Fog Creek Software
Discussion Board

writing a paper on software developed?

I have been involved in writing a data mining system for information retrevial and text parsing of very large databases.
In course I have developed some heuristic algorithms combining and modifying other people's work for the requirements of the project.
In the end my results are by and large comparable with the software which is the industry standard in use, developed by some people at one of the world's top universities.
I am not an academic(BS in cs), but in the course of the project read many papers and came up with interesting approaches to various problems.
I think the work is quality enough that it can be published as a paper, unfortunately this is a real world live system developed for a company which owns the work.
Publishing a paper would be something which my boss won't definitely allow.
But it's something I feel will help advance my career, as the whole system has been developed by me from scratch( design,development and testing).
While it's not perfect, it rivals the industry standard software tool most of the time.
Is there any way I can publish a paper or article on this work? My boss will definitely say no, and I don't want to do anything unethical.
Anybody have an idea on this? the source code is owned by the company, but all the ideas,algorithms, techniques have been developed from scratch using a bunch of AI and statistical techniques.
Any comments appreciated

Sunday, February 1, 2004

Most employers rightfully expect to review anything job related that you publish, mostly just to ensure it doesn't contain anything proprietary or unflattering. Try selling the idea of a paper to your manager; it should be good publicity for the company.

Tom H
Sunday, February 1, 2004

You are right that your work is probably worthy of a paper. A lot of serious software development is similar to academic work.

However it would be commercial suicide for companies in competitive markets to disclose the details of their work. If you want to write papers, you need to get a job in a university.

Sunday, February 1, 2004

Is the ability of the algorithms something by which your company directly earns money? If so, give it up now. Think of the reaction if the head chemist at Coca-Cola wanted to publish a paper on "a new taste developed by combining simple household ingredients" [grin]

However, if the algorithms are tangential or even orthogonal to how the company earns profits (for example, their HR system or an accounting system) then there may be room to negotiate a paper.

Best of luck,

Sunday, February 1, 2004

No it's not a way that they earn money currently, but they might incorporate it with their revenue model

Sunday, February 1, 2004

You are probably correct that your work is worth of publication.

However, you need to be aware that most employment contracts state that intellectual property that an employee generates belongs to the company (even if it is developed in their own time). This is even the case in academia (in the UK, if not elsewhere). It sucks, but that's what happens.

It you give away company secrets, you may not only get fired, but possibly sued, too.

In the US, I believe that it is possible to patent ideas such as yours (check with an IP lawyer!). After getting a patent, you could then publish the paper, and would be protected under patent law. This has several benefits:

1. You get a publication!
2. You get a patent!
3. Your company possibly gets to license the technology to other companies.
3. Your company gets to claim it has state-of-the-art technology.
4. Your company may get interest from companies and universities.
5. Your company may get to collaborate with a university.

While I don't advocate patents on mathematical ideas (since that allows patent holders to prevent others from using mathematics, which is a very bad idea), in your situation, this may be to your benefit. If you enjoy this kind of thing, have you though about doing a PhD?

C Rose
Sunday, February 1, 2004

> In the US, I believe that it is possible
> to patent ideas such as yours

It is worth noting that patents are considered intellectual property, so any patentable idea would more than likely belong to the company.

Caliban Tiresias Darklock
Sunday, February 1, 2004

Maybe you could talk this over with a professor at a local university who works on database stuff?

Warren Henning
Sunday, February 1, 2004

If your boss wants to patent the ideas / algorithms he'll need to disclose most of the valuable IP anyways.  After that's happened, the incremental effect of a paper is minimal.

I'd convince the boss to patent the algorithms,  then once he's done, sell him on the 'prestige' of getting the company's work published.

Sunday, February 1, 2004

I don't know how it works in the US, but here in France, as long as you have designed something on company time, those ideas don't belong to you, but to the company who paid for them.

The contrary would be like paying a scientist to come up with a new drug only to have him leave and build his own shop when he got the formula right.

As sad as it can be, the product of your work belongs to the one who bought it. :)

Renaud Martinon
Monday, February 2, 2004

While I understand the company owns your ideas, I have problems understanding how are the creative people rewarded to bring new ideas to the table. You can't just expect somebody to give the company a very lucrative idea and get nothing in return.

I find myself in this position oftentimes when I have an idea that would save my company big money and I have to decide: do I give my idea to the company for free or just keep it safe and eventually start an ISV to collect some ROI?

Is this a lose-lose situation?

Monday, February 2, 2004

"""it would be commercial suicide for companies in competitive markets to disclose the details of their work."""

Oh please.  Researchers at Oracle, Sybase, IBM, and other companies regularly publish papers on their work on databases and darn-near everything else.  If it's patentable, they'll patent it, sure, but regardless of that, they certainly publish.

Phillip J. Eby
Monday, February 2, 2004

"You can't just expect somebody to give the company a very lucrative idea and get nothing in return."

A salary is something in return.

Monday, February 2, 2004

"A salary is something in return."

This is a good point and I will try to refine my previous assertion.

I can see two types of employee contributions:
1. The management has a specific requirement and the employee provides a solution. This is where the "work for a salary" belongs.
2. The employee comes with a novel and lucrative idea. The idea can be a better implementation, a new approach, a new line of products or a tool to increase productivity. Overall the company makes a lot of money after that idea is implemented.

If you don't reward somehow the employee no. 2 you either lose him or the employee will avoid bringing creative contributions altogether. Most likely both of them.

My company for instance does not care about workplace creativity. They prefer to throw more people at problems instead to reward in-house contributions and to save money.

Let's imagine you are Microsoft and you develop a new search engine. How do you retain the talent when the creative types can open their own shop or drive over to Google's campus?

Monday, February 2, 2004

Oh yeah, I knew I was forgetting at least one big company I'd seen whose researchers published papers...  Microsoft.  :)

Phillip J. Eby
Monday, February 2, 2004

If the company you are working for refuses to allow you as a professional to publish papers, then you seriously need to find another place to work. And I assure you that in the cases where you are clever enough to have new ideas and persistent enough to implement them and coherent enough to present them in a journal quality paper, there are LOTS of companies that definitely want you to come work for them and publish all the papers you can handle.

Not allowing professionals to publish is a monster sign of radically incompetant and unprofessional management. The only exception is national security issues. If its so valuable that the company wants to protect it then they should be getting it patented if they have the least bit of sense and they should actively be seeking and expecting you to try to get published.

What has the world come to that companies don't want their researchers to publish? That's just incredibly stupid.

Dennis Atkins
Monday, February 2, 2004

When the company allows the results to be published is a positive sign that the company is interested in creative developers not in code monkeys.

There are a lot of companies who don’t care at all but some others us good incentives – stock options, separate offices, co-patents, articles in professional journals- to reward creativity.

Monday, February 2, 2004

*  Recent Topics

*  Fog Creek Home