You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@opennlp.apache.org by Markus Jelsma <ma...@openindex.io> on 2024/03/15 13:32:09 UTC
Re: Text classification using embeddings / centroids
I would certainly want to see if this algorithm performs any better than
the ones that are shipped with OpenNLP.
Is there already a Jira ticket associated with this task?
Op do 29 feb 2024 om 20:28 schreef Michael Wechner <michaelwechner@gmail.com
>:
> Hi Vaijanath
>
> Thanks for your feedback!
>
> I just received your message, but my previous reply I did not receive.
>
> I am also subscribed to other ASF mailing lists and don't experience this
> issue there.
>
> Thanks
>
> Michael
>
> Am Do., 29. Feb. 2024 um 21:10 Uhr schrieb Vaijanath Rao <
> vaiju1981@gmail.com>:
>
> > Hi Michael,
> >
> > I am not sure of delay, but It's possible ( for the delay), if you have
> > subscribed to options such as all the emails or daily digest.
> >
> > --Thanks and regards
> > Vaijanath
> >
> > On Thu, Feb 29, 2024 at 6:10 AM Michael Wechner <
> michaelwechner@gmail.com>
> > wrote:
> >
> > > Thanks for your feedback! I will have a closer look at
> > > https://github.com/jelmerk/hnswlib
> > >
> > > Btw, do others also experience strange issues re this mailing list?
> > > For example I am subscribed with another email, but do not receive
> > messages
> > > or it took a very long time to receive my own post.
> > >
> > > Thanks
> > >
> > > Michael
> > >
> > > Am Do., 29. Feb. 2024 um 14:45 Uhr schrieb Vaijanath Rao <
> > > vaiju1981@gmail.com>:
> > >
> > > > That might be good. We use https://github.com/jelmerk/hnswlib for
> > local
> > > > testing and other vendors for real time usage. With Lucene,
> > > > speed, scalability and updates were an issue for us.
> > > >
> > > > --Thanks and Regards
> > > > Vaijanath
> > > >
> > > > On Thu, Feb 29, 2024 at 12:50 AM Michael Wechner <
> > > michaelwechner@gmail.com
> > > > >
> > > > wrote:
> > > >
> > > > > Hi
> > > > >
> > > > > I recently found the following article re using text embeddings for
> > > text
> > > > > classifications
> > > > >
> > > > >
> > > > >
> > > >
> > >
> >
> https://medium.com/@juanc.olamendy/unlocking-the-power-of-text-classification-with-embeddings-7bcbb5912790
> > > > >
> > > > > Has anyone tried this?
> > > > >
> > > > > I am considering to implement this using Lucene Vector search.
> Could
> > > this
> > > > > be a contribution to OpenNLP?
> > > > >
> > > > > Thanks
> > > > >
> > > > > Michael
> > > > >
> > > >
> > > >
> > > > --
> > > > I am feeling fine, healthier and Happier, what about you
> > > >
> > >
> >
> >
> > --
> > I am feeling fine, healthier and Happier, what about you
> >
>
Re: Text classification using embeddings / centroids
Posted by Michael Wechner <mi...@gmail.com>.
Thanks for your feedback Vaijanath!
Can you give a concrete text example why you see this difference?
Thanks
Michael
Am Fr., 15. März 2024 um 20:03 Uhr schrieb Vaijanath Rao <
vaiju1981@gmail.com>:
> Embedding based classifiers are good for some use cases ( especially if you
> are modeling for similarity ), but opennlp use case ( initial ones ) are
> more for understanding text as in tokenization, Part of Speech, Chunking,
> Dependency and so on. I would not say one can replace the other.
>
> We use both in our pipeline, especially when we want to understand.
>
> --Thanks and Regards
> Vaijanath
>
> On Fri, Mar 15, 2024 at 8:09 AM Michael Wechner <mi...@gmail.com>
> wrote:
>
> > I have done an implementation within Katie
> >
> >
> >
> https://github.com/wyona/katie-backend/blob/main/src/main/java/com/wyona/katie/services/ClassificationServiceEmbeddingsCentroidsImpl.java
> >
> > and I am currently testing it and will try to compare it with the Max
> > Entropy Implementation of OpenNLP.
> >
> > Will keep ypu posted
> >
> > Thanks
> >
> > Michael
> >
> > Am Fr., 15. März 2024 um 14:33 Uhr schrieb Markus Jelsma <
> > markus.jelsma@openindex.io>:
> >
> > > I would certainly want to see if this algorithm performs any better
> than
> > > the ones that are shipped with OpenNLP.
> > >
> > > Is there already a Jira ticket associated with this task?
> > >
> > > Op do 29 feb 2024 om 20:28 schreef Michael Wechner <
> > > michaelwechner@gmail.com
> > > >:
> > >
> > > > Hi Vaijanath
> > > >
> > > > Thanks for your feedback!
> > > >
> > > > I just received your message, but my previous reply I did not
> receive.
> > > >
> > > > I am also subscribed to other ASF mailing lists and don't experience
> > this
> > > > issue there.
> > > >
> > > > Thanks
> > > >
> > > > Michael
> > > >
> > > > Am Do., 29. Feb. 2024 um 21:10 Uhr schrieb Vaijanath Rao <
> > > > vaiju1981@gmail.com>:
> > > >
> > > > > Hi Michael,
> > > > >
> > > > > I am not sure of delay, but It's possible ( for the delay), if you
> > > have
> > > > > subscribed to options such as all the emails or daily digest.
> > > > >
> > > > > --Thanks and regards
> > > > > Vaijanath
> > > > >
> > > > > On Thu, Feb 29, 2024 at 6:10 AM Michael Wechner <
> > > > michaelwechner@gmail.com>
> > > > > wrote:
> > > > >
> > > > > > Thanks for your feedback! I will have a closer look at
> > > > > > https://github.com/jelmerk/hnswlib
> > > > > >
> > > > > > Btw, do others also experience strange issues re this mailing
> list?
> > > > > > For example I am subscribed with another email, but do not
> receive
> > > > > messages
> > > > > > or it took a very long time to receive my own post.
> > > > > >
> > > > > > Thanks
> > > > > >
> > > > > > Michael
> > > > > >
> > > > > > Am Do., 29. Feb. 2024 um 14:45 Uhr schrieb Vaijanath Rao <
> > > > > > vaiju1981@gmail.com>:
> > > > > >
> > > > > > > That might be good. We use https://github.com/jelmerk/hnswlib
> > for
> > > > > local
> > > > > > > testing and other vendors for real time usage. With Lucene,
> > > > > > > speed, scalability and updates were an issue for us.
> > > > > > >
> > > > > > > --Thanks and Regards
> > > > > > > Vaijanath
> > > > > > >
> > > > > > > On Thu, Feb 29, 2024 at 12:50 AM Michael Wechner <
> > > > > > michaelwechner@gmail.com
> > > > > > > >
> > > > > > > wrote:
> > > > > > >
> > > > > > > > Hi
> > > > > > > >
> > > > > > > > I recently found the following article re using text
> embeddings
> > > for
> > > > > > text
> > > > > > > > classifications
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> https://medium.com/@juanc.olamendy/unlocking-the-power-of-text-classification-with-embeddings-7bcbb5912790
> > > > > > > >
> > > > > > > > Has anyone tried this?
> > > > > > > >
> > > > > > > > I am considering to implement this using Lucene Vector
> search.
> > > > Could
> > > > > > this
> > > > > > > > be a contribution to OpenNLP?
> > > > > > > >
> > > > > > > > Thanks
> > > > > > > >
> > > > > > > > Michael
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > I am feeling fine, healthier and Happier, what about you
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > I am feeling fine, healthier and Happier, what about you
> > > > >
> > > >
> > >
> >
>
>
> --
> I am feeling fine, healthier and Happier, what about you
>
Re: Text classification using embeddings / centroids
Posted by Vaijanath Rao <va...@gmail.com>.
Embedding based classifiers are good for some use cases ( especially if you
are modeling for similarity ), but opennlp use case ( initial ones ) are
more for understanding text as in tokenization, Part of Speech, Chunking,
Dependency and so on. I would not say one can replace the other.
We use both in our pipeline, especially when we want to understand.
--Thanks and Regards
Vaijanath
On Fri, Mar 15, 2024 at 8:09 AM Michael Wechner <mi...@gmail.com>
wrote:
> I have done an implementation within Katie
>
>
> https://github.com/wyona/katie-backend/blob/main/src/main/java/com/wyona/katie/services/ClassificationServiceEmbeddingsCentroidsImpl.java
>
> and I am currently testing it and will try to compare it with the Max
> Entropy Implementation of OpenNLP.
>
> Will keep ypu posted
>
> Thanks
>
> Michael
>
> Am Fr., 15. März 2024 um 14:33 Uhr schrieb Markus Jelsma <
> markus.jelsma@openindex.io>:
>
> > I would certainly want to see if this algorithm performs any better than
> > the ones that are shipped with OpenNLP.
> >
> > Is there already a Jira ticket associated with this task?
> >
> > Op do 29 feb 2024 om 20:28 schreef Michael Wechner <
> > michaelwechner@gmail.com
> > >:
> >
> > > Hi Vaijanath
> > >
> > > Thanks for your feedback!
> > >
> > > I just received your message, but my previous reply I did not receive.
> > >
> > > I am also subscribed to other ASF mailing lists and don't experience
> this
> > > issue there.
> > >
> > > Thanks
> > >
> > > Michael
> > >
> > > Am Do., 29. Feb. 2024 um 21:10 Uhr schrieb Vaijanath Rao <
> > > vaiju1981@gmail.com>:
> > >
> > > > Hi Michael,
> > > >
> > > > I am not sure of delay, but It's possible ( for the delay), if you
> > have
> > > > subscribed to options such as all the emails or daily digest.
> > > >
> > > > --Thanks and regards
> > > > Vaijanath
> > > >
> > > > On Thu, Feb 29, 2024 at 6:10 AM Michael Wechner <
> > > michaelwechner@gmail.com>
> > > > wrote:
> > > >
> > > > > Thanks for your feedback! I will have a closer look at
> > > > > https://github.com/jelmerk/hnswlib
> > > > >
> > > > > Btw, do others also experience strange issues re this mailing list?
> > > > > For example I am subscribed with another email, but do not receive
> > > > messages
> > > > > or it took a very long time to receive my own post.
> > > > >
> > > > > Thanks
> > > > >
> > > > > Michael
> > > > >
> > > > > Am Do., 29. Feb. 2024 um 14:45 Uhr schrieb Vaijanath Rao <
> > > > > vaiju1981@gmail.com>:
> > > > >
> > > > > > That might be good. We use https://github.com/jelmerk/hnswlib
> for
> > > > local
> > > > > > testing and other vendors for real time usage. With Lucene,
> > > > > > speed, scalability and updates were an issue for us.
> > > > > >
> > > > > > --Thanks and Regards
> > > > > > Vaijanath
> > > > > >
> > > > > > On Thu, Feb 29, 2024 at 12:50 AM Michael Wechner <
> > > > > michaelwechner@gmail.com
> > > > > > >
> > > > > > wrote:
> > > > > >
> > > > > > > Hi
> > > > > > >
> > > > > > > I recently found the following article re using text embeddings
> > for
> > > > > text
> > > > > > > classifications
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> https://medium.com/@juanc.olamendy/unlocking-the-power-of-text-classification-with-embeddings-7bcbb5912790
> > > > > > >
> > > > > > > Has anyone tried this?
> > > > > > >
> > > > > > > I am considering to implement this using Lucene Vector search.
> > > Could
> > > > > this
> > > > > > > be a contribution to OpenNLP?
> > > > > > >
> > > > > > > Thanks
> > > > > > >
> > > > > > > Michael
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > I am feeling fine, healthier and Happier, what about you
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > I am feeling fine, healthier and Happier, what about you
> > > >
> > >
> >
>
--
I am feeling fine, healthier and Happier, what about you
Re: Text classification using embeddings / centroids
Posted by Michael Wechner <mi...@gmail.com>.
I have done an implementation within Katie
https://github.com/wyona/katie-backend/blob/main/src/main/java/com/wyona/katie/services/ClassificationServiceEmbeddingsCentroidsImpl.java
and I am currently testing it and will try to compare it with the Max
Entropy Implementation of OpenNLP.
Will keep ypu posted
Thanks
Michael
Am Fr., 15. März 2024 um 14:33 Uhr schrieb Markus Jelsma <
markus.jelsma@openindex.io>:
> I would certainly want to see if this algorithm performs any better than
> the ones that are shipped with OpenNLP.
>
> Is there already a Jira ticket associated with this task?
>
> Op do 29 feb 2024 om 20:28 schreef Michael Wechner <
> michaelwechner@gmail.com
> >:
>
> > Hi Vaijanath
> >
> > Thanks for your feedback!
> >
> > I just received your message, but my previous reply I did not receive.
> >
> > I am also subscribed to other ASF mailing lists and don't experience this
> > issue there.
> >
> > Thanks
> >
> > Michael
> >
> > Am Do., 29. Feb. 2024 um 21:10 Uhr schrieb Vaijanath Rao <
> > vaiju1981@gmail.com>:
> >
> > > Hi Michael,
> > >
> > > I am not sure of delay, but It's possible ( for the delay), if you
> have
> > > subscribed to options such as all the emails or daily digest.
> > >
> > > --Thanks and regards
> > > Vaijanath
> > >
> > > On Thu, Feb 29, 2024 at 6:10 AM Michael Wechner <
> > michaelwechner@gmail.com>
> > > wrote:
> > >
> > > > Thanks for your feedback! I will have a closer look at
> > > > https://github.com/jelmerk/hnswlib
> > > >
> > > > Btw, do others also experience strange issues re this mailing list?
> > > > For example I am subscribed with another email, but do not receive
> > > messages
> > > > or it took a very long time to receive my own post.
> > > >
> > > > Thanks
> > > >
> > > > Michael
> > > >
> > > > Am Do., 29. Feb. 2024 um 14:45 Uhr schrieb Vaijanath Rao <
> > > > vaiju1981@gmail.com>:
> > > >
> > > > > That might be good. We use https://github.com/jelmerk/hnswlib for
> > > local
> > > > > testing and other vendors for real time usage. With Lucene,
> > > > > speed, scalability and updates were an issue for us.
> > > > >
> > > > > --Thanks and Regards
> > > > > Vaijanath
> > > > >
> > > > > On Thu, Feb 29, 2024 at 12:50 AM Michael Wechner <
> > > > michaelwechner@gmail.com
> > > > > >
> > > > > wrote:
> > > > >
> > > > > > Hi
> > > > > >
> > > > > > I recently found the following article re using text embeddings
> for
> > > > text
> > > > > > classifications
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> https://medium.com/@juanc.olamendy/unlocking-the-power-of-text-classification-with-embeddings-7bcbb5912790
> > > > > >
> > > > > > Has anyone tried this?
> > > > > >
> > > > > > I am considering to implement this using Lucene Vector search.
> > Could
> > > > this
> > > > > > be a contribution to OpenNLP?
> > > > > >
> > > > > > Thanks
> > > > > >
> > > > > > Michael
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > I am feeling fine, healthier and Happier, what about you
> > > > >
> > > >
> > >
> > >
> > > --
> > > I am feeling fine, healthier and Happier, what about you
> > >
> >
>