In 2004 we sold our company Polexis to SYS Technologies. I was the Polexis CTO, but had taken on the CEO title shortly before we sold the company. So I continued in a management role at SYS after the sale. That meant I had very little time to "get my hands dirty" in code. After a couple of years, I missed coding and I missed being an entrepreneur.

In mid-2006, I left SYS and started Appeligo (aka Know'bout). I recruited a few of the best developers I had worked with (Chris Exline, David Almilli, and Jake Fear) to work part time. Chris eventually joined me full time, investing his time for free as well. As is typical with startups, we created our first live web app, but discovered too many roadblocks to getting it to market; so we trashed most of that first app, and started over, creating [flip.tv][http://www.richkadel.com/fliptv "Video of flip.tv"].

We had some good opportunities to talk with VC's, and even present to a large VC audience at [VentureNet 2007][http://www.venturenet.org/previous.html#2007] but ultimately decided we weren't getting enough traction and funds were running low. It was time to get a job again, so we folded the company on December 31, 2007.

When my partners and I joined forces, we agreed (through a mutual Letter of Agreement) that if the company did not take off, we would post the code to open source. As there was no time limit and a possibility that someone might be interested in the code, as one of the few remaining assets, I held off for a while, hoping there might actually be a buyer for the code itself. Now I'm rectifying that situation and posting the code on github. Hopefully someone will find some of it useful.

I want to get this code posted quickly, so I'm not going to spend a lot of time going through the details. I may come back and provide more details later. If someone is really interested in this, I encourage you to use and/or contribute to this codebase. Maybe it will find a life after all.

The basic building blocks included a combination of Java and native C code to parse live streaming TV closed captions and Java code to periodically download and index licensed TV guide information. (Note that the closed caption extraction code is based on analog TV standards. Digital TV and closed caption capture capabilities were not available in a way that we could control from our Linux servers, and at the time, analog TV was still prevalent...and still is available to many cable subscribers. But digital TV capture would be an important upgrade for someone willing to take it on.) We developed that code and then built the applications to use the collected data.

Some of the remnants of the original app that were interesting but didn't make it into the final Flip.TV app are still here. The original app analyzed incoming closed captions for significant words and phrases, and then passed those significant terms to various search engines as if the user was trying to find out more information on the topic. Jake did an excellent job of adapting open source Natural Language Processing (NLP) code to generate reasonable extracted topics. We had a great time watching this work live, seeing search results with added context based on the words the actors had said seconds before on the TV. But to make it more than a novelty would have required changing the way people watch television, and it was too much of a stretch in 2006. Online video was certainly gaining in popularity, but closed captions were few and far between, and the implemention would have required a lot of agreements with other web app providers. So that's why we nixed the original concept.

The final Flip.TV app extracted closed captions from multiple channels in real time, compared the incoming text streams to standing queries from users (to generate email or SMS alerts), and then indexed the closed captions, along with TV guide information, to support on-demand queries via a google-like query page. The TV guide data, along with the time sequence information from the captured caption stream, provided metadata to support complex searches, and to also provide guided sub-query and related query links on the search page.

We used Lucene to index the closed caption text and TV guide metadata, and Chris came up with the idea to create separate search indexes for long-term data versus the most recent stream (I believe the default was about 1 hour). This supported the ability to update the short term index in real time, so users could query on dialog that had just occurred. As each period expired, a new short term index would be created, and the previous one would be merged into the long term index.

We also exprimented with video capture and conversion, and there are options in the config files to turn on video clip capture, to save short clips along with the caption data. There is an open source flash viewer and server (red5) that we adapted, so we could easily extend the web app to show brief clips from a program.

This software was originally managed with svn. There are some convenience scripts and maven scripts that may refer to that dependency.

The software builds with maven, and includes some JUnit tests.

The maven POM files reference internal and external libraries. I have a personal repo from my most recent builds, so if for some reason the external repo does not respond to the dependency requests, let me know via github mail, and I may be able to upload the old jars.

The TV guide information downloader, parser, and indexer (epg - for electronic program guide) depends on a circa 2007 formatted set of daily program information from Tribune Media Services. This was made available to us by paid subscription, and incuded proprietary information. Neither the format documentation nor any samples could be published legally by me, but you might find information from MythTV (the open source Linux DVR software), which as I recall may have made a deal with Tribune Media at one point.

As mentioned, the closed caption capture software and video capture software uses analog video capture devices. The closed caption software was developed against a generic interface, and adapted to both a Linux driver and a Windows driver. Our production software was deployed on Linux (CentOS), but we could also test on Windows using this software.

I can't guarantee this will work (and it probably won't), but you might start out trying to build the software with the following commands:

	cd cc4j
	maven dev clean deploy
	cd ../ccdataweb
	maven dev clean deploy
	cd ../epg
	maven dev clean deploy dist
	sudo ./installtarget
	cd ../showfiles/
	maven dev clean deploy
	cd ../websites/search/
	maven dev clean deploy

There are a few config files as well. Samples are in various "src/deploy/etc" directories.

Apache License, Version 2.0, appears to be a reasonable license for this software. Unless otherwise stated (which may be the case if certain external dependencies were purposefully or inadvertantly included in this distribution), I am asserting this Apache License, Version 2.0 with this initial open source publication of the Appeligo, Flip.TV source code.