Batch request via GPT-4o or mini does the trck

On Friday, December 18, 2015 at 10:04:02 AM UTC+3 Jeremy Maitin-Shepard 
wrote:

> On Thu, Dec 17, 2015 at 7:49 PM, Martin Blais <[email protected]> wrote:
>
>> Ideally, a classifier would operate on imported Beancount transactions 
>> directly rather than at the import level, so that it's reusable across a 
>> variety of different importers. In other words, the automatic 
>> categorization / completion of a transaction is a feature orthogonal to 
>> that of its creation via import.
>>
>
> I completely agree that the classification should operate independently of 
> the source, but there are some complications:
>
> - Classification is only easy to define for the case where you have a 
> transaction with exactly two postings: one posting from a known "source 
> account", and one unknown posting for the opposite amount in an unknown 
> account, to be determined by the classifier.
>
> - You need some way to determine, for each imported posting, a description 
> string (or other source of features), that will be used along with the 
> source account, and possibly the amount and date, for classification.
>
> - You need to be able to determine for each existing posting in the 
> beancount journal that was imported (or manually entered but later matched 
> with an imported posting) what the original description string used to 
> import it was.  This is used to train the classifier based on a beancount 
> journal.  Just using the payee/narration field of the transaction isn't a 
> good idea, since the user may have edited it, and we only want to train 
> based on the description strings seen when importing.
>
> A good way to achieve this would be to have a single import tool that 
> supports arbitrary backends that provide the actual transactions to import.
>

-- 
You received this message because you are subscribed to the Google Groups 
"Beancount" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/beancount/32d70532-7212-48ad-8732-fae96547644dn%40googlegroups.com.

Reply via email to