Ideas

I am rather busy at the moment but I could lend some programing assistance towards the end of the month.

Others here are also programming inclined and might be able to help sooner. Software guys, chime in. There is a pretty well defined set of requirements and very well defined problem space.

--------------

tnt1842, if this is made it probably should start with whatever type of email you are using and then branch out to oter providers.

Are you using a free mail account like gmail, yahoo or hotmail? or is this through a work email (i.e. through Outlook). The CH business SideFinder already has some code working to tag outlook messages, I am sure they could modify their engine to accommodate this task. Might be worth pinging them to see. (User ChuckNorris runs SideFinder)

I had an underscore missing on the user name, it is Chuck_Norris.

Or just go here: http://www.cambrianh...view/seedomainbelow/

From a programming standpoint, it's all pretty well defined except #2. If you really want to do some kind of linguistic digestion to produce a meaningful summary that is not already existing verbatim in the text, that's a pretty sophisticated task. Many (most...) human writers have an extraordinarily hard time accurately summarizing text, and having a computer do it for you requires that the software be able to parse the text, assign meaning to each chunk, determine what the dominant meaning of the entire text is so it can discard the less meaningful pieces, and then create new text describing that dominant meaning.

I've looked at exactly the same task for another project but haven't gotten to the point of implementing it yet. I don't think it's impossible and there are some very good linguistic processing tools available, but from what I've seen so far it is not a straightforward task.

Now if all you need is tagging, which only requires identifying basic keywords and synonyms, the task is significantly easier.

Have a feeling that I have read about a similar tool being posted here at CH.

I do not know how you would make money

I agree with Micco.

The idea is good except for some technical problems as stated by micco but you can still make this feasible. Good luck!

what about instead of sending emails you just both type onto a shared google doc? and make a no-back-editing rule.

hm, i can imagine another way of this happening, where there's an online document that could have an email address, and then you both always cc the document on your emails, and it takes only the newest response in the string and plops that onto the bottom of the document. the 'document having an email' thing is already there - that's pretty much what backpack does, in a quite sophisticated way.
http://backpackit.com/

these options are probably out for your current predicament though - looks like you'll just have to buckle down to some serious editing :(

doublelibra: you could easily set up a shared document or update system like you describe, but then you have to know in advance what conversation is going to be meaningful and worth keeping. Most of these good threads start as something mundane and evolve over time. Are you going to start using a documenting system for every email you send? Would the candidate tnt1842 was corresponding with have even participated if it had been a more formal document and not "just" an email exchange?

I think for most situations, you need a post-processor for this because you can't know in advance what will be worth keeping.

Agreed, post processing is sometimes the only way.

I would think that sometimes the only a snippet might be useful from an email and you stitch many of those snippets together to tell a great story. And yes the summary could be the difficult part. At first you could just start out with a keyword density parser. (like how a web search spider summarizes a page) It wouldn't have to be perfect, just a reminder of what was in the message.

You could even take it a little further and start trying to divide up specific content. For example you could extract lines that have "I think", "I believe" to identify sections with opinions. You could further and pull out any messages in the candidates replies that contain "my", meaning that they refer to something that belongs to them. Pull out text that has "your" in your responses so you can capture things you are saying about the candiate.

There are so many ways you could cut things up, though I imagine those would be put in place later once the basic framework was established. (i.e. focus on txt1842's idea description, not my suggestions for additional features)

I think that the technical issues already pointed out are not trivial and could unfortunately render this nice idea unfeasible.

There are also some privacy concerns here:

a) When you say the user can enter several email addresses, do you mean addresses belonging to the user or to other people as well? If other people's email archives will also be searched, you will need to ensure that their email system supports your app, that they also have the app installed and that they give permission to search their mailboxes.

b) Even if you were only searching your own mailboxes, I expect you might need the other parties' permission to publish their contributions in a book or online - since at the time they contributed to the conversation, they did not do so with the intention of having it published.

How so?