Can someone explain how to create a PTB Dataset And/Or Train my own model using StanfordNLP?

I'm learning about sentiment analysis and I can't seem to find anything online that outlines how to create a PTB Dataset. I'm using StanfordNLP with Java. I've downloaded the test, dev and validate data that they used and I can't get my head around how these have been outlined:

test.txt:

(3 (2 (2 The) (2 Rock)) (4 (3 (2 is) (4 (2 destined) (2 (2 (2 (2 (2 to) (2 (2 be) (2 (2 the) (2 (2 21st) (2 (2 (2 Century) (2 's)) (2 (3 new) (2 (2 ``) (2 Conan)))))))) (2 '')) (2 and)) (3 (2 that) (3 (2 he) (3 (2 's) (3 (2 going) (3 (2 to) (4 (3 (2 make) (3 (3 (2 a) (3 splash)) (2 (2 even) (3 greater)))) (2 (2 than) (2 (2 (2 (2 (1 (2 Arnold) (2 Schwarzenegger)) (2 ,)) (2 (2 Jean-Claud) (2 (2 Van) (2 Damme)))) (2 or)) (2 (2 Steven) (2 Segal))))))))))))) (2 .)))

I figure that numbers are aligned to sentiment value but I'm still not sure how it works.

TLDR; I'm trying to develop my own model for news analysis and have seen that the StanfordNLP model has been trained on movie reviews which is leading to poor sentiment analysis so, I thought to attempt to develop my own but I can't find anything online that teaches what each element is or how to even do this.

At best; outlined on this page: https://nlp.stanford.edu/sentiment/code.html

Is the dataset available and the code to train.

Models can be retrained using the following command using the PTB format dataset:

java -mx8g edu.stanford.nlp.sentiment.SentimentTraining -numHid 25 -trainPath train.txt -devPath dev.txt -train -model model.ser.gz

I have the data that I need to parse ready.

Solution

Okay.. So I've done some more digging and have started to finally understand (some what) as how to create a Dataset Tree and will try to break it down for anyone who stumbles upon this post with the same troubles as I've been having.

Step 1.

Find your data. (In my case it's news articles about the UK housing market)

UK renters: are you living with someone you’ve fallen out with?
UK property asking prices stagnating, lifting hopes of softer landing for housing market

Step 2.

Annotate your data

2 UK renters: are you living with someone you’ve fallen out with?
1 fallen out with
1 fallen out
2 UK renters
2 living with someone
3 fallen
2 :
2 ?
2 living with
2 someone

3 UK property asking prices stagnating, lifting hopes of softer landing for housing market
2 UK property
3 asking prices stagnating
2 asking prices
4 lifting hopes
2 hopes
4 lifting hopes of softer landing
3 softer landing for housing market
2 housing market
2 lifting
2 landing
2 ,

Annotation Meanings

Very Positive= 4
Positive = 3
Neutral = 2
Negative = 1
Very Negative = 0

Structure

2 UK renters: are you living with someone you’ve fallen out with?
   //Overall sentiment

1 fallen out with
   // Negative

1 fallen out
   // Negative

2 UK renters
   // Neutral

...etc..

Save the annotated data to a .txt (sample.txt)

Step 3:

Locate your stanford-corenlp-4.5.2.jar
- example ~/.m2/repository/edu/stanford/nlp/stanford-corenlp/4.5.2

Step 4:

Open Bash and run
- java -cp "*" -mx5g edu.stanford.nlp.sentiment.BuildBinarizedDataset -input /c/Users/rusku/Desktop/StanfordNPL/rusSample/sample.txt
- replace the above data location

Step 5:

Result

(2 (2 (2 (2 UK) (2 renters)) (2 :)) (2 (2 (2 (2 are) (2 you)) (2 (2 living) (2 (2 with) (2 (2 someone) (2 (2 you) (2 (2 ▒ve) (1 (1 (3 fallen) (2 out)) (2 with)))))))) (2 ?)))
(3 (3 (2 (3 UK) (3 property)) (2 (3 asking) (3 prices))) (3 (3 (3 stagnating) (3 (2 ,) (4 (2 lifting) (2 hopes)))) (3 (3 of) (3 (3 (3 softer) (2 landing)) (3 (3 for) (2 (3 housing) (3 market)))))))

Resource: Train Stanford CoreNLP about the sentiment of domain-specific phrases

This is as far as I've currently gotten.

Hope this helps.