Directory Error

[Spruces]

Hello, I tried to extract the features with extract_features.py

and the result is just like this.

I don't get where the open_data_type_78_header_valid.csv's location is.

• Can I just use this model as easier as Sato Web Demo, just upload and get return value?

Too complicate to use, and not much instructions

[jason022085]

Same error occurs to me. I think the solution is to understand the usage of Viznet, because there is similar code in it. https://github.com/mitmedialab/viznet

[Spruces]

Same error occurs to me. I think the solution is to understand the usage of Viznet, because there is similar code in it. https://github.com/mitmedialab/viznet

Thanks bro, it's weird. To execute it, you need certain .csv files(like open_data_type_78_header_valid.csv) that doesn't provided by sato.

and I had examined the codes, you need certain type of NLTKs to generate open_data_type_78_header_valid.csv.

Also, after I found certain types of NLTKs, I couldn't figure out where exactly the NLTK should be located in certain directory.

Poor guidelines.

[Spruces]

And how they can detect semantic columns with NLTKs?

Like my ID, and yours. They don't have any semantic things as they are just nouns.

I think that for detecting 'ID' columns based on our id(Spruces,jason022085) they need something else like the lengths of inputs or how many blanks they had, and the number of separator comma(address).

cus there's no meaning in our names of addresses as themselves.

[jason022085]

p.s. Remember to download VizNet data first(produced by retrieve_corpora.sh) and set RAW_DIR to the path of raw data form Viznet In my case, it is os.environ['RAW_DIR'] = r"D:\viznet-master\raw"

Since other dataset are too large, I am working on "manyeyes" dataset.

And I found there is a extract_header.py. It may help

I got it ! You have to extract header first, and then extract features

[RichaMax]

Hello, from what I understand you can download open_data_type_78_header_valid.csv and the sherlocks features they used using their script ./download_data.sh

[horseno]

Sorry about the confusion. If you wish to extract the features, you'll first need to run extract_header.py to get the valid headers (in the 78 types after canonicalization) from the dataset.

[iganand]

I have empty header files generated when I try extract_headers.py for manyeyes. Can you please help? What it might be. Please let me know