inpho/topic-explorer

Web interface for displaying LDA topic models generated by the VSM framework

Neeraj Kumar

@kumarneeraj2005

Hello..please guide me how to run inpho/topic-explorer on new data

Jaimie Murdock

@JaimieMurdock

Hi Neeraj, glad to hear from you

Sorry I wasn't online earlier - the easiest way to prepare your data is to create a folder consisting of PDFs or TXT files

the topic explorer can then be pointed at that folder:
topicexplorer init <YOUR PATH HERE>

I can give more help if I know whether you are on Windows, Mac, or Linux

Ed Kaplan

@dropshot_twitter

Great project! I just started using it, and am examining my first corpus of data. The topic modeling goes fine, but I can't open a document. The document entry page "select a random document" button doesn't pull anything up, and when I click on the fingerprint next to a topic bar, nothing happens.

Jaimie Murdock

@JaimieMurdock

Hi Ed, thanks for the contact.

Which OS are you using?

When you run "topicexplorer launch" in your console it should log requests - if you set up your computer so you have the explorer open in a browser and can still see the console, you'll notice that everytime you go to a new page, it prints a line in the console

Here's an example of some lines:

127.0.0.1 - - [17/Apr/2017:08:58:32 -0400] "GET /40/topics.json HTTP/1.1" 200 15161 "http://compy386:8000/" "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.133 Safari/537.36"
127.0.0.1 - - [17/Apr/2017:08:58:32 -0400] "GET /60/topics.json HTTP/1.1" 200 22759 "http://compy386:8000/" "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.133 Safari/537.36"

Jaimie Murdock

@JaimieMurdock

I actually just found an error on my machine: uBlock Origin will block the fingerprint.js script that tries to load

That's an extension on Chrome and Firefox - I'm not sure why it's blocking right now, but will file an issue and look into it

@dropshot_twitter will resolve at #198

Ed Kaplan

@dropshot_twitter

ah...I'm running uBlock too

The first problem is resolved by disabling uBlock, but I've run into a second problem which uBlock disabling doesn't fix: I get no response when I put a term into the search box.

Jaimie Murdock

@JaimieMurdock

(I'm pushing a patch to permanently solve the uBlock issue, while keeping uBlock on)

Ed Kaplan

@dropshot_twitter

Here's the relevant entry from the console:

10.211.55.2 - - [17/Apr/2017:09:16:29 -0400] "GET /topics.json?q=cuba HTTP/1.1" 400 25 "http://edwards-macbook-pro.local:8000/topics?q=cuba" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.133 Safari/537.36"

Just clicking on "view topic clusters" without a search term works fine.

Jaimie Murdock

@JaimieMurdock

The issue there is that cuba may have been stop-listed during the prepstage, especially if it was a very frequent or infrequent word in your corpus

Ed Kaplan

@dropshot_twitter

ah...that's probably it...documents from the kennedy library

thank you!

btw...how do I pull the patched version?

Jaimie Murdock

@JaimieMurdock

ah, that makes sense - cool corpus too! I should definitely add a GUI message about the terms - the browser sees a 400 Bad Request but for the user it just fails silently.

in about 5 minutes I should have the patch uploaded to the server and topicexplorer update will update the software

Ed Kaplan

@dropshot_twitter

Thanks! I'm just getting into topic modeling and the digital humanities generally. I study nuclear and airpower history. I'd like to trace the development of certain ideas (like ways of thinking about limited war) over time. I think this will give me great insights into my documents.

Thanks for making it a bit easier.

Jaimie Murdock

@JaimieMurdock

That's really fascinating, would love to hear more about the research and always available to help with software issues!

The patch is available via topicexplorer update now and I've opened an issue for the warning (#199) and will fix by Wednesday

Ed Kaplan

@dropshot_twitter

Thank you for the prompt help!

Do you have a roadmap for future development? Does it perhaps include interactive topic modeling (https://github.com/travisbrown/itm)?

Jaimie Murdock

@JaimieMurdock

the three biggest features in development are:

a Python 3.x port
internationalization, esp Chinese to facilitate some ongoing work on the Han'dian classics https://arxiv.org/abs/1702.00860
a new interface for the prep stage, as it's really hard both to understand and explain in current form

Ed Kaplan

@dropshot_twitter

cool stuff! thanks for taking this on

Jaimie Murdock

@JaimieMurdock

should be a more open roadmap posted sometime in May - in the middle of changing affiliations, so I've been piecing together a long-term support structure now :)

Ed Kaplan

@dropshot_twitter

You definitely have an excited user in me.

Jaimie Murdock

@JaimieMurdock

@dropshot_twitter just closed #199 so a topicexplorer update should get the proper warning messages! Sorry for the 1 week delay.

YvonneZhen

@YvonneZhen

Hi Jamie. Do you guys still have a lab or group at IU that I can talk to about using the InPho topic explorer for my text?

Jaimie Murdock

@JaimieMurdock

Hi Yvonne, the project is still active with some big improvements coming this month. Prof. Allen has relocated to Pittsburgh, so most efforts are there. I'm still a PhD candidate at IU, but am not currently on-campus

Both he and I are happy to talk though!

YvonneZhen

@YvonneZhen

Hi Jamie! Thanks for getting back to me so quickly!

I see. Do you know when you will be back on campus?

I am a PhD student in pre-modern Chinese literature with a minor in Didital Humaniteis. And I really hope to cooperate with you guys for my project of some medieval Chinese literary texts.

Please let me know!

I can also contact you via email if you prefer : )

Jaimie Murdock

@JaimieMurdock

Hey Yvonne, if you wouldn't mind sending me an email, I might track better - jammurdo@indiana.edu

Where communities thrive

People

Repo info

Activity