aaai/ainews

This is the NewsFinder software, designed to automatically crawl the web for news related to artificial intelligence, filter, categorize, and rank the news, and publish to a wiki, mailing list, and RSS feeds.

ainews's Issues

Use FeedBurner or similar

Push RSS feeds into such a service so that we can track usage of the RSS feeds.

Improve URLs

Add redirects. Change /AITopics/pmwiki-2.2.26/pmwiki.php/AITopics/CognitiveScience to /AITopics/CognitiveScience

Make bookmarklet for AINews submissions

This is after the new parser is put in place.

Consider pushing "AI Handbook but online" for AITopics

Investigate MEAD and ROUGE

MEAD: an automatic summarizer (http://www.summarization.com/mead/) -- also look at its "addons" for formatting text inputs

Acquire ROUGE (evaluator of automatic summarizers) from http://berouge.com/contactus.aspx and set up according to http://kavita-ganesan.com/rouge-howto

Include multi-part HTML/text AIAlerts

Allow a user's email program to choose which to show.

Fix Forbes source (ad block)

An ad is probably blocking Forbes (http://search.forbes.com/search/find?&start=1&tab=searchtabgeneraldark&MT=artificial+intelligence&sort=Date)

Fix that and add it to the sources list (http://www.aaai.org/AITopics/pmwiki/pmwiki.php/AITopics/NewsSources)

Figure out what files can be deleted or archived away

E.g. stuff in the Archive folder.

Weekly processing

Determine if AINews can be run weekly, and not daily.

Read the AITopics original article

In AI Magazine, referenced by Liang's paper.

Try multiple category membership

Create a branch of the software that puts a news story in several high-ranking categories. See if that helps the "appearance" of too many robot stories (a story labeled as "Vision, Robots" won't appear to be a robot story misclassified or dominating).

Recognize multiple story ratings from same user

In retrospect, the obvious solution to working around NAT firewalls is to use a cookie with a unique identifier embedded in it. The server can automatically generate it when contacted by a client that doesn't provide one, and return it with the page markup for the NewsFinder page. Thereafter, the browser will supply the cookie back to the server, so you can track the activity of an unknown user.

This also solves the problem of users moving around the Internet and connecting to Newsfinder with different DHCP-assigned IP addresses.

Consider setting up proper auth system

Right now it's stored in config.php. Look at PmWiki cookbook solutions.

Write survey about AIAlert and AITopics

Look at old survey first (report is in email).

Questions related to AIAlert:

• How do you use AIAlerts
• How broad a range of stories do you want? (broader range penalizes "active" research about robots)

Questions related to AITopics:

• How do you use AITopics?
• Have you contributed to AITopics?
• Would you like to contribute?
• What is the quality of AITopics (specifically related to your field of expertise)?

Output logs to PmWiki structure

Output crawling logs, training logs, classification logs, etc.

Spam submissions filter?

Suppose somebody submits a news article via the website, and Bruce is emailed but opts not to upload the submission on the wiki. Does the AINews software respect this decision or does the software still process the submission regardless?

Update mission, vision statements

vision stmt: only 1st sentence

strategic:

• relevant interesting news delivered

guiding principles: students first; 10-12 stories is good

deliverables: add reference to quality of news; also "feedback from users"

Produce another example execution

Check again the effect of my changes.

Find Tom's documentation

Check out video submissions page

Article, audio, and video submissions from http://www.aaai.org/AITopics/pmwiki-2.2.26/pmwiki.php/AITopics/SubmitNewContent

Test "goose" article extractor

https://github.com/jiminoc/goose

Document various algorithms

Use markdown, put it in the github wiki? (where is github wiki content stored?)

Switch to SVM for category membership

Instead of a centroid classifier, classifier articles in categories with SVMs; try multiclass (choose one category) and one-against-all (choose several good categories).

Capture outbound clicks

Email Tom & Liang about what's being done there (investigate first). Figure out a way to capture outbound clicks and plot it (put into Google Analytics?)

Find a screen-sharing solution

Mikogo did not work.

Consider automated journalist for AIMag

Like AIMag articles by Glick, in folder assets/; consider building an automated journalist that processes the AINews stories and summarizes them, for publication.

Set up booth at IAAI

Check out prior survey.

Look at Liang's documentation

Ask him if there is an updated version.

Put it on the site if it is not already there.

Consider adding link to email for ratings

"Rate this story" in email (as a link); also "This is irrelevant" links

Try A/B testing.

Test "tagger"

https://github.com/apresta/tagger

Extract content of prior articles

For all articles in archive, extract article content & keywords, built corpus.

http://www.aaai.org/AITopics/pmwiki-2.2.26/pmwiki.php/AITopics/NewsArchive

Take out newpagebox cookbook recipe

Is only included in wiki.d/Main.WikiSandbox

Check all crawlers

See if they actually pick up the stories.

Determine if newsspider cookbook recipe is used

Fix PCWorld parser

Also prevent crashing if any other parsers break.

Finish upgrade to PmWiki 2.2.26

Also tell Reid, Bruce about this upgrade.

Look at site search

Figure out why the Google Custom Search is not catching the abduction page, for example.

Look at AIAlert subscribers

Get a sense of who is subscribing.

Consider reviving the xword

See if it can be generated automatically (from the news & wiki content).

Mobile browser support

Determine if people using AITopics are using mobile devices.

Video support, etc.

Check out RSS pages

Make sure, in the code, that the AINews stuff is actually appearing, correctly. Also, what is the process for truncating based on time? Is it doing that?

Send message to Bruce to summarize meeting with Reid

Draft AINews vision statement, etc.

Mission statement, vision statement, guiding principles, strategic value, deliverables. Look at http://www.reidgsmith.com/Strengthening_Information_Management_for_Reduced_Risk_20081118.pdf slide 18.

Try a link checker

Do stemming on whitelist

Use NLTK to stem whitelist words.

Look at video integration

See how well it is integrated with the rest of AITopics.

Check sources on wiki

Make sure the sources list is up to date on the wiki.

Also ask Feigenbaum, others if they know of more sources.

Modify sources page

http://www.aaai.org/AITopics/pmwiki/pmwiki.php/AITopics/NewsSources

Say "Here are the AINews sources." Also show some other options.

Update The Times URL

Finish adding Reid's keywords

cognitive science (or cognitive sciences … May I assume that stemming will obviate the need for singular and plural terms)
brain science
cognition
induction
causal reasoning
automatic translation
automatic translator
machine translator
translation system
machine translation … or even translation, given that we are already selecting sources
expert system
poker program
co-occurrence of “computer” and “brain” or “computer” and “brain” or “computer” and “recognition” or “computer” and “think”
computer vision
machine vision
pattern recognition
pattern analysis
image recognition
image analysis
voice recognition
speech recognition
avatar
computer music
computer art
neural network
semantic analysis
semantic search
recommendation system
recommendation engine
artificial thinking … question … does NewsFinder remove quotes prior to analysis? I ask because the phrase that appeared was: artificial 'thinking head'.