syntaxnet | Syntaxnet Parsey McParseface wrapper for POS | Natural Language Processing library

I am new to python and

I am trying to test syntaxnet in python using this github repo.

and in the section "how to run the parser" is as following :

1. git clone https://github.com/spoddutur/syntaxnet.git
2. cd
3. python main.py
4. That's it!! It prints syntaxnet dependency parser output for given input english sentence

through some research, I understood that first one indicates that I need to install the syntaxnet package in my cmd so I did and the package was successfully installed. but I don't understand how to perform the second one what does cd do, and where and how should I use it?

also in main.py,

I am trying to get a dependency tree in JSON format from SyntaxNet but all I get from the examples is a Sentence Object which is providing no accessors to access the parsed object or even iterate through the items listed.

When I run the examples from the docker file provided by TensorFlow/SyntaxNet, what I get is as below

Currently I'm trying to build syntax-aware NMT model.
In this project, I need the sequence of one of three transition actions (SHIFT, REDUCE-L, REDUCE-R)

Similar to what is in the image a

This chunk represents the transition-based dependency for 2 sentences(1 for 1 chunk split by empty lines)

I'm using Syntaxnet to get the dependency parse tree first, but it doesn't directly provide that transition action sequences.
It's results are as follows,

b

Is it possible to get the action sequences similar to this image? Is it possible to convert what is achieved from this image to the original image's format.

I have a large dataset of sentences (i.e., ~5.000.000) in raw text which I want to process using SyntaxNet already trained for English. That is, I just want to process the sentences using a SyntaxNet model, I don't want to train any new model.

Setting up a processing environment with GPUs will have any effect on performance ?

I understand that most of the heavy CPU operations is on estimating the parameters and weights of the network/model, once these are estimated, applying the trained network should be faster than training.

Nevertheless, I've never worked before with Tensorflow and I don't know whether GPUs are used when one applies an already trained model to data.

Also, does anyone knows any easy way to setup SyntaxNet as a daemon or web-service, so that batch processing can be made easily?

I want to retrain existing models for SyntaxNet/DragNN and looking for some real numbers how long does it take to train models for any language (it will give me good baseline for my languages). What hardware have you used during this process?

Thank you in advance!

I exhausted my virtual memory when trying to install SyntaxNet from this Dockerfile using the Docker Toolbox. I received this message when compiling the Dockerfile:

I installed bazel and upgrade it to 0.4.4 recently.

I want to try tensorflow/models/syntaxnet but it requires bazel 0.4.3.

So how can I downgrade bazel 0.4.4 to 0.4.3?

I know the more data, the better it's but what would be a reasonable amount of data required to train SyntaxNet?

I annotated a corpus using pre-trained syntaxnet model (i.e. using Parse McParseface). I am having a problem understanding the output. There are two metrics reproted in the output. Are those for POS tagging and dependency parsing? If yes, which one is POS tagging performance and which one is for dependency parsing performance?

Here is the output:

INFO:tensorflow:Total processed documents: 21710 INFO:tensorflow:num correct tokens: 454150 INFO:tensorflow:total tokens: 560993 INFO:tensorflow:Seconds elapsed in evaluation: 1184.63, eval metric: 80.95% INFO:tensorflow:Processed 206 documents INFO:tensorflow:Total processed documents: 21710 INFO:tensorflow:num correct tokens: 291851 INFO:tensorflow:total tokens: 504496 INFO:tensorflow:Seconds elapsed in evaluation: 1193.17, eval metric: 57.85%