Animesh Pathak's Crazy Ideas

A [small] idea to [further] improve Google Docs for Collaboration

2011-03-22T15:25:00.000-07:00

I love Google Docs. I use it for a lot of things --- to store recipes my mom sent to me when I was in grad school, taking instantly-ready meeting minutes collaboratively, as well as discuss paper-writing plans with my colleagues, among others.

When Google released Wave, I loved it; but sadly people like me were in a minority and the project was killed. However, I am very happy to note that Google has kept on pulling the much-liked features of Wave into Docs, including instantaneous collaborative editing (April 2010) and, most recently, advanced discussion/commenting features (March 2011).

However, there is still one extra feature which I wish they could add, and this one would not even take much coding -- just some basic JavaScript. Here is the idea:

I often use Google docs in a conference call. It is a great way to conduct meetings, with attendees posting their questions directly in the doc, and updating it with answers as they are agreed upon in the meeting. The same thing goes for action items. And this way, the minutes are ready at the end of the call, thus saving a TON of time.

However, what I miss is the "follow $user's cursor" feature, using which I can make my screen auto-scroll to the portion of the document that my colleague is currently seeing on his browser. This would help immensely when people on a conference call are trying to read/discuss a document together, and thus need to be on the same page, literally :). [ To those who have played Quake III or other multiplayer FPS games, this concept should not be new :). ]

Implementing this should not be a problem either. The Docs page already sends cursor positions periodically to its servers, we just need to add screen boundaries to this information in some manner. On the receiver's side, it will simply involve some JavaScript to scroll the screen to the appropriate position.

So, what say?

Who Tweeted That? A Proposal for inter-disciplinary Research

2011-01-09T03:16:00.000-08:00

[No, this is not about the latest means of communication, rather, about something which has been in vogue for about 150 million years.

Motivation and Context

Often as I am walking in my neighborhood or near my lab (which is next to a forest), I hear bird calls, and wonder what bird is trying to communicate with its peers. Now, if it were a celestial object, I would have whipped out my phone and used Google Sky Maps, but alas, there is no such app for bird calls. Hence this proposal/idea.

The proposed system

My idea is rather simple, at least in terms of a user interface. Quite similar to apps like Shazam which let users find out more about a song they are hearing. Basically, it should work like this from the user's perspective:

I install the app
When I hear a birdcall, I click "record" and point my phone toward it
When done, I click the "upload" button
I get the information about the bird (Name, Wikipedia link, samples of bird calls. etc.) on my phone :)

Paying for the app

Now, apps like Shazam make money when people end up buying the songs that they recognize using the app. For the app being proposed here, there is no such revenue stream. However, the users can 'pay' by uploading the location of the phone (and consequently, the bird) when they upload the bird-call. Of course, the system should not store any other data which is personal, but the location can be very useful to secure funding to create this system (see below).

The players behind the curtain

I feel that in order to provide the functionality proposed above, several unanswered questions will need to be addressed by specialists of several disciplines:

Ornithologists, who can provide the information about existing samples of bird calls, as well as locations where they are commonly found, so as to quicken the search. They will, in turn, gain valuable data about where the various bird calls were heard (by virtue of the location uploaded by the users), which might lead to interesting discoveries.
Computer Scientists, who will have to solve the problems of performing the search at large scales while being responsive enough so as not to bore a smart phone user :).
Mathematicians/Digital-Signal-Processing researchers, who will have to propose ways to model bird calls for easy searching, as well as propose compression techniques suited for audio which is neither voice nor music.

So, do you think the above idea makes sense? Is there an NSF/NIH/EC call where this can be proposed? Do you know of researchers who would be interested in working on this? Let me know in the comments.

-A

One way to cite them all

2008-04-24T00:28:00.000-07:00

Being a researcher, I am always on the lookout of new work by fellow researchers. To that end, each of us usually "knows" who else is working in the area (people we have met in conferences, groups recommended by our advisors, researchers found on google). And we visit their websites once, reading their work, and trying to make sure that our own is a bit different :).

Invariably, the submission time of a new paper is a time of tension. Do all my experimental results look right? Did I polish the text enough? And finally, did I consult and cite all the work of the guys who matter? [don't believe me? see this for reference]

So one goes on the google process again, scouring through well and ill-maintained university websites, google, scholar, citeseer, ~~imdb~~, you know... the works.

Wouldn't it be nice if there was a service that would let you know whenever a new paper had been published by someone whose papers you are interested in? Note that checking their websites on a daily basis is not a good idea, since I might be interested in a HUGE number of researchers, and they might publish only sporadically.

And that pattern led to think of all my friends who maintain non-regularly-updated blogs. Back in the day, it was a pain to check their blogs once in a while, but since I started using Google Reader, I don't need to do it anymore. If there is a new post on Joie-de-vivre, I would know, thanks to the magic of RSS/ATOM feeds.

So, without further ado, here is my new crazy idea. Create a new site [or downloadable software for a site, kinda like wordpress], where:

each researcher has a login
when a paper is accepted/published, they can fill in a simple form, essentially the contents of a BibTeX entry, and upload/link to a PDF/PS copy of their paper
each user gets a unique url for their "paper feed", which they can then publish prominently on their websites, encouraging others to subscribe to it. There can even be different feeds for each "track" the user publishes in, and each paper can belong to multiple tracks
when a new paper is published, the site/software automatically creates a nicely formatted feed entry, which the subscribers' readers can pick up. It can even create a BibTeX entry :)

Now, I am sure there are similar things around, and there are researchers who publish an RSS feed of their work, but nothing mainstream that I have come across.

Google, Citeseer, Arxiv, facebook, are you listening?

So... dear readers, what say? Kaisa laga mera naya idea?

Next step for google video

2006-02-10T16:26:00.000-08:00

The google videos site is super cool, and a lot of you are going to it nowadays to see zany, funny and cool videos.

However, the searchability is not so good. For example, I may have looked for the song "My name is Anthony Gonzalvez", only to not find it but then later find it with the "Amar Akbar Anthony" tag.

So, there is a problem, too little meta-text... uploaders do not have too much time to type.

But viewers do... and that brings us to my next idea: Improving google videos using public editing.

In short, google should let people post comments/annotations to the videos they see on google videos -- I am sure people would not mind typing out while they are seeing a video.. and that will help boost the metadata to search the video with.

Of course, there can be malicious uses of this "wiki" method, to which I propose 2 solutions:
1. Allow only people with Google IDs to post. This limits spamming and violators stand to lose their accounts!
2. Better still, when someone searches, allow them to search from "uploader's tags" or "complete text" using a radio button next to the search-box. Simple.

So.. what say? [are Page and Brin listening?]
-A

Idea #4: Public crime reporting using cellphones

2005-10-03T18:05:00.000-07:00

Acknowledgement: This idea is due to Prof. Gandhi Puvvada, who suggested this as a reaction to this incident.

Think of the scenario : a murder/mugging is happenning in a street in Mumbai. Normally, noone will come forward as a witness, so the police cannot book the culprit.

However, with the current ubiquity of cellphones with cameras in India, members of the public can easily and anonymously contribute to the booking of the offender.

Step 1: Take picture of act with camera phone
Step 2: Go to a well-known-website [to be launched] which hosts pictures of incidents.
Step 3: Create a "case" on the site and upload the picture.
Step 4: Done.

The site can be set up to notify the relevent police deptt and/or major news agencies upon upload of the picture. Also the site can be run by someone in the USA so that the person is out of reach of the desi hooligans.

So, what say people? As always.. your comments are eagerly awaited!

Luv and Luck,
-A

How to tell the sex of a visitor to your website?

2005-04-03T17:32:00.000-07:00

WARNING : ADULT CONTENT... above 18 only.

I actually got 90 hits on my blog yesterday (April 2). I wonder what caused it ?!

That set me .. and more importantly, M, thinking about how many of them were heterosexual females, since that is the species I am on the lookout for. M suggested that I put something up on the website that would decipher that info for me.

So here it is - the first audience-speak of this blog. Dear readers, please remember to read all the comments and append your idea at the end. If you want to, please leave your email address behind in the comment, so that due credit (or brickbats) can be given.

Here's my idea #1:
Post a link to "sexy pictures of [insert name of hot female celebrity here] " on the blog. Then count ratio of visiters who visit the link. The only problem is that lesbian girls may also click on the above. But come to think of it, I should not care about it since they won't be interested in me anyways :-).

Idea #2: Post a link to "online window shopping". The number of visiters going to that link should give a pretty close estimate of the number of females visiting the site [or is this a mere lower bound, since some guys might also be interested in shopping (ewwwww) ].

Finally, and the best (?) idea:
Just put a poll on the site saying "are you male or female?". Trust in people's honesty, and assume that the number of whackos giving wrong answers is the same across the sexes. Then the anomalies will cancel each other out and we have the answer to the question about my life, universe and everything.

So, dear reader, how did you like this one... crazy huh?

I know all the above methods have loopholes. Maybe a method that uses a combination of the above strategies will work best. I don't know. Do you?

Please do post your comments and let others know what you think about this. I am sure this will help a lot of people who have their blogs :D. Also let me know if you have thought about this problem before.

waiting for comments... and signing off.
-A
P.S. So, what is your sex?

Idea #2: The birthday virus!

2005-03-04T16:57:00.000-08:00

Hi all readers,
I have been asked about my next idea for some time now, so hear goes:

Since yesterday (march 3) was my birthday, I received a lot of e-cards and emails [thanks to all of you who sent them]. However, I realised that the occasion made me check my mail more frequently - and more prone to opening attachments.

So, imagine that it is your birthday and you receive an email from one of your friends with Happy Birthday as the subject and a generic birthday message in content. Attached is a zip file which (allegedly) is a nice card for you.

Now, ask yourself as to how likely you are to open that attachment, something that you will not do in a general scenario to a random email.

Scary, isn't it? The thought that this can be exploited by a virus/worm/trojan.

The algorithm is very simple:
1. Scan infected machine for calendar app (MS Outlook, Mozilla Sunbird etc.) and glean out birthdays and other important events (anniversaries etc).
2. Send a copy of the infectious program in a zip file or something with a generic birthday (or other relevent) message to appropriate people in the address book.

That's it!

Major Con:
1. The virus cannot spread so fast since only so many people you know have the same birthday.
A. Well, if I get a card even today or tomorrow from a friend, I still have a high probability of opening it, 1-2 days after my birthday, so that increases the infection circle. Also, a virus sitting on a person's machine and not getting cleaned will have the chance to wish a lot of his/her friends in the months that it stays in that machine. Basically, given the high probability of infection, maybe this way of spreading can be an extra way by which the malware can hop hosts.

So here it is people, be careful of those birthday emails. I have not seen something like this yet, but very soon... someone might decide to add this to their arsenal.

As always, comments are most welcome. The ones on the previous idea were pretty good!

cheers
Animesh

First Idea Online!! a Novel method of Employee Evaluation

2005-01-10T17:36:00.000-08:00

Hi all,
Here goes.. and I got this during the CS599 lecture today. I was thinking of what may actually be taking place at google... if you have seen this earlier, please post a link to it as comment.

Suppose that a manager has some (say 10) employees who are all capable of doing similar stuff (e.g. software engineers writing code). (S)he calls all of them and gives them a problem. Then each person in the team is asked to "bid" on the amount of time they will take to complete that job (definition of "completeness" pre-ascertained). The 5 (or some other number) people who claim to be the fastest are assigned that job. When they do it, they are awarded points which can be positive or negative depending on how timely and good their work was.

The method is very similar to a card game called "french bridge" I played as a kid where you were given a set of cards each and had to bid how many hands you will make. Accurate bidders are given +ve points and over- or under-bidders are penalized.

I believe this can be used for employee evaluation someplace, if not used already.

Any takers?
Please comment on this.. I am waiting for feedback.

Animesh