topic Can File locations be provided in Visual Text analytics? in SAS Data Science

Can File locations be provided in Visual Text analytics?

SundareshS — Mon, 02 Jul 2018 22:26:34 GMT

I have a feeling the answer is no ... but isn't there a way we can provide a location to a filename as part of the data source in Visual Text Analytics (much like we provided a file URL / point to a folder containing text files in Contextual Analysis / Text miner) ?

I feel it will be a convenient feature in a future release, though there might need to be some thought devoted on how to load that file (rather than just a location to that file) to CAS.

regards

Sundaresh

Re: Can File locations be provided in Visual Text analytics?

RussAlbright — Tue, 03 Jul 2018 02:45:15 GMT

No, we currently do not have a way to do this with text actions. Typically the server side actions only have access to content through a table and not through a file system. With the variable length characters, VIYA does not have the 32k limit. So once you have the content in a VIYA table, you can process the longer text variables. Also, if you can get the files into rows of a flat file such as a csv file, then you can use the loadTable action to create a Viya Table that holds the entire content of your observations from the csv file.

Re: Can File locations be provided in Visual Text analytics?

SundareshS — Tue, 03 Jul 2018 04:05:49 GMT

Thank you. Yes, your comment about the server side actions ( ... have access to content through a table and not through a file system ) makes it clearer. At some point of time, you would need that data (the text content) in memory anyways (as a CAS table) so I now realize that there isn't much to be gained (apart from a small convenience) from what I originally wished for.

One alternative (if I want to be stubborn) is to probably make use of the python-swat package, start with a file containing location of documents , and have a python program carry out the work of collect the required text from all mentioned documents, and load it to a CAS table. Or a SAS program which does the same thing.

Thanks again.

Sundar