may- / cnn-re-tf Goto Github PK

Dear all,

I need help to understand these codes.

I would like to use these codes to make predictions. My data contains two class labels, namely 'Cat' and 'Bunny'. If I would like to pick "Bunny" as positive class, shall I edit the \er\target.txt and setting all instances whose class is "Bunny" as "1 0" and the others (cat) as "0 1"?

Moreover, how can I get the actual predictions?

Thank ahead for your time.

Hi there, I am back again. This time I am trying to train a context cnn model using sentence and some other attributes. For example, including sentence itself and demographic attributes of writer (age, income, education level, etc) into a source data entry: "The fox chase a bunny , 23 , 24000 , high school"

I am not quite sure whether I could just added these other attributes as left, right file, correct me if I am wrong:

source.left:
the fox chase a bunny
source.middle:
23,24000
source.right:
high school
target.txt
1 0

BUT how abount source.att? How to decide the values between 0.0 and 1.0?

Hi, can you please explain how I can form my own dataset for training MLMICNN. I'm confused with the source.att and target files and some tokens in the other files, such as , , etc.
Also is it possible to check the relation prediction for a single sentence after training is completed ?

Where do you download the entities.pickle file to use for distant supervision? Need help with this

hello!
(base_path = "http://en.wikipedia.org/w/api.php?format=xml&action=query"
query = base_path + "&list=random&rnnamespace=0&rnlimit=%d" % limit)
what is mean????
I can not get useful infomation from this url.

Hi,

I am doing a simliar project as you did.
I took a look at your text_cnn looks like the loss function you use is cross_entrophy.
I wonder how does precision and recall look like when your loss start to converge?
In my case loss start to be very small but Precision and Recall is still high, not sure what need to be done.
Did you optimize F1 specifically?
Thanks!

(.venv) ub16hp@UB16HP:/ub16_prj/cnn-re-tf$ python distant_supervision.py
===== step 1 =====
[1/4] Downloading wiki articles ...
===== step 2 =====
[1/4] Downloading wiki articles ...
===== step 3 =====
[1/4] Downloading wiki articles ...
===== step 4 =====
[1/4] Downloading wiki articles ...
===== step 5 =====
[1/4] Downloading wiki articles ...
===== step 6 =====
[1/4] Downloading wiki articles ...
===== step 7 =====
[1/4] Downloading wiki articles ...
Traceback (most recent call last):
File "distant_supervision.py", line 693, in
main()
File "distant_supervision.py", line 681, in main
positive_examples()
File "distant_supervision.py", line 452, in positive_examples
ret = loop(step, doc_id, limit, entities, relations, counter)
File "distant_supervision.py", line 343, in loop
docs = download_wiki_articles(doc_id, limit)
File "distant_supervision.py", line 73, in download_wiki_articles
pages = bs(r, "html.parser").findAll('page')
File "/home/ub16hp/ub16_prj/cnn-re-tf/.venv/local/lib/python2.7/site-packages/bs4/init.py", line 246, in init
elif len(markup) <= 256 and (
TypeError: object of type 'NoneType' has no len()
(.venv) ub16hp@UB16HP:/ub16_prj/cnn-re-tf$

Hello, Thank you for the code.

I have been trying to recreate dataset using same instructions here (https://github.com/may-/cnn-re-tf/issues/3#issuecomment-309293662) it works great till the very end and gives the following error :

Traceback (most recent call last):
File "./distant_supervision.py", line 693, in
main()
File "./distant_supervision.py", line 687, in main
extract_negative()
File "./distant_supervision.py", line 666, in extract_negative
subj = '<' + entities[row['subj'].encode('utf-8')][0] + '>'
KeyError: 'Bell\xc3\xaame'

Any thoughts? thanks in advance