The cacm from xuanyuan14

cacm's Issues

Inconsistency between argparse and cmd

In run.py line 62-67, these three argument types should be string or store_true. With current format, these auguments would set True regardless of cmd "--use_knowledge True" or "--use_knowledge False".

Embedding data generation

Hi, I have a question regarding the data generation for node2vec. I got data from TianGong-ST page here by contacting [email protected]. But the following command seems not working well due to missing vtype for each document interaction.

python TianGong-ST-CACM.py --xml_clean --dict_list --txt --node2vec --human_label_txt_for_CACM --dataset sogousessiontrack2018.xml --input ../ST/data/ --output ./data
===> cleaning xml file...
  - start reading from xml file...
  - read 50265215 lines
  - start removing useless lines...
  - remove 1000000 lines...
  - remove 2000000 lines...
  - remove 3000000 lines...
  - remove 4000000 lines...
  - remove 5000000 lines...
  - remove 6000000 lines...
  - remove 7000000 lines...
  - remove 8000000 lines...
  - remove 9000000 lines...
  - remove 10000000 lines...
  - remove 11000000 lines...
  - remove 12000000 lines...
  - remove 13000000 lines...
  - remove 14000000 lines...
  - remove 15000000 lines...
  - remove 16000000 lines...
  - remove 17000000 lines...
  - remove 18000000 lines...
  - remove 19000000 lines...
  - remove 20000000 lines...
  - remove 21000000 lines...
  - remove 22000000 lines...
  - remove 23000000 lines...
  - remove 24000000 lines...
  - remove 25000000 lines...
  - remove 26000000 lines...
  - remove 27000000 lines...
  - remove 28000000 lines...
  - remove 29000000 lines...
  - remove 30000000 lines...
  - remove 31000000 lines...
  - remove 32000000 lines...
  - remove 33000000 lines...
  - remove 34000000 lines...
  - remove 35000000 lines...
  - read 50265215 lines
  - totally 356252 iteractions
  - totally remove 35981452 lines
===> generating dicts and lists...
  - start parsing xml file...
  - generating infos_per_session...
Traceback (most recent call last):
  File "TianGong-ST-CACM.py", line 526, in <module>
    main()
  File "TianGong-ST-CACM.py", line 510, in main
    generate_dict_list(args)
  File "TianGong-ST-CACM.py", line 121, in generate_dict_list
    vtype = doc.getElementsByTagName('vtype')[0].childNodes[0].data
IndexError: list index out of range

I'm new to this dataset and not sure where to find the cause for this. Would you please give more information about what is vtype and what might be the possible cause for this? Thanks a lot!

Support for negative weight in node2vec

For generating node embeddings using node2vec, I use the official python-version code: https://github.com/aditya-grover/node2vec

But I was caught in errors that node2vec can not support negative edge weights because there are normalization operations over weights. It turns out that, instead of the python version, we should use C++ node2vec tools which support negative weights.

So I submit this issue as a reminder.

Data preparation

Can you give some examples about how to use node2vec to get embeddings and how to map queries ,documents into node identifiers?

And I can't access the dataset through your link http://www.thuir.cn/tiangong-st/

Parameters of node2vec

In order to better reproduce your wonderful work, could you provide the training parameters of node2vec, the known parameters are -d:64 -dr -w, how do you set other such as -l,-r,-k,-e,-p,-q?

xuanyuan14 / cacm Goto Github PK

cacm's Introduction

cacm's People

Contributors

Stargazers

Watchers

Forkers

cacm's Issues

Inconsistency between argparse and cmd

Embedding data generation

Support for negative weight in node2vec

Data preparation

Parameters of node2vec

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent