For the current moment (Apr.2023) we have got the working cloud service with subscription. It's
contain the functions for list operations (import, export, comparing, cleaning) usually
needed as regular MDM system operations, capable of comparing lists of items like product descriptions from different vendors. Those lists are commonly created manually (without the help of automated systems) and though contain variety of fields and
contained types of information. Those data usually contains a lot of mistakes, typos etc. Two lists with 1 thousand positions each give us 1 million pairs to compare. Our associative index in acouple of minutes helps to determine only really needed pairs to compare (it is tenth of percent
of all positions usually), so the similarity is computed in minutes instead of hours on the basic PC with 8 CPUs and 8 GB RAM
It would appear to me to have significant benefits, if it were augmented by a library of user customizable filters in reduction of "biases" in training datasets submitted to generative AI instances.
Jim Overbyiii
What parameters would you suggest for these filters?
At the moment we have made a cloud service that allows you to automatically reconcile lists and various directories with each other.
I have taken the liberty of forwarding a suggestion to another researcher who will contact you for a more in depth discussion of the development of the filters if he chooses to participate in our discussion. I would expect him to contact you soon. However, to answer your question, I would expect the filters to be "callable" as needed by something like an event dispatcher or object request broker, with the capability to adapt the sequence of filtering dynamically to the situation requirement and/or prior results. Eventually resulting in a "library" of filters available for reuse.
- Badges
- Science topic
- Similar topics
- Indexes