Quece Private Beta Launch: The Q-Phrase Pursuit of Natural Language Search
This article is an exclusive first look at Quece, the private beta natural language search project from Q-Phrase, the makers of the ConceptQ series of search products (written about in Q-Phrase: Contextual Relevance Tech Unlocks Deeper Value from SERPs).

This article, further, is the result of several email interviews and a phone interview of an hour and a half. Please forgive its somewhat sprawling nature, which follows this structure:

• Quece in Brief
• SEL/Q-Phrase History
• Reconstructions from the introduction-to-Quece phone interview with Andy Miller and Danny Espinoza
• Quece as Described by Andy Miller and Danny Espinoza in eMail
• The Q-Phrase Crew (from their resumes)
• Q-Phrase patents (with links)

Quece in Brief
Andy Miller, president of Quece's parent company, describes its function this way: "Quece interactively deciphers and continuously refines a user's intentions and thought processes to produce highly-targeted results not achievable with traditional latent semantic indexing and page rank techniques."

In my words Quece takes natural language user queries and asks questions (disambiguates) to better determine searcher intent. It disambiguates through instant messaging clients and is built around the chat metaphor.

Quece has been three years in the making and is built up from patents won in 1998.

Quece forced me to superficially investigate natural language processing.

SEL/Q-Phrase History:
I first heard from Q-Phrase when Google purchased the Orion patent and hired its creator in April, '06. At that time Q-Phrase president Andy Miller and Chief Software Architect Danny Espinoza held that Google's purchase of the concept-parsing Orion pointed at a coming shift towards more contextual Google SERPs.

At that time Q-Phrase had a single product line available to the public: Concept Q, a desktop app that enables users to unlock the structure and key concepts from a given body of information.

Concept Q creates instant Cliffs notes of whatever body of text you aim it at. Put a couple academic papers in Concept Q, let it rip and you'll have a stronger understanding of the papers' core concepts than the abstracts and the indexes could give you.

Reconstructions from the introduction-to-Quece phone interview with Andy Miller and Danny Espinoza

G: Some CEOs at major search engines (Ask for example) characterize searchers as lazy. What are your thoughts on searcher laziness given the work required in Quece to disambiguate down to the appropriate page?
Andy Miller: I think there is a lazy portion, but the bigger portion is unsophisticated; they don't know what they're supposed to do to get the results they want. A user's typical reaction to a search in Google looks like this: "all I know is I typed in Cancer and got back a billion sites." They're uninformed, not lazy.

G: Help me understand the core technology differences between Quece and ConceptQ.
Danny Espinoza: They are certainly different technologies. If 0 = "I have an idea" and 100 = "I've maximized the potential" then we're at 80 with ConceptQ and with Quece we're between 5 and 10.

When we reach Quece's full potential, there will be a significant opportunity to challenge current keyword-based search engines.

G: QPhrase, ConceptQ, Quece. It's hard to tell these apart. Why the separation?
Andy Miller: We're focused on keeping the companies and the patents they're built around legally protected.

We will merge the technologies as they mature.

G: What about APIs as a growth strategy?
Danny Espinoza: APIs are definitely on the feature list. They're something we've thought of as an enterprise thing though, where a company had a proprietary database that only their programmers understand how to query... we could provide some kind of API for their data.

In the consumer space we can't provide appropriate developer and tech support. Right now we have to be careful about where we spend our development time.

G: Talk to me about funding.
Andy Miller: First off we have a revenue component with ConceptQ. Secondly we have a group of Angel investors who are excited about the idea of new search technology and are eager to be involved in something that could significantly alter the search landscape.

G: Where does QPhrase fit in an industry dominated by Google:
Danny Espinoza: The state of search has remained the same for a long time. It appears to be that there hasn't been a lot of changes in the way that they do
rankings or SERPS since inception

They're clearly doing keyword search better than anyone and they've hired hundreds of PhDs - they're not going to just rest on their laurels.

We look at them and try to understand who they're hiring and what they're trying to do, which is why we found the Orion purchase interesting. We strongly suspect that Orion approaches contextual relevance the same way that WE do it. Google has an amazing snapshot of web - their index is the perfect playground for algorithms...

The problem we solve now that isn't currently addressed by Google, is how to extract contextual relevance quickly. Clearly the infrastructure isn't in place yet for Google to introduce this level of complexity into their search, so there is a window where this problem can be solved before Google solves it.

Quece as Described by Andy Miller and Danny Espinoza in eMail:

Why develop this out for AIM or instant messenger?
The primary driver behind Quece was to build a search application that would eliminate the frustration associated with traditional keyword search engines.

Specifically, users can become frustrated when their keyword search returns thousands, if not millions, of results and they must spend time clicking links - often at random depending on the quality of the snippets - to validate whether or not a result is something of interest.

In our opinion, the only way to provide more accurate contextual results is to mirror human interaction. If you think about any conversation, one person makes a statement and another asks clarifying questions until each person understands one another.

While this was the premise behind Ask Jeeves and other Q&A search applications, none of those solutions understood how to properly “chat” with the user. Leveraging Q-Phrase’s patent portfolio, we developed Quece to do just that.

Given that Quece works by engaging the user to discuss topics of interest and contextual search results, AIM/IM seemed like a logical platform to integrate with Quece via our QueceBot. And although QueceBot is available via AIM, users can also interact with Quece through the Quece Search Portal (www.quece.com).

Doesn’t this limit the full functionality of your product?
Not at all, in fact, chat functionality is the cornerstone of Quece technology. As the user “chats” with QueceBot via AIM, Quece provides a link to view results every time it responds to the user.

This way, the user may elect to view results anytime throughout the course of the conversation. This is actually a very important point because it enables the results to shift as the conversation deviates from the original scope of the search query.

When the user clicks the “Results” link from within AIM, a Quece One-Page Result screen is opened, alongside AIM, in the user’s default browser. The Quece One-Page Result screen displays the categorized results associated with the user’s QueceBot conversation.

Contextual relevance in the semantic Quece?
Quece uses search result content to focus its discussion with a user around the topics related to the user's search query. However, this approach to contextual relevance is unique in that it uses a semantic analysis to understand the terms that are related to the original search query.

Therefore, instead of calculating a statistical baseline, Quece builds a custom dictionary that defines the concepts that are expressed in the search result text (e.g. snippet parsing).

We think that both of our approaches to contextual relevance, statistical in our ConceptQ and CQ web products and semantic in Quece, are important to solving the problem of understanding the concepts that define many search queries. And, it's only in understanding what search results mean that can we try to improve search relevance.

Q-Phrase patents:
Device for Storage and Retrieval of Compact Contiguous Tree Index Records
U.S. Patent 5,829,004

Method to quantify abstraction within semantic networks
U.S. Patent 5,937,400

Topological Methods to Organize Semantic Network Data Flows for Conversational Applications
U.S. Patent 6,778,970

Semantic Network Models to Disambiguate Natural Language Meaning
Application

The Q-Phrase Crew: (from their resumes)
Lawrence Au: Chief Scientist, Q-Phrase
Au is an expert in large-scale software development, computational linguistics and artificial intelligence with over five granted U.S. patents, three of them affecting the field of computational linguistics. With over 25 years of cross-disciplinary professional experience, he combines deep research, commercialization and team leadership experience to cross-fertilize, validate and deliver leading-edge computer applications.

Andy Miller: President, Q-Phrase
Miller is a marketing and product development executive with over 13 years of professional experience and a proven track record of significant contributions to world-class financial services and technology driven organizations.

Danny Espinoza: Chief Software Architect, Q-Phrase
Espinoza is a software designer and application developer with over 20 years of professional experience. His expertise is in providing full life- cycle development from conception to delivery of sophisticated desktop application software.
Garrett French - MarketSmart Interactive | permalink | +marktd |