> > (Confirming, you did say that the table of brushes would contain

500,000

> >items, right?)
> Yah. Right now the database will have 500 000 brushes and will

increase in

> the future because every time they identify a competitor brush or
> manufacturer brush, they will put it on the database. Therefore, in

the

> future the database will have 1 million or more. I talk with the guys

here

> and we are assuming 1.5 million the worst case.

Then the uniqueness of the require properties will significant impact on
how a complete solution might be constructed.

You haven't answered my previous question regarding the commonality of
the required properties -- i.e. what is the smallest and largest number
of brushes which will share the same required properties.

> >Also, while you may have 200 clients connected to the database, I

can't

> >believe that they all will be searching for matches simultaneously.
>
> Well, in fact this is true because this is a sales software. This

mean

> that every day sales team will use it to identify the brush and sell

it.

> But this can be resolved by multiple databases. So each sales

department

> have its own database and therefore just 10 clients will be accessing

the

> database at the same time. Do you know if firebird has tools for

maitain

> multiple databases synchronized?

Not built-in, but there are some third-party tools which can do this,
you can also build your own for a single table.

> >Is there any way to perform the similar-to search by looking for an
> >exact match based on all the properties, and then dropping the match
> >criteria based on an order of properties which have the least

relevance

> >for similarity.
>
> Is not so simple. The objective of this search is identify my company
> brushes when you input competitor or manufacturer information of the
> brush. Sometimes is easy because you have all information, but most

of

> the times you don't have anything. Just a competitor part number. So

my

> job if provide all brushes that can be similar with the competitor

brush.

But the part number will yield you the specs for the "target" brush,
from their you will need to find the possible matching brush from your
firm's inventory. This seems to be an even easier task, since there
maybe 1.5 Million brushes, your firm will only have a small portion that
it sells. Thus is further reduces the number of 'candidates' which need
to be evaluated.

> Yah. I want to list all brushes with similarity higher than 85 %. The

user

> also has the option to select the minimum level between 85% and 100%.

Understood, but I am sure that there are some properties which will have
a higher importance/value to the similarity value, so by evaluating
these properties first, you can quickly reduce the size of the set for
complete evaluation.


Sean