 |
Description  |
|
|
BACKGROUND OF THE INVENTION
The invention relates to a computer-assisted system for brokering of goods
or services.
In certain markets for goods or services, for instance hiring personnel,
purchasing decisions are difficult because diverse characteristics of the
goods or services must be evaluated and compared, but information on the
goods or services is incensistent and dispersed.
SUMMARY OF THE INVENTION
The invention provides a computer-based system to facilitate any
transaction where review of diverse information is a part of the buyer's
decision-making process. It allows information in a number of forms to be
submitted by the seller, compiled in a database and reviewed by the buyer
with the assistance of an interactive, expert system based, networked
computer system.
In general, the invention features a computer-implemented system for
brokering transactions between sellers and a buyer of goods or services,
the system including a database, a Seller's Interface, and a Buyer's
Interface. The database contains information, including multimedia
information, descriptive of respective ones of the goods or services. The
Seller's Interface enables the sellers to interactively enter information,
including multimedia information, into the database. The Buyer's Interface
provides a knowledge-based interactive protocol, enabling the buyer to
select and review the descriptive information from the database, and makes
perceptible the multimedia information in response to an interactive buyer
request.
Preferred embodiments of the invention may include the following features.
The Seller's Interface enforces entry by the seller of at least a
predefined minimum set of information about each of the goods. The
descriptive information includes profile vectors of optional information.
The information of each profile vector is associated with other
information in the profile vector but is independent of information of the
other profile vectors for the same good or service. The Buyer's Interface
records actions of the buyer in an action log for later use, and a report
generator extracts information from the action log to provide feedback
information to the buyers and/or sellers. At least one of the Seller's
Interface and the Buyer's Interface has two modes, a first mode having
relatively slower interactivity for use with a low-bandwidth
communications channel, and a second mode having relatively faster
interactivity for use with a high-bandwidth channel. The system may have
automatic notification elements for notifying the buyer of descriptive
information newly-entered into the database that matches selection
criteria previously specified by the buyer. The Buyer's Interface may also
have, two modes of operation, a first mode for specifying selection
criteria for selecting descriptive information from the database, and a
second mode allowing detailed study of the selected descriptive
information. The knowledge-based protocol includes an
approximate-comparison system, for presenting to the buyer, goods or
services that approximately match selection criteria entered into the
Buyer's Interface. In one approximate-comparison system, the buyer is
presented those goods or services that meet user-defined required
criteria, and closely meet user-defined desired criteria.
The multimedia information may consist of audio and/or video, possibly in
combination with text, still images, or other digital computer data. The
information is edited and compiled into a coherent database with links to
allow navigation through the varied portions of each product's
information. The compiled information is presented to the buyer with the
assistance of an expert system that selects relevant information in
accordance with specified criteria. The interactive nature of the process
allows rapid review of a large quantity of information, comparison of
various alternatives and changes to criteria as the decision process
progresses.
This system can be used in a variety of transaction applications which
include, but are not limited to:
Purchase or rental of real estate, automobiles, aircraft or yachts;
Purchase or rental of consumer goods or services;
Hiring of full or part time personnel, executives, and consultants;
Exploration of travel, colleges, business opportunities, investments,
alliances or technology transfers.
Among the advantages of the invention are the following. The buying process
is made more productive for both the buyer and the seller by the system.
It provides a database of consistent and relevant information. The
seller's information is collected in an automated process to assure the
completeness of the required information of a Product Profile. The seller
may provide information in a variety of formats, including still images
and multimedia (including audio and/or video). To further provide complete
information, the system also collects information from other sources, such
as personal references, product evaluations or reports. The buyer is
provided an automated process to aid in selection from among a large
variety of products. The information available to the buyer is presented
in a consistent form. His "first cut" is made in accordance with the
characteristics of his Buyer's Profile matched by the database manager
against a database of Product Profiles gathered from all sellers; it is
therefore both accurate and complete. Further information is accessible in
various media so that a more enriched review can begin without delay. The
system provides more assistance, in addition to providing common
information, when the buying decision is made by a team. A record of the
buying process is also made by the Action Log that contains a record of
the products presented, information reviewed, buyer's scoring decisions,
etc. The system allows a buyer to make a purchasing decision more quickly,
both in terms of calendar days and in terms of the amount of time actually
expended in the search process.
BRIEF DESCRIPTION OF THE DRAWING
FIG. 1 is a block diagram of the broker system in accordance with the
invention.
FIGS. 2a-2f show database tables used in accordance with the invention.
FIG. 3a is a block diagram of a workstation for the Seller's Interface for
the broker system of FIG. 1.
FIG. 3b is a flow-chart of the Seller's Interface for the broker system of
FIG. 1.
FIGS. 4a-4q are screen displays for the broker system of FIG. 1.
FIG. 5 is a flow-chart of the operation of the Buyer's Interface for the
broker system of FIG. 1.
FIGS. 6a-6e and 7a-7n are screen displays of the Buyer's Interface for the
broker system of FIG. 1.
FIGS. 8a-8b show representative database tables for use with the invention.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to FIG. 1, a multimedia database 200 has a Seller's Interface 300
and a Buyer's Interface 500. Database 200 is arranged to provide Product
Profiles, each a description of a product (goods or services) that
includes the information that is important to both buyer and seller.
Seller's Interface 300 automates the process of "interviewing" sellers
about their products, and for each product produces a corresponding
Product Profile in database 200. The Buyer's Interface 500 assists the
buyer in selecting likely Product Profiles from database 200, and in
evaluating and comparing products to make a purchasing decision. A
Selection Engine 580 acts as an interface between the Buyer's Interface
and a database manager 201 for database 200.
In one example application, personnel search, the candidate (seller) is
interviewed by the Seller's Interface 300 to supply information for his
Product Profile. The interview collects information such as the position
the seller seeks, his desired salary and geography, and his experience. In
addition to this information, free-form text such as a work sample, still
images such as a resume, or multimedia information may be incorporated
into his Product Profile in database 200.
The Buyer's Interface 500 presents interactive screen queries to a hiring
manager (buyer) so that he can construct a "first cut" description, called
a Buyer's Profile, of the products that should be studied in greater
detail. Buyer's Interface 500 uses this Buyer's Profile to retrieve
product profiles from the multimedia database 200, for instance using SQL
queries. The system presents the candidates that closely match the Buyer's
Profile. The manager can then review further information and use other
facilities of Buyer's Interface 500 to assist his hiring (purchasing)
decision.
The Multimedia Database
Referring to FIGS. 2a-2d, in the personnel search example, multimedia
database 200 is the repository for the Product Profile information
collected by the Seller's Interface. Database 200 may use a known database
manager 201 using known technologies such as relational architecture and
SQL access. Database 200 is organized around a product identifier as the
primary database key. Each product is represented in the database by a
Product Profile. Each Product Profile has two parts: a required part, and
additional optional part with information included as the seller sees fit.
FIG. 2a shows the Product Table 202 of the Product Profile database in
which each product occupies a row. Each table column represents a
characteristic of a product. Thus, in the personnel search application,
the Profile Template table for the "product" might contain information
such as:
the candidate's name 204, address 206, and telephone number 208
last position or position sought 210: individual contributor (I.C.),
manager (Mgr.), vice president (V.P.), etc.,
compensation 212--for example, minimum and maximum values
education 214--bachelors (Bach.), masters (Mast.), etc.,
Subsidiary tables of multimedia database 200 contain other information,
organized as required by the nature of the information itself, for
instance, multimedia information stored for each product.
Referring to FIG. 2b, products may have variable amounts of multimedia
information, and thus this multimedia information is not stored in the
Product Table of FIG. 2a but rather within tables associated with each
type of media. FIG. 2b shows a table 220 indexing available video files.
The table shows that each of products P00001, P00003, P00005, P00006, etc.
has an associated video data file 228 containing a compressed video image,
and product P00008 has four associated video clips. The file's name
extension (".VID" in the example) may identify the particular technology
used in its compression. This table may also include a text description
222 of the video. For instance, a video answer to the question "What is
your favorite activity?" would be linked to the text of the question, for
display on Seller's Interface to prompt the candidate to speak, and in the
Buyer's Interface to show the buyer what video clips are available. These
text strings may be unique to a specific video or shared among videos
associated with several products.
When the Buyer's Interface presents a particular screen, it queries the
database for all video references associated with a particular product and
a particular screen display 224. The database row yields the icon 226 for
the video clip and the screen location at which to display the icon and
the text describing the video clip, and the name of the data file 228 that
contains the compressed video.
FIG. 2c illustrates a table 240 for audio information The "associated text"
column 243 describes text describing the clip itself for instance, "What
do you want to be doing in five years?" The Screen ID 244 and screen
position 245 tell what Buyer's Interface screen and which icon on that
screen are associated with the specific clip. See, for instance, FIG. 7g.
Tables similar to those of FIGS. 2b and 2c could identify text and images
associated with a Product Profile.
The database designer will specify certain information as being required
for a Product Profile. For the personnel search example, this information
might include:
the candidate's name
at least one contact mode: either a phone number or an address, for
instance
at least one experience entry: either an industry/function vector
(discussed below in connection with FIGS. 2d-f) or a description of the
candidate's education.
A candidate with "Production experience in the Biotechnology industry" and
"Design experience in the Aerospace Industry" should not match a search
for "Design experience in Biotechnology." Thus, the industry, function,
and skill set for each resume item of a candidate is logically correlated
in the database, and the data for each resume item are kept logically
separate from the data for the candidate's other resume items.
FIGS. 2d-2f show how this information describing the candidates might be
stored as a table 260 of "Industry Experience," a table 270 of
"Functions," and a table 280 of "Skill Set." Each of these three tables is
related to a screen of the Seller's Interface. The entries in the three
tables are each correlated to the candidate by a "Product ID" column 262,
272, 282. The rows of the three tables are indexed by "Product ID" and
"Component ID" columns 263, 273, 283. Associated rows for a single resume
entry are correlated to each other by "Prior Component ID" columns 264,
274, 284. The actual candidate information is stored in an
"Industry--Broad" column 265, an "Industry--Narrow" column 266, a
"Function--Broad" column 275, a "Function--Narrow" column 276, and a
"Skill Set" column 285.
SQL queries constructed by the Buyer's Profile access these experience
tables. These queries may attempt to match an entire row, for example,
"Industry is Finance/Loans/Consumer" or may match only a partial row, for
example, "TeleMarketing" experience regardless of industry.
The logical relationships between rows of the three tables are noted by the
"Component ID" columns 263, 273, 283 and "Prior Component ID" columns 264,
274, 284. In the example tables of FIGS. 2d-2f, rows 296, 297, and 298 are
correlated to each other. Row 298 is indexed, in columns 262 and 263,
"Product P00001, Component 1." Row 297 is labelled "Product P00001,
Component 2," and in link column 274, links to Component 1. Because the
"Component ID" is assigned sequentially to each table row for a specific
Product, "Product Component" pairs are unique among the three tables. A
table join operation will discover that the "link" field of row 297 links
row 297 to row 298. Similarly, row 296 links to row 297. Row 298 has a
null link field 264, indicating that no further searches are required to
link profile records for this experience item for Product P00001.
The Seller's Interface enforces relationships between the rows of the
tables. For instance, each row of "Industry" table 260 must have at least
one associated "Function" row in table 270. Each "Skill Set" row in table
280 must have at least one associated "Function" row in table 270 (and
thus one associated "Industry" row).
In the example discussed above, a candidate with "Production experience in
the Biotechnology industry" and "Design experience in the Aerospace
Industry," is seen to be described in two vectors, stored respectively in
rows 291 and 292 and rows 294 and 295. Rows 291 and 292 are linked
together by link field 274 of row 292, and rows 294 and 295 are linked
together by link field 274 of row 295.
Linked rows from the three tables are considered as an integral unit when
matching against a Buyer's Profile. Thus, a Buyer's Profile that specified
"must have Production experience in the Aerospace industry" would not
match candidate P00002 because there is no link between the linked rows
291 and 292 and the linked rows 294 and 295. As will be discussed below in
connection with FIG. 6e, a Buyer's Profile that merely gave weights to the
industry or experience, without using a combination rule, would match.
In an alternate database organization, the experience information of the
three tables of FIGS. 2d-2f can be combined into a single table. Instead
of explicitly representing links between a candidate's data, the interdata
correlations are noted by storing correlated data in a common row. This
single-row alternate representation requires extra space for fields that
are always present though unused (for instance, a "Skill Set" entry is not
required for every "Industry" and "Function" pair in the representation of
FIGS. 2d -2f, but would be stored as a null entry with an
"Industry/Function" pair), or is redundantly represented when a single
copy suffices in FIGS. 2d-2f (where, for instance, multiple "Function"
rows may be associated with a single "Industry" row). But the alternate
representation frees space for the "Component ID" and "link" fields.
Further, searching the single-table representation for candidates with
designated combinations of "specific function within a specific industry"
experience is improved because of reduced disk access. If a search is
desired in which one column is a "don't care," that column can be left out
of the WHERE clause of the SQL SELECT statement. The choice between the
organization of FIGS. 2d-2f will depend on circumstance and experience.
The "Industry" and "Function" codes may be designed by the database
designer, or for instance, taken from the Department of Labor's index of
job classifications.
The Seller's Interface
Referring to FIG. 3a, the Product Profile information is collected from
(and about) the candidate/seller/product by the automated Seller's
Interface 300 to efficiently collect complete and consistent Product
Profiles. In the simplest case, this collection may be performed by
distributing Seller's Interface 300 as a diskette-based application to be
run on a seller's personal computer. The multimedia information is
collected at a workstation or automated kiosk with appropriate input
devices, for instance a microphone 302 for audio input, a video camera 304
or VCR 306 for video input, a FAX machine or other scanner 308 for input
of documents, and/or disk or tape input 310. Either the seller's computer
or the computer that manages database 200 can also be used to edit and
link (as noted at 320) the input obtained from devices 302-310. The
editing computer will also generally include keyboard 322 to control its
operation and enter text data, and a communications port 324 to attach to
database 200.
FIG. 3b shows the process that the Seller's Interface uses to obtain a
Product Profile. Seller's Interface 300 first collects the information of
the Product Table of FIG. 2a. In step 352, the seller enters the product
name. In step 354, the Seller's Interface presents a top level screen
(FIG. 4a) showing the categories of information of a Product Profile. In
step 356, the seller chooses a particular category (402 of FIG. 4a) and
the Seller's Interface presents the profile characteristics appropriate to
that category (FIG. 4b).
In steps 358 and 360, the seller chooses characteristics (FIG. 4c) and
continues through successive screens, guided by the Product Profile
structure incorporated into the Seller's Interface. In step 362, the
Seller's Interface validates that at least the required information has
been entered. This validation, at the time of information entry, reduces a
buyer's need to later complete or clarify information. At step 361, prior
to validation at step 362, (FIGS. 4d, 4g, 4i, 4k, 4m), the seller may
select an additional category of information to enter, and the Seller's
Interface will prompt the seller with characteristics to select among
(FIGS. 4e, 4f, 4h, 4l, 4n, 4o, 4p, 4q).
Still referring to FIG. 3b, after step 362 has verified that at least the
required information has been entered, the seller may either exit or enter
optional additional information. In step 368 the seller selects from among
the categories of additional information to add. These categories include
multimedia information. In step 370, the Seller's Interface presents a
series of questions such as "What was your most significant accomplishment
in your most recent job?". The candidate's response to each question is
captured as audio and/or video information in step 372. Some of the
interview questions may be submitted to the candidate before the
interview, others may not, allowing for both rehearsed and unrehearsed
responses. Some general questions may be asked of all candidates while
others may be asked depending on specific characteristics of the
candidate's Product Profile.
In step 374, scanner 308 of Seller's Interface 300 can accept scanned-in
documents, for instance work samples or a resume. The Seller's Interface
may optionally run the document through a character recognizer (step 376)
to produce free-from text. Selected keywords may also be identified at
this time for use in retrieval. To correct character recognition errors,
Seller's Interface would present the resulting text to the seller for
review and correction (step 378).
The seller can enter other information in response to Seller's Interface
requests, for instance contact names for personal references (step 380).
Referring to FIGS. 4a-4c, once a seller has chosen a specific category of
information, for example, "Industry" 402 of FIG. 4a, the Seller's
Interface presents levels of expanded characteristics for more-detailed
selection, as shown in FIG. 4b. Pop-up menus may be used to allow
additional choices, as shown in FIG. 4c. Similarly, FIGS. 4d-4f show
another three-level choice of job function. In this case, the seller
selects the "Function" 412 category in FIG. 4d, and then from among five
sub-groupings: "Administrative" "Marketing/Sales" "Manufacturing",
"General" and "Technical" in FIG. 4e. The choice of "Marketing/Sales" 414
causes Seller's Interface 300 to present "Advertising", "Marketing" and
"Sales" as third-level choices. In this fashion the information of the
Product Profile is gathered by the Seller's Interface into a
highly-structured form suitable for database query by a buyer. Throughout
the process the seller is presented with new choices only in response to
his choices of higher-level criteria and is not presented with choices
that are not significant to him.
FIGS. 4g-4q show further examples of hierarchical choices available to the
Seller.
The Seller's Interface correlates entries by the seller to form the vectors
of FIGS. 2d and 2e, for instance to form one vector indicating "Design
experience in the Aerospace Industry" and another vector for the same
product indicating "Production experience in Biotechnology."
Following the information capture by Seller's Interface 300, a new Product
Profile is added to the Product Profile database 200. Multimedia
information is stored in compressed form in the multimedia database. This
may be in a variety of formats, e.g., JPEG, Group IV facsimile, MPEG or
Indeo, to accommodate various presentation devices and network facilities.
The Buyer's Interface
Referring to FIG. 5, Buyer's Interface assists the buyer in selecting the
Product Profiles of interest from among all Product Profiles in database
200, and then in reviewing this "first cut" in more detail. The hiring
manager, in the illustrated embodiment, is further assisted in scoring and
otherwise comparing candidates. The automated process helps to reduce
subjective and accidental biases.
Upon establishing a search session, in step 502 the system presents the
buyer with a series of screens, similar to those of FIGS. 4a-4q. The
buyer, assisted by Buyer's Interface, specifies search criteria (the
Buyer's Profile), indicating characteristics for selection or exclusion of
products. To provide additional capability to the search, the buyer may
specify "must have" criteria, weighted "want to have" criteria, and
unstructured word associations (see discussion of FIGS. 7a-7d, below). The
buyer may also specify a target number of "hits," that is, a target number
of candidates whose profiles he would like to retrieve from the database
for detailed review.
In step 504, the Buyer's Profile is transmitted to the database server 200.
This transmission may either occur in a "batch mode" after the buyer has
completed a full Buyer's Profile, or interactively as each sub-criterion
is specified. The interactive mode enhances responsiveness because the
database manager can begin its search with partial information. It also
allows the buyer to tailor his search in response to the partial results
retrieved from the database 200. In step 504, database server 200
retrieves database rows for all products that meet the "must" criteria.
This step in the search selects from the total population of the database
a workable set of products for further review.
If at least one product meets all the "must" criteria (step 506), in step
508 the database system and Selection Engine cooperate to select a set of
Product Profiles that represent the close matches to the Buyer's Profile
"want" characteristics. In step 508, the Selection Engine 580 and database
manager 201 cooperate to select those products that match all "want"
characteristics. If the number of hits meets or exceeds the buyer's target
number, then the Buyer's Interface can proceed directly to step 524. (If
the number of hits is very far away from the target, for instance if no
products meet all "must" criteria or if a permissive Buyer's profile gives
far too many hits, the Selection Engine may heuristically adjust the
Buyer's Profile. This feature is discussed below, in connection with FIG.
6e).
But if too few products are perfect matches (meet all "musts" and "wants"),
then in step 522 the Selection Engine selects all the products that meet
the "must" criteria, and evaluates the selected products by summing the
"want" weights of the characteristics that do match. The selected products
are rank-ordered according to these sums of weights.
In step 524, the perfect matches from step 508 or the ordered "near misses"
from step 522 are presented in summary form to the buyer.
In steps 526 and 528, the buyer may optionally modify the target number,
the weighting factors for the "wants," or the "must" criteria to adjust
the number of products selected from database 200.
In step 540, the Buyer's Interface presents product information in several
levels of detail from the most-general to the most-specific to assist the
buyer in selecting products for further review. See the discussion of
FIGS. 7e and 7f, below, for further details. The buyer may also choose to
modify the Buyer's Profile and thus the selected set of products (steps
526, 528) if the selected product set is too large, too small, or
otherwise inappropriate.
In step 540, the buyer selects an individual product for review. Buyer's
Interface 500 presents initial information on the selected product and
indicates at step 542 whether additional information about the product is
available, as discussed below in connection with FIGS. 7f-7i. The
product's additional information may be viewed by scrolling windows, or
selecting additional pages or associated multimedia information, by "point
and click" on an icon. The navigation through the information provided is
interactive, and information may be reviewed in any order and at the
buyer's desired pace. It may be re-reviewed or, in some cases, saved for
later perusal and comparison.
In step 550, the buyer may choose to review more detailed information for
the selected product. From step 550, the buyer may select steps 552, 554,
or 556 to obtain more information, or may loop back to steps 526, 528, or
540 to select a different candidate or set of candidates.
In step 552, the buyer may review the Profile Template information
describing the selected product--a presentation of all the structured
information on the product. Because this information is stored in
structured form, it provides consistency and facilitates comparison
between products.
At step 554, the buyer may also request to review multimedia or other data
associated with the product. For example, in the personnel application,
these would include the candidate's or references' answers to particular
questions, a scanned image of a resume, a representative piece of work, or
a presentation. Text information may also be searched to explore areas not
included in the structured information of the database tables of FIGS.
2a-2d.
At step 556 of the illustrated embodiment, upon completion of the review of
an individual product, the buyer chooses to record his impressions and
decisions. The Buyer's Interface accepts "scoring" information, further
actions and other annotations to assist in ranking products and also, for
example, for communication within teams of buyers. Reasons for elimination
of a product from further consideration can also be recorded. This phase
is discussed in more detail in connection with FIGS. 7j-7m, below.
The buyer can navigate among steps 552, 554, 556, 526, 528, and 540
described above: re-examining information, changing the Buyer's Profile,
or comparing products.
If a selection of a product is made (step 570), a complete Action Log of
the session is stored within the database (step 572). This will allow
subsequent analysis, feedback to the seller and tracking of the buyer's
actions and buying criteria. This is important in the personnel
application, for example to show conformance to employment regulations. It
can also allow the seller/candidate to modify his presentation or identify
desired training or other improvement needs.
Note that the flow chart of FIG. 5 shows interaction with a single buyer. A
team of buyers may each perform the buying session described in FIG. 5,
simultaneously if desired, and the Buyer's Interface will assist in
correlating the impressions of each member of the team, as discussed in
more detail in connection with FIGS. 7j-7m, below.
A more-detailed consideration of the screen displays of the Buyer's
Interface and the buyer's interactions follows.
Referring to FIGS. 6a-6d, the Buyer's Interface presents an opening
greeting. In box 612, the buyer identifies himself and gives a name to the
session 614 (either a new name, or the name of a previous session to
resume). Sessions might be named for a position or requisition number.
Security information is also entered to validate the session.
Referring again to FIGS. 4a-4q, the buyer enters Buyer's Profile search
criteria with more-detailed choices following selection from among broader
choices, using screens similar to those with which the seller entered a
Product Profile.
Referring to FIG. 6e, the conditions specified by a buyer at step 502 or
528 of FIG. 5 are not limited to a simple AND conjunction of one industry,
one function, one skill, etc. He may specify "and" or "or" conjunctions
between his criteria, and may specify some as "must" and others as
"wants." The "wants" can be weighted relative to each other. For instance,
the buyer may want a person with marketing or sales experience in the
electrical subsystems or semiconductor devices industries (for instance,
because the position's requirements are not specific to one industry). The
buyer can specify these complex conjunction rules graphically. For
instance, in FIG. 4c, the buyer has clicked two industry rules:
"Subassemblies/Electrical Subsystems" and "Subassemblies/Semiconductor
Devices." In FIG. 4f, the buyer has clicked both "Marketing" and "Sales."
In FIG. 4h, the buyer has clicked on "Direct," "OEM," "Rep.," and
"Telemarketing."
Still referring to FIG. 6e, the buyer can also note that some criteria are
"musts" and some are "wants." After going through all screens of FIGS.
4a-4q, a summary screen is displayed showing all criteria of the Buyer's
Profile. All selected characteristics are initially "wants" with weight
10. Characteristics can be made "musts" by clicking on the "M" button 660.
When a characteristic is selected as a "must," the associated "must"
button is highlighted, as shown for "Function."
A product is required to match all "musts" in the Buyer's Profile to be
selected. For instance, if multiple characteristics are made "musts" in a
category, a Product Profile will only match the Buyer Profile if the
Product Profile has all characteristics (for instance, as described in the
experience vectors of the tables of FIGS. 2d-2f). If two nested
characteristics within a category (for instance, "Telemarketing Sales" is
nested within "Marketing/Sales" within category "Function") are both made
"musts," then the database search can ignore the broader characteristic
and select only on the narrower.
Note that "must" characteristics have no numeric weight.
The weight 662 of a "want" can be raised or lowered by clicking on arrows
664. If multiple characteristics are selected in a single category (for
instance "Electrical Subsystems" and "Semiconductor Devices" as discussed
above), weights can be individually assigned to each characteristic by
scrolling the selected characteristics of the category with scroll arrows
666 and adjusting the weight to be associated with each of the
characteristics.
A selected characteristic can be assigned a weight of zero, for instance as
shown in "Location" 668. This has the same effect as if the characteristic
is assigned the lowest possible weight. If some characteristics in a
category are given zero weights and others given non-zero weights, then
those with non-zero weights will be given preference over those with zero
weights.
Combination characteristics can be specified in the combination box 670.
Characteristics are selected for combinations by clicking on the
characteristic selection boxes 672 (using scroll arrows 666 to display the
characteristics of a category) and pasting selected characteristics into
combination box 670. In the example of FIG. 6e, a combination has been
defined calling for "Marketing/Sales" with "Subassemblies." The
combination has been given a high weight of twenty. Note that in the
"Function" box, "Marketing/Sales" has been made a "must." The industry
selection, subassemblies, has been given a low weight of five. By using
these features, a very flexible selection structure can be configured. The
terms of a combination are ANDed together if the buyer selects "EVERY" 674
or ORed together if the buyer selects "ANY" 676. Combinations, may be
selected and combined into further combinations, and given weights or be
made "musts." The product will be given the weight score of the
combination if all characteristics of the combination match, or a zero
score if any of the elements of the combination fail to match. The product
will remain unselected if it fails to match any characteristic in a "must"
combination.
It may be desirable or necessary that either the buyer or the Selection
Engine treat the selected characteristics of one or more categories as
"musts" so that the number of database hits in step 504 is smaller than
the entire database. If the buyer specifies no "must" characteristics,
then the Selection Engine can use one of several heuristics to adjust the
Buyer's Profile to reduce the number of hits to a workable number.
Complements of these same heuristics can be used to convert "musts" to
"wants" if there are too few hits.
For instance, if electrical subsystems is preferred but semiconductor
devices is acceptable, the buyer might assign "Electrical Subsystems" a
weight of eighty, and "Semiconductor Devices" a weight of seventy-six.
These weightings will allow the database manager 201 in step 504 to gets
hits for candidates with no experience in these industries, but it will
cause the ordering of step 522 to assign candidates matching these
characteristics a high ranking that "swamps out" all other lower ranking
(and hence lower weighted) "wants." If the buyer limits the number of
categories with large-weighted characteristics, this will have the effect
of picking a pool of candidates with experiences matching at least one
characteristic in each of the categories given large weights. Note that
"musts" have the effect of ANDing together the selected characteristics;
large weights in a category have the effect or ORing together the
large-weighted characteristics. The Selection Engine may incorporate
special logic to recognize when large weights have been assigned to the
selected characteristics of a category, and to internally treat the
weighted selections as a "must" for the category when generating the SQL
select statement. This has the advantage of reducing the number of
database hits in step 504.
A second heuristic forms a union of intersections of characteristics. For
instance, if the buyer has specified a "must" on a characteristic from
category A and "wants" on characteristics from four other categories B, C,
D, and E, the Selection Engine may, when generating an SQL SELECT, treat
this Buyer's profile as one that matches all but one of the wants: "A and
((B and C and D) or (B and C and E) or (B and D and E) or (C and D and
E))." If this results in too few hits, the Selection Engine may try again
with another SELECT that specifies two of the four want categories: "A and
((B and (C or D or E)) or (C and (D or E)) or (D and E))." If there are
still to few hits, the Selection Engine may try again with a still-broader
match criterion, or give up and report what it found to the user.
In a third heuristic, a Buyer's Profile could be broadened by substituting
broader characteristics for narrower ones. For instance, if the Buyer's
Profile specifies "Subassemblies/Semiconductor Devices," the Selection
Engine may substitute the broader characteristic "Subassemblies." Or, the
database system may have tables telling which characteristics are
near-synonyms. In cases where the database query generates too few hits,
the Selection Engine may request from these tables near-substitutes for
the Buyer's Profile characteristics, and thus automatically generate a
broader database query.
A fourth filtering method may be combined with or used instead of the three
heuristics. For instance, if one of the heuristics gets too many hits,
this filtering method may be used to pare down the number of products
presented to the buyer. If a heuristic method gets too few hits, the
Selection Engine and Database Manager may try again with a looser match
criterion, and use the filter. In the filter, each product selected by the
"musts" of the Buyer's Profile (if there are no "musts," all products will
be selected) is evaluated against the weighted characteristics of the
Buyer's Profile, with the product given a "score" equal to the sum of the
weights of the matching characteristics, or the number of "want"
characteristics matched. A buffer equal to the target size, or possibly
somewhat larger, is maintained in sorted order. Each product is scored; if
the score is better than the score of the product at the bottom of the
buffer, then that bottom product is dropped from the buffer, and the
newly-scored product is inserted into the buffer at the appropriate point.
Referring to FIGS. 7a and 7b, the buyer may select an "other" category 702
of selection criteria. This choice presents a list of optional entries
presented in a scrolled list format. The buyer can thus choose from less
frequent entries. Since these same selections have been presented to the
seller in creating the Product Profile, accurate matches are possible. For
instance, a buyer may choose a "Location" using a combination of a button
followed by selection of, for example, a metropolitan statistical area
from a scrolled list. FIG. 7b shows choosing specific characteristics such
as companies, educational institutions, or a keyword 704. This allows
searching for keywords in text data associated with a product, for
instance in a resume.
Referring to FIG. 7c, when the Buyer's Profile is complete, the buyer may
define specific "exclusion criteria" 706 for products known not to be of
interest, for instance individual candidates or employees of specific
companies.
Referring to FIG. 7d, after the database query of steps 504 and 508, in
step 524 the Buyer's Interface presents a summary 708 of the database
hits. The summary lists the total number 710 of Product Profiles reviewed,
the number 712 that met all "must" criteria, and of those, the number 714
that also | | |
