‘dog‘. This ‘concept’ is what is produced by the Encoder, and it is irrespective of any language. In fact, some articles went so far as to say that Google’s AI had invented a language of its own :-D.

Providing the target language as input allows GNMT to easily use the same neural network for training with any pair of languages, which in turn allows zero-shot translations. As a result, the new Translate gets closer than ever before to the way humans perform translations in their mind.

Heres some references if you want to read further on this subject 🙂 :

12 thoughts on “Understanding the new Google Translate”

1. Tien Ping Tan says:

Hi, thanks for the very informative explanation. Based on your stacked LSTM figure, if a sentence has 10 words, the number of LSTM cells will be 10/level. But the number of cells can normally be set to a number e.g. 256, 512 … In this case, how’s the architecture will look like? Thanks.

1. Theoretically, it should not matter since its practically the same processing unit. But I guess we could pad the input on the right with some default tokens? The paper does not go into these details AFAIK, but the diagram still pretty much remains the same :-).

2. Quick ‘two ways’ text translation to russian and back demonstrates that every 2 of 10 sentences can be distorted at this moment

Each horizontal line of pink/green boxes is an ‘unfolded’ LSTM on its own. The above figure therefore has 8 stacked LSTMs in a series. The input to the whole architecture is the ordered set of tokens in the sentence, each represented in the form of a vector. Mind you, I said tokens – not words. What GNMT does in pre-processing, is break up all words into tokens/pieces, which are then fed as a series to the neural network. This enables the framework to (atleast partially) understand unseen complicated words. For example, suppose I say the word ‘Pteromerhanophobia‘. Even though you may not know exactly what it is, you can tell me that it is some sort of fear based on the token ‘phobia‘. Google calls this approach Wordpiece modeling. The break-up of words into tokens is done based on statistical learning (which group of tokens make most sense?) from a huge vocabulary in the training phase.

Каждая горизонтальная линия розовых / зеленых коробок является «развернутом» LSTM сам по себе. В связи с этим выше цифра имеет 8 сложенных LSTMs в серии. Исходные данные для всей архитектуры является упорядоченный набор лексем в предложении, каждое из которых представлено в виде вектора. Имейте в виду, я сказал лексемы – не слова. То, что делает GNMT в предварительной обработке, является разбить все слова на лексемы / куски, которые затем подают в виде ряда к нейронной сети. Это позволяет основу для (по крайней мере, частично) понять невидимые сложные слова. Например, предположим, что я говорю слово “Pteromerhanophobia”. Даже если вы не можете точно знать, что это такое, вы можете сказать мне, что это какой-то страх основан на лексемы “фобии”. Google называет этот подход к моделированию Wordpiece. Распад слов в лексемы осуществляется на основе статистического обучения (какая группа лексем наибольший смысл?) Из огромного словаря на этапе подготовки.

Each horizontal line of pink / green boxes is “unfolded» LSTM itself. In connection with this figure is above 8 LSTMs stacked in series. Initial data for the whole architecture is an ordered set of tokens in a sentence, each represented as a vector. Keep in mind, I said tokens – not words. What makes GNMT preprocessing is to break all the words into tokens / pieces, which are then fed to a number of neural network. This allows for the basis (at least partially) invisible difficult to understand speech. For example, suppose I say the word “Pteromerhanophobia”. Even if you can not know exactly what it is, you can tell me that this is some kind of fear is based on the token “phobia”. Google calls this approach to modeling Wordpiece. The collapse of the words in the token is based on statistical learning (which group tokens most sense?) Of a huge dictionary in preparation.

1. Aah. Its worse than 2/10, tbh. Though its not _that_ bad, considering the differences in grammar between the two languages, and the difficult/technical content. Its far from perfect per say, but cool nonetheless :-).

3. professional hair says:

Thankfulness to my father who told me regarding this blog,
this blog is truly awesome.

4. Thanks for the good writeup. It in truth used to bee a leisure account it.
Glance complicated to more introduced agreeable from
you! However, how could we keep in touch?

5. Pretty! This has been ann extremely wonderful article.
Thanks for providing this information.

6. I am regular visitor, how are you everybody? This article poated at
this skte is iin fact good.

7. Igoumenitsa Ancona says:

If some one wants to be updated with newest technologies therefore he must
bee ppay a visit his siite and be up to date everyday.