How to process new data? What should be the format of input file about disan HOT 3 OPEN

That all depends on what kind of data you're using. A general procedure is:

1. load raw data, including input and corresponding label
2. tokenize the natural language and get the token list
3. build vocab for your dataset
4. map the natural language token to index by using the vocab
5. go with your network

from disan.

but in your code their is extensive use of trees. i dont understand its purpose

trees = []
with open(pjoin(data_dir, 'STree.txt'), encoding='utf-8') as file_STree,
open(pjoin(data_dir, 'SOStr.txt'), encoding='utf-8') as file_SOStr:
for STree, SOStr in zip(file_STree, file_SOStr):
sent_tree = []
STree = list(map(int, STree.strip().split('|')))
SOStr = SOStr.strip().split('|')

and for standard datasets we don't have tree.txt

from disan.

Similarly not all these files are available for every dataset, can you please give a simple example of text classification where we have train test and valid dataset files only

from disan.