iTrieve

iTrieveTM is a new type of semantic search application: one that retrieves exact answers from plain-language requests, whether from PCs or from smart devices.  While not competing with Google as the "goto app" for common information search, iTrieve is designed to be the user's "goto app" for actionable information.

Semantic Search (for Exact Answers) has finally arrived

People who see the phrase Semantic Search either concentrate on the word "Search" and think of current (Google-like) search technology, or if they're technical the word "Semantic" may steer them to a second branch of semantic technology (Ontology).  Our iTrieve technology actually includes both branches, but we'll concentrate on the "Search" part of the equation first: We look at Semantic Search as a methodology to retrieve a higher degree of exact answers than conventional search engines.  Until now, this branch of Semantic Search has been more experimental than real, and here's why: since Exact Answers (facts) generally reside in relational databases, Semantic Search needs to incorporate a new type of user oriented database access technology -- one that retains a true "end user" access method (note: SQL is NOT an end user access method).

iTrieve's inventor, a true pioneer in Natural Language Processing (NLP) and end user computing (inventions listed here), has solved some inherent problems that have kept Semantic Search from reaching its long-envisioned 'holy grail' status.   Using iTrieve, people will finally be able to retrieve Exact Answers from computers and mobile devices, and in the near future will be able to command and query other "information providing" devices (TV sets, etc.) and automobile systems.

Our Breakthrough Solution

iTrieve accomplishes its goal of retrieving Exact Answers by coupling semantic plain-language requests with a proprietary "visual request specification" interface: (Request Specifications), together with an invention called Rules-based Join Navigation.  For a Technology Overview, select Technology on the Navigation menu.

A key design feature of iTrieve is combining the power of natural language search with a novel Visual Request Specification method that serves as a "backstop" to natural language: if the semantic parser doesn't fully understand the request, the user is prompted to "finish" the request specification visually.

Close window

Technology


iTrieve incorporates advanced technology in both its Semantic Search process and in its underlying Taxonomy-based Metadata Architecture.  Both of these underlying technologies are described below.


I. Semantic Search Technology


iTrieve's Semantic Search product is based on the combination of three different technologies: 1) Semantic Search, 2) a visual Request Specification interface (RequestSpecs), and 3) a patent-pending innovation called Rules-based Join Navigation.


Why is it necessary to combine these three technological innovations?

Research in Human Computing Interaction (HCI), as it relates to end user information retrieval, has demonstrated success with two main end user interface methods: plain-language requests (speaking or typing) and visual methods (point and click; drag and drop).  Each user interface method has its own problems:


1) Speaking or typing 'plain-language information requests' (Semantic Search) suffers from major factors that often result in the Search tool not being able to understand and/or carry out the request:

details on Semantic Search problems>

2) Visual interface method: Many database query tools (B.I. tools) show the user a 'tree' of information objects (database tables and columsn; conceptual entities and attributes). The user can point and click (or drag and drop) on the information objects he or she wants in a query (report or graph).
>>details on Visual User Input problems>>


iTrieve combines its Semantic Search technology (speaking or typing a request) with its intuitive Visual Request Specification interface, giving the user an iron-clad way to "specify" a user request correctly.
>>details on Request Specification>>


The third innovation necessary to get Exact Answers from end-user requests, Rules-based Join Navigation, is a technological feature that works "under the covers": the regular users of iTrieve are not aware that it's happening. Nevertheless, this feature is a vital cog in the magic of iTrieve's ability to retrieve exact answers upon request. >>details on Rules-based Join Navigation>


II. Taxonomy-based Metadata Architecture


iTrieve's advanced metadata architecture consists of hierarchical concept layers: a taxonomy.  What this taxonomic metadata structure means to iTrieve's users is a unique type of ETL ("Extract, Transform and Load") capability, baked into the product itself (see the Usage Benefits menu for more details).


Close window

Close window

iTrieve App Store


The iTrieve Semantic Search product has been "mobified", with special browser screens tailored for various classes of smart phones and tablet PCs.  The iTrieve App Store offers a low-cost way to "showcase" an iTrieve application or vertical market "template app" to the world of mobile device app buyers.

The makers of iTrieve are planning to provide a dedicated cloud server with replicated iTrieve Enterprise Edition customer application databases, or subset "showcase" portions of these application databases, and will share a portion of revenue from apps sold through the App Store.


Close window

Close window

Cloud-Client Architecture


iTrieve is a web application with no software downloads needed.  Its user-facing app has been certified across most known browsers.  The "back-end" of the application is hosted on a cloud-based server, but the overall platform architecture is "distributed", in the sense that iTrieve's cloud-based Metadata Repository contains the structural information (schemas) of clients' target databases, while the actual target databases remain on client-based servers or on the clients' own cloud host servers.

- Metadata Repository: iTrieve's metadata repository captures all client database metadata, through a one-time setup process called Semantification (described under IT Admin). This metadata repository supports "multi-tenancy", so that one organization (or VAR) can provide user access to different organizations or divisions.

- Targeted Client databases: Each client organization sets up one or more "targeted" databases to be accessed by its users through iTrieve. These targeted databases usually reside totally on-site behind the organization's firewall, but can reside on a cloud platform. User access through iTrieve is "read-only", enforcing data security (also see IT Admin).

- Request Processing: When a user has selected or created a User Request and has pressed an action button to generate and execute a report or graph, the target database metadata provides the necessary information to iTrieve's Query Generator to generate the proper SQL command, which is then sent via VPN to the client's target database. The result set of this executed SQL command is returned to iTrieve, which formats and presents the appropriate report or graph to the user.

- Client IT admin control: The client organization's superadmin controls all access of its target databases (see IT Admin).

Close window
Close window

IT Admin


iTrieve basically allows "IT-free" usage by non-technical users, and that's true for public or social databases (all IT work is handled by iTrieve's own IT staff). For corporate usage, however, there are some IT processes set up and administered by an assigned superadmin, usually an IT staff member of the corporation. These admin processes included the following:

- Setting up a VPN: The client corporation signs up one or more corporate databases to be accessed by its business users. Each targeted corporate database is accessed by iTrieve through a VPN.

- Semantifying corporate databases: Utilizing the VPN between iTrieve's cloud-based metadata repository and a target database, the superadmin invokes a front-end process called semantification, in which the superadmin essentially connects to the database through a normal URL database VPN connection. At that point, all of the target database metadata is added automatically to iTrieve's metadata repository.

- Managing Organization-level Data: The client organization's superadmin enters and controls top-level database tables:

- Assigning User Group data access privileges: The client organization's superadmin controls which user groups can access which "database model views" of its target databases.

Close window

Previous Close window

Details on Semantic Search problems


- Using Natural Language input alone violates a major principle of User Interface Design. This principle has been identified thusly: "Recognition is better than Recall". In other words, speaking or typing a Plain Language request with no visual clues about the content of a target data source leads to wrongly named objects (is it "Customer" or "Client"?), plus the normal typing and spelling errors.



Previous Close window

Previous Close window

Details on Visual Query Interface problems


Visual User Interface tools are intuitive -- they usually show a tree or list of available "information elements" or data objects for the user to select in generating an ad hoc report.  But there is a giant problem trying to use this type of ad hoc query tool to obtain facts from databases: in addition to selecting or filtering data elements for a report, the user is expected to know or learn a technical part of database query languages: how to "join" tables such that the correct query is generated.


Vendors of eport generation tools such as Crystal Reports and Access try to convince their users that "joining tables" is easy: using their Graphic User Interface, a user can "just drag a Parent Table's Primary Key column(s) to point to a Child Table's Foreign Key column(s)".


What they don't tell you is that often there are dozens or even hundreds of tables in a target database, and the non-database savvy user has no idea of how to create the correct set of Joins needed to produce 'the correct answer'.   Even worse: what if the user TRIED to produce the correct Joins, but ended up with an incorrect set that always generated INCORRECT ANSWERS for the rest of the user base?  (for our solution to this problem, return to the previous section and click on >>details on Rules-based Join Navigation>>).


Actual studies have shown that, far from being "end user" tools, the users of these tools overwhelmingly are database-savvy IT professionals, with a few brave "power users" making up the rest of their user base.


If there's a truism about "end users", it's that they don't want to have to learn any techie jargon such as "parent table", "primary key", etc.



Previous Close window

Previous Close window

Details on Request Specifications


Visual Request Specifications (RequestSpecs) provide a user with a way to construct a precise Specification of an information request.



Previous Close window