Re: Toward a web standard for XAI?

My research group has proposed a machine learning classification approach exploiting automatic ontology-based annotation of input data to merge computational statistics with non-standard reasoning.

A paper has been accepted in the Semantic Web journal and is in press [1]. Pre-press version can be found in [2].

Basically, the typical classification problem of ML is treated as resource discovery by means of semantic matchmaking. Outputs of classification are endowed with machine-understandable OWL descriptions, while the adopted reasoning procedures for matchmaking allow logic-based result explanation.

In our early tests classification performance is not so bad w.r.t. the state of the art, but both models and outcomes are explainable: considering the reference questions in the DARPA XAI initiative [3], our approach fully addresses "Why did you do that?", "Why not something else?" and "How do I correct an error?/Why did you err?".

We are currently working to make the approach more amenable to large distributed sensors network/IoT scenarios and to improve both classification and computational performance.

Best regards,
Floriano

[1] https://content.iospress.com/articles/semantic-web/sw314
[2] http://www.semantic-web-journal.net/content/machine-learning-internet-things-semantic-enhanced-approach-1
[3] https://www.darpa.mil/program/explainable-artificial-intelligence

--
Floriano Scioscia, Ph.D.
Information Systems<http://sisinflab.poliba.it/> Research Group
Department of Electrical and Information Engineering<http://dei.poliba.it/>
Polytechnic University of Bari<http://www.poliba.it/>
Home page: http://sisinflab.poliba.it/scioscia/
Informativa Privacy - Ai sensi del Regolamento (UE) 2016/679 si precisa che le informazioni contenute in questo messaggio sono riservate e ad uso esclusivo del destinatario. Qualora il messaggio in parola Le fosse pervenuto per errore, La preghiamo di eliminarlo senza copiarlo e di non inoltrarlo a terzi, dandocene gentilmente comunicazione. Grazie. Privacy Information - This message, for the Regulation (UE) 2016/679, may contain confidential and/or privileged information. If you are not the addressee or authorized to receive this for the addressee, you must not use, copy, disclose or take any action based on this message or any information herein. If you have received this message in error, please advise the sender immediately by reply e-mail and delete this message. Thank you for your cooperation.