LexisNexis Group (http://www.lexisnexis.com) announced today at the KMWorld 2001 conference that it was partnering with iPhrase Technologies (http://www.iphrase.com) to develop new natural language high-relevance interfaces for two leading online directories. LexisNexis will integrate iPhrase's One Step natural language service into its Advertising Red Books (http://www.redbooks.com) and Directory of Corporate Affiliations (DCA; http://www.corporateaffiliations.com). Insiders report that the partnership represents a major move by LexisNexis toward the application of more sophisticated end-user tools to retrofit its massive arrays of data. According to Tom Derry, who manages the division of LexisNexis Group that handles reference publications, the development is also consistent with a larger strategy of concentrating its investments on "proprietary, ‘must-have' information."
LexisNexis owns the two publications through its Reed Elsevier parent. Reed Elsevier Ventures, the venture capital arm of the multinational publishing giant, has made a major investment in iPhrase Technologies, leading a financing round last July that garnered $25 million for the new company. The current deal represents the beginning of a broader collaboration that could marry iPhrase's search technology with a wide range of Reed Elsevier's information services.
The iPhrase natural language search and navigation technology allows users to pose typical questions that the system can interpret to locate precise results. Its software can even recognize follow-up questions from users and answer them in the context of what has gone before. It can tap into all the fields available in a database structure and present the results in a variety of attractive, useful formats. In the course of adopting the technology, the content provider works extensively with iPhrase to customize the knowledge base behind the One Step system, helping it learn the types of questions users will ask, the jargon of the trade, synonyms, taxonomies, etc.
LexisNexis Group has chosen to apply iPhrase's One Step first to the development of stand-alone service to the two directories. The new interfacing, scheduled to become available early in the first quarter of next year, will work for subscribers to the data on the individual Web sites—http://www.redbooks.com and http://www.corporateaffiliations.com—but not for subscribers to the general LexisNexis service. Clearly, the Web routes for the two directories represent the most end-user-oriented outlets. LexisNexis Group does not plan to add any additional charges for using the new software interface. However, according to Derry, it will be available to subscribers through the Web sites and not pay-per-use through Telebase's bizlinkage service or through Hoover's online access route—at least not for now. "Ask us again after 6 months," said Derry.
The Advertising Red Books, produced by National Register Publishing, have been the bible of the advertising industry for over 80 years. They provide competitive intelligence and prospecting data through detailed profiles of some 13,500 U.S. and international agencies and information on some 24,000 U.S. and international advertisers who each spend over $200,000 annually on advertising. The Agency Database includes information on accounts represented by each agency, fields of specialization, breakdown of gross billings by media, contact information on agency personnel, etc. The Advertiser Database offers advertising expenditures by media, current agency, fiscal year-end and annual sales, contact information on key personnel, brand-name information, SIC classifications, etc.
The Directory of Corporate Affiliations provides corporate linkages for over 150,000 public and private companies in the U.S. and worldwide with company trees to affiliates, subsidiaries, and divisions. In addition to corporate linkages to the eighth level, its records include information on finances, assets, liabilities, sales, net income, net worth, number of employees, key personnel, contact information, business description, etc. The DCA product is currently in the midst of a push to expand coverage. Derry told me: "We are investing heavily in the database itself. We expect to grow DCA 22 percent in the next 6 months and improve its quality at the same time. We will take internal snapshots at a number of records, checking for age, completeness, and so forth, and measuring our progress against metrics. Internally we will have a multiple-touch strategy and ensure we look at each record three, four, or five times a year." Together, the quality expansion and improvement of end-user access software represent a heavy investment, according to Derry.
The most exciting aspect of this announcement may be its influence on the general market for advanced search interfacing. Sue Feldman, lead analyst in this area for IDC, points to iPhrase Technologies as having really "new technology. But now that it has been extensively tested internally by LexisNexis, that in itself constitutes a kind of validating process. Other people, other vendors may now decide to jump in, arguing that ‘if it's good enough for LexisNexis, it's good enough for …' This could be a significant market development for next-generation search technology, depending on how successful they are."
LexisNexis seems committed to the new technology. Derry indicated that this effort constituted only the first implementation, "a proof of concept" for LexisNexis. Diane Noble, managing director of Reed Elsevier Ventures, said: "As the owner of deep value-added information services like LexisNexis and Cahners, Reed Elsevier's business is dependent on unlocking the value of content for its customers. iPhrase has developed breakthrough solutions that will connect users to the information they want in one fast, intuitive, and accurate step. We're very excited about our investment in iPhrase and the strategic benefits it holds for Reed Elsevier."
Bill Pardue, CEO of LexisNexis' Corporate and Federal Markets, said: "This investment is another move in our continuing efforts to make our content easily accessible by customers. iPhrase is an enterprise-class solution—exactly the market we are serving with our LexisNexis Customized eSolutions."
The iPhrase One Step product suite supports the following functions:
The iPhrase One Step platform even supplies that critical element (so long requested by this searcher): a Feedback form. At the end of each search, users are asked, "Was the response satisfactory?" To which they may respond, "Yes/No/Partially" and be referred to a direct contact for further assistance as desired.
- Interpreting—Natural language processing targeting the underlying meaning of a user's question
- Retrieving—Gathering data from both structured and unstructured sources
- Presenting—Designing a specific, tailored response for each query, dynamically summarizing and collating results, and providing contextual guidance on refining or modifying requests for users
- Learning—Collecting and analyzing user questions and responses to improve content management
Describing the company's technology, Eric Sall, senior vice president of marketing at iPhrase Technologies, said: "The core of our product is a Smart Decision Engine, a customer-specific knowledge base we use to interpret queries and represent content. We are very different from others. Our software looks for the meaning behind the question." Sall particularly pointed to the software's presentation of results. "The number of ways to present results depends on different situations. We can create a dynamic table, which will very much be the approach in this [LexisNexis project] application. For example, we can identify the certain characteristics sought, convert them into a database query, pull information, and create a results page on the fly. If there are follow-on questions, we can interpret them in context and may add a column to the result table. The software can figure out if the second query makes sense in the context of the first; if not, then it designates it as a new query."
When I asked Sall whether the system could recognize when its information was insufficient or defective—in other words, could it learn to say, "I don't know,"—he said that it depended on the knowledge base and the tuning. However, he indicated one feature that might help. The Voice-Over feature is "information returned with the results that explains to a user how the query was interpreted, repeating it in terms of the context it mapped to, describing what it looked at and why. This helps the user understand…. If the system retrieves two ambiguous answers that seem equally relevant, it may return both."
Sall admitted that features designed to protect user interests, such as the Voice-Over and Feedback form, were not mandatory. Vendors acquiring the iPhrase One Step tools did not have to use all of its features. However, he indicated that iPhrase "works with customers on deployments. We want them to use the functionality the best they can, to help us with knowledge-base creation and tuning. By generating terms relevant to the dataset, this can improve the user experience."
As to the company's prospects with LexisNexis, Sall could not go into details, but he did say that iPhrase was "talking to a number of divisions for all kinds of different markets and needs, all with the idea of making access easier." He also indicated that iPhrase had some installations with other partners coming soon, including one with a major Web player.
LexisNexis Group serves the legal, corporate, government, and academic markets with online data services as well as print and CD-ROM publications. The LexisNexis service accesses over 3 billion documents from thousands of sources.
Former MIT researchers and business leaders formed the iPhrase Technologies company to serve enterprises by providing advanced technology for internal and external enhancement of information's usability. The company has offices in Cambridge, Massachusetts, and San Mateo, California. Its clients include Charles Schwab and CNET Networks, for whom it supports the Notebook Assistant service that provides information to users who are trying to decide which laptop computer to buy.