This folder contains the code of running the user interface.
To start the local server, you can run python server/server.py inside this folder.
Then you can visit http://localhost:7070/ to interact with the user interface.
The code of the system of iSEA is organized as below:
-
The
data/folder contains the pre-computed data. We describe thedata processing step in the pre-process/ directory. For a given <data_name> (e.g., "twitter", "mnli_government" as we describe in the paper), the following files should be included to run the web application successfully:-
<data_name>_binary/list.json: the binary rule (based on tokens). -
<data_name>_binary/test.json: the tokens. -
<data_name>/doc.jsonl: the actual documents. The format of the this file is slightly different from the one used during data process when used for the user interface. Check/pre-process/02-doc_processfor more details. -
<data_name>/model_output.csv: model predictions and ground truth. -
<data_name>/model_stat.json: model performance statistics used for the overall stat. view. -
<data_name>/hfeat_stat.csv: extracted high-level features. -
<data_name>_hfeat/list.json: the extracted rules based on high-level features. -
<data_name>_hfeat/test.json: features used as high-level features, and thresholds for low/medium/high values. -
<data_name>/sentence_tsne.csv: 2-dimensional projection of document embeddings for each document (2d positions in the projection view).
-
-
The
server/folder contains the code of the server side, include app routing function supported byFlask, and other real-time computation functions. -
The
static/folder contains all the front-end code, including the visualization and interactive functions.