When Google pushed into the enterprise search market with its pizza-box-shaped Google Search Appliance, the market and competitors took notice. The new appliance offered the simplicity of the Google Web search experience with fairly affordable pricing. However, many said it couldn't handle the complex environments and information demands of enterprise searching. Many high-end search systems offer concept clustering and other advanced contextual capabilities to increase precision, as well as other sophisticated business intelligence capabilities. Now, one enterprise search vendor has found a way to improve upon Google's keyword search capabilities by bringing linguistic understanding into the mix and offering Google Search Appliance customers an augmented solution for enterprise search operations.
Inxight Software recently announced a partnership with Google to integrate their respective enterprise search technologies. Companies and government agencies that use the Google Enterprise Search Appliance can now use Inxight's federated search solution, SmartDiscovery Awareness Server, to federate searching across multiple sources and filter the search results set by categories. They can also use the Inxight SmartDiscovery Analysis Server to automatically identify concepts, people, organizations, places, and other information, providing structure to unstructured text as XML-out for use in routing, categorization, search, and business intelligence applications.
I recently sat in on a Webinar (sponsored by Inxight and KMWorld magazine) so I could hear about the Google-Inxight solution. According to Catherine van Zuylen, director of product management for Inxight, Google offers fast, easy-to-use search indexing but can't handle many tasks required in an enterprise environment, such as looking for information not contained in a Google index (full-text analyst reports, existing Verity/Autonomy/FAST indexes, Salesforce.com data, PubMed, LexisNexis, patent filings, SEC data, etc.), monitoring competitors' Web sites or the content of blogs, or accessing internal corporate skills information.
Inxight SmartDiscovery Awareness Server is a federated search, entity filtering, and alerting solution. It allows users to query just once for information locked in existing (multiple) Google Search Appliances, FAST or Verity/Autonomy indexes, Documentum and other systems and combine it with results from subscription sources, the open Web, patent databases, news sources, SEC filings, analyst reports, and other information sources to deliver a single unified screen of information. It handles 600 out-of-the-box sources and can be customized for additional sources. It lets enterprise customers monitor brand chatter from a variety of blogs, news articles, and other sources and be alerted when competitors' Web sites change. The system offers both page and search term alerts. It can even suggest sources a user might not know about.
This is fairly sophisticated stuff. Here's what it does:
- Translates and sends a query out to each source
- Manages passwords and authentication
- Compiles results
- Re-ranks for relevance
- Summarizes and categorizes
- E-mails you when the search is complete if you don't want to wait for complex operations like querying 100 sources at once.
Entity (people, companies, places, etc.) and concept filters cluster results documents for easy and fast navigation. The real difference comes down to understanding the language and context of a query. A keyword search system like Google's can't differentiate the meaning in the following two sentences:
"Bank of California floods mailboxes with checking account offers." "Banks of California river breached in flood."
Inxight's technology understands language on a deep level (identifying people, companies, etc.)—in more than 30 languages.
Inxight not only handles extensive entity and concept extraction out of the box, but it can accommodate custom entities to cluster, including list-based (terrorist names, project names, etc.) and pattern-based entities (SKUs, chemical compounds, etc.). The product also works with pre-defined taxonomies. A general news taxonomy is included with the product. Users can import taxonomies from Factiva, Intellisophic, KAPS Group, etc., and can plug in custom taxonomies created with SmartDiscovery Analysis Server.
Inxight is also about to announce a new version 4 of ThingFinder, its faster and more advanced entity extraction solution. It is available as an SDK and will be available as a component within SmartDiscovery Analysis Server in Q2.
By the way, according to van Zuylen, more than 550 people participated in the Webinar that day, and the company received nearly three times the normal number of follow-up inquiries. She said that Google and Inxight have a significant number of overlapping customers and they are seeing interest from both sides.
When I asked whether Google might incorporate Inxight's clustering technology in its general Web search engine, van Zuylen explained: "While Google has been experimenting with clustering since 2004, they have so far not incorporated it in their standard Google.com search engine or their Enterprise Search Appliance. However, we have seen significant interest from both enterprise and government customers in combining Google's search index results with Inxight's entity/concept clustering technology."
During the Webinar, van Zuylen showed a Desktop Search prototype of Inxight working with Google—at this point dubbed InXbot. The developing product is targeted for possible rollout in Q3 or Q4. If you're interested in learning more, the Inxight-KMWorld Webinar is archived and available for 90 days at http://webinars.kmworld.com/inxight/16feb2006.
In other news this week, Inxight and Verity (now owned by Autonomy) announced that their contract and trade secret dispute that had begun back in October 2004 has been resolved out of court. While terms of the settlement were not disclosed, the resolution will allow each party to continue to serve and support its existing mutual customer base without disruption.