Now that the basic concepts of predictive coding have seen some judicial approval, those who attended last week’s LegalTech New York exhibit got to see some fresh angles on the controversial technology.

Predictive coding — software using data analytics, guided by attorneys, to find relevant case documents — was all the rage at LegalTech New York in January 2012. By the end of the year, almost every major e-discovery company developed its own system, or licensed software from companies such as Content Analyst Co., Equivo Inc., and OrcaTec. A trend shaping up for 2013 are systems that offer alternatives to the current approaches.

Several companies began describing their approaches ahead of the conference and plan to reveal details in the days, weeks, and months ahead.

For example, Catalyst Repository Systems Inc. is one of dozens that license Equivio’s Relevance product. But at the show, Catalyst introduced its own predictive coding system, calling Insight Predict. "It’s probably as different as you can get," and is based on query technology developed in a joint venture between Japan’s Fuji and Norwalk, Conn.-based Xerox Corp., CEO John Tredennick said.

"The system’s built in a big cloud, where we’re ranking all your millions of documents, every time," said Tredennick, who revealed the project in August. "Some other systems use 10,000 documents and call that representative, which we question," he continued. "The whole key is we can do this against any volume of documents quickly, and you just can’t with the appliances."

Denver-based Catalyst unveiled its new application at LegalTech and plans to ship it by late February. Catalyst will continue supporting Equivio Relevance if customers request it, but, "Our focus is on Insight Predict," Tredennick added.

John Felahi, senior vice president of products at Reston, Va.-based Content Analyst, said his company is starting to differentiate itself by making its technology compatible with open-source data visualization software. Content Analyst’s predictive coding software is licensed by name brands such as iConect, Ipro Tech Inc., and kCura Corp.

Michael Schubert, vice president of software development at IPro, added that his version is due in April and will suggest data relevance decisions based on what other system users have previously decided.

Nuix also debuted an alternative predictive coding approach last week, based on technology from records management techniques and spam filters.

What if predictive coding isn’t necessarily the best way to perform a technology-assisted review? Andy Kraftsow, chief scientist at RenewData Corp., said at LegalTech in 2012 that he believes traditional keyword search can be just as efficient as predictive coding. Now Kraftsow said he’s got "a new trick" to make keyword search potentially more efficient than predictive coding.

Kraftsow acknowledged that keyword searches takes longer, and therefore can cost more, than predictive coding. But he said that’s because predictive coding gives incomplete results. To get the same efficiency as a human reviewer’s keyword search, a predictive coding process would need more time to configure the software than some users expect or than some vendors advertise, he asserted — leading to the same costs required by more traditional searches.

Something new, Kraftsow said, is that Austin, Texas-based Renew’s software now includes a hybrid approach to proximity search, which evaluates information’s relevance based on nearby words, and synonym search, which considers how words are related to each other. It also adds frequency search from highlighted sentences. The result, he said, is more efficiency without increasing the $350 average cost per gigabyte.

It’s not that Kraftsow is against predictive coding, he explained. He’s just fed up with the hype. "I’m an old guy and I kind of resent the nonsense that’s going on that predictive coding is the only way to accelerate review [and] that they’ve got to turn the coding over to machines to get the cost-and-time savings that they want," he said. "It’s just absurd, and it’s wrong."•