Mat Kelly
@machawk1.bsky.social
Assistant Professor - Information Science
Drexel University College of Computing & Informatics (CCI)
WS-DL & ODU CS alum
Most things web archiving-related
Drexel University College of Computing & Informatics (CCI)
WS-DL & ODU CS alum
Most things web archiving-related
Now presenting our #jcdl2025 paper exploring IPARO, a decentralized web-archival system on IPFS/IPNS that embeds version links into archived content. We evaluate linking policies and storage/lookup/replay tradeoffs. @ipfs.tech
Slides: matkelly.com/presentation...
Paper: matkelly.com/papers/2025_...
Slides: matkelly.com/presentation...
Paper: matkelly.com/papers/2025_...
December 18, 2025 at 8:33 PM
Now presenting our #jcdl2025 paper exploring IPARO, a decentralized web-archival system on IPFS/IPNS that embeds version links into archived content. We evaluate linking policies and storage/lookup/replay tradeoffs. @ipfs.tech
Slides: matkelly.com/presentation...
Paper: matkelly.com/papers/2025_...
Slides: matkelly.com/presentation...
Paper: matkelly.com/papers/2025_...
My PhD student Chris Rauch presented OntExtract @ #JCDL2025! Web-based NLP w/ PROV-O provenance—comparing manual vs LLM pipelines & track semantic change.
Demo: ontextract.ontorealm.net
Slides: matkelly.com/presentations/2025_jcdl_ontextract.pdf
Paper: matkelly.com/papers/2025_jcdl_ontextract.pdf
Demo: ontextract.ontorealm.net
Slides: matkelly.com/presentations/2025_jcdl_ontextract.pdf
Paper: matkelly.com/papers/2025_jcdl_ontextract.pdf
December 16, 2025 at 6:02 PM
My PhD student Chris Rauch presented OntExtract @ #JCDL2025! Web-based NLP w/ PROV-O provenance—comparing manual vs LLM pipelines & track semantic change.
Demo: ontextract.ontorealm.net
Slides: matkelly.com/presentations/2025_jcdl_ontextract.pdf
Paper: matkelly.com/papers/2025_jcdl_ontextract.pdf
Demo: ontextract.ontorealm.net
Slides: matkelly.com/presentations/2025_jcdl_ontextract.pdf
Paper: matkelly.com/papers/2025_jcdl_ontextract.pdf
Our article "Liberation of LMS-siloed Instructional Data" was just released by the Code4Lib Journal!
The work documents our efforts in extracting IMLS-funded bootcamp content from Blackboard for future reuse in library carpentries & for reference by students.
📝 journal.code4lib.org/articles/18462
The work documents our efforts in extracting IMLS-funded bootcamp content from Blackboard for future reuse in library carpentries & for reference by students.
📝 journal.code4lib.org/articles/18462
October 22, 2025 at 3:49 PM
Our article "Liberation of LMS-siloed Instructional Data" was just released by the Code4Lib Journal!
The work documents our efforts in extracting IMLS-funded bootcamp content from Blackboard for future reuse in library carpentries & for reference by students.
📝 journal.code4lib.org/articles/18462
The work documents our efforts in extracting IMLS-funded bootcamp content from Blackboard for future reuse in library carpentries & for reference by students.
📝 journal.code4lib.org/articles/18462
That's a wrap on ACM Hypertext 2025! #ACMHT2025 #HT2025 #HT25
Next year's Hypertext 2026 will be in London at School of Advanced Study at University of London and the British Library. #HT2026
September 8-11, 2026
The theme will be Hypertext as Method.
Next year's Hypertext 2026 will be in London at School of Advanced Study at University of London and the British Library. #HT2026
September 8-11, 2026
The theme will be Hypertext as Method.
September 18, 2025 at 9:32 PM
That's a wrap on ACM Hypertext 2025! #ACMHT2025 #HT2025 #HT25
Next year's Hypertext 2026 will be in London at School of Advanced Study at University of London and the British Library. #HT2026
September 8-11, 2026
The theme will be Hypertext as Method.
Next year's Hypertext 2026 will be in London at School of Advanced Study at University of London and the British Library. #HT2026
September 8-11, 2026
The theme will be Hypertext as Method.
At #ACMHT2025 day 4, Yavuz Selim Kartal (from @gesis.org) discusses supplementing social media posts with AI-generated summaries, which are perceived as most understandable form of enrichment as compared to metadata enrichment or pull quotes. @acmht.bsky.social #HT2025
📄 dl.acm.org/doi/10.1145/...
📄 dl.acm.org/doi/10.1145/...
September 18, 2025 at 2:41 PM
At #ACMHT2025 day 4, Yavuz Selim Kartal (from @gesis.org) discusses supplementing social media posts with AI-generated summaries, which are perceived as most understandable form of enrichment as compared to metadata enrichment or pull quotes. @acmht.bsky.social #HT2025
📄 dl.acm.org/doi/10.1145/...
📄 dl.acm.org/doi/10.1145/...
Almost all participants in their interviews mentioned fairness in musicRecSys came down to paying the artists. Secondarily was alg. bias &popularity bias as interview insights.
Not everyone agreed that imbalance in the sys meant it was unfair. Many also wanted transparency & control.
#ACMHT2025
Not everyone agreed that imbalance in the sys meant it was unfair. Many also wanted transparency & control.
#ACMHT2025
September 18, 2025 at 2:24 PM
Almost all participants in their interviews mentioned fairness in musicRecSys came down to paying the artists. Secondarily was alg. bias &popularity bias as interview insights.
Not everyone agreed that imbalance in the sys meant it was unfair. Many also wanted transparency & control.
#ACMHT2025
Not everyone agreed that imbalance in the sys meant it was unfair. Many also wanted transparency & control.
#ACMHT2025
At Day 4 of #ACMHT2025, Karlijn Dinnissen describes the role of fairness & diversity in user choices and perception of music playlists.
* What do end users think makes a fair MusicRec sys?
* How does that impact their choices?
* How should they be informed?
#HT2025
📄 dl.acm.org/doi/10.1145/...
* What do end users think makes a fair MusicRec sys?
* How does that impact their choices?
* How should they be informed?
#HT2025
📄 dl.acm.org/doi/10.1145/...
September 18, 2025 at 2:15 PM
At Day 4 of #ACMHT2025, Karlijn Dinnissen describes the role of fairness & diversity in user choices and perception of music playlists.
* What do end users think makes a fair MusicRec sys?
* How does that impact their choices?
* How should they be informed?
#HT2025
📄 dl.acm.org/doi/10.1145/...
* What do end users think makes a fair MusicRec sys?
* How does that impact their choices?
* How should they be informed?
#HT2025
📄 dl.acm.org/doi/10.1145/...
Ágnes Horvát highlights in her @acmht.bsky.social keynote that the frequency of certain words in PubMed abstracts has drastically increased due to authors massaging their texts with LLMs. #HT2025 #ACMHT2025
September 17, 2025 at 7:35 PM
Ágnes Horvát highlights in her @acmht.bsky.social keynote that the frequency of certain words in PubMed abstracts has drastically increased due to authors massaging their texts with LLMs. #HT2025 #ACMHT2025
At the #HT2025 Day 3 keynote, Ágnes Horvát asks, "How is science effectively promoted in online social spaces? Does self-publicity online lead to a tangible increase in citations?"
When we account for venue, popularity of authors, etc., there still seems to be a net positive.
When we account for venue, popularity of authors, etc., there still seems to be a net positive.
September 17, 2025 at 7:11 PM
At the #HT2025 Day 3 keynote, Ágnes Horvát asks, "How is science effectively promoted in online social spaces? Does self-publicity online lead to a tangible increase in citations?"
When we account for venue, popularity of authors, etc., there still seems to be a net positive.
When we account for venue, popularity of authors, etc., there still seems to be a net positive.
In his #nht25 prez @ #HT25, David Millard (@hoosfoos.bsky.social) ids the parallels between hypertext & those that will come as a result of gen AI. Who controls meaning? What are the boundaries of text? We typically think of AI as an assistant, the human being's still the wordsmith.
September 17, 2025 at 3:24 PM
In his #nht25 prez @ #HT25, David Millard (@hoosfoos.bsky.social) ids the parallels between hypertext & those that will come as a result of gen AI. Who controls meaning? What are the boundaries of text? We typically think of AI as an assistant, the human being's still the wordsmith.
@eastgate.bsky.social describes resurrecting Information Cities at #NHT25 as a narrative, a journey toward understanding, a central theme in open world games like Elder Scrolls or Cyberpunk 2077 #ACMHT2025 #HT2025 @acmht.bsky.social
September 17, 2025 at 2:37 PM
@eastgate.bsky.social describes resurrecting Information Cities at #NHT25 as a narrative, a journey toward understanding, a central theme in open world games like Elder Scrolls or Cyberpunk 2077 #ACMHT2025 #HT2025 @acmht.bsky.social
They utilized the timestamp of a tweet to estimate the time that a tweet has been archived. They then established left & right time boundaries to limit search space to make resolving captured tweets from the web archive more efficient.
@tarannum44.bsky.social #ACMHT2025
📄 doi.org/10.1145/3720...
@tarannum44.bsky.social #ACMHT2025
📄 doi.org/10.1145/3720...
September 16, 2025 at 9:58 PM
They utilized the timestamp of a tweet to estimate the time that a tweet has been archived. They then established left & right time boundaries to limit search space to make resolving captured tweets from the web archive more efficient.
@tarannum44.bsky.social #ACMHT2025
📄 doi.org/10.1145/3720...
@tarannum44.bsky.social #ACMHT2025
📄 doi.org/10.1145/3720...
1. Reconstruct a tweet URL based on the embedded parameters in a tweet 2. Search @archive.org. Web archives are most useful when the Tweet (or account) is no longer on the live web. - @tarannum44.bsky.social #ACMHT2025
September 16, 2025 at 9:52 PM
1. Reconstruct a tweet URL based on the embedded parameters in a tweet 2. Search @archive.org. Web archives are most useful when the Tweet (or account) is no longer on the live web. - @tarannum44.bsky.social #ACMHT2025
Another reason fake tweets are shared is for satire. 100k likes and no comments is suspicious. How can you tell it's real? Web archives serve as a good src of evidence that it's real. Their goal is to build an automated method for evidence that a tweet is authentic @tarannum44.bsky.social #ACMHT2025
September 16, 2025 at 9:49 PM
Another reason fake tweets are shared is for satire. 100k likes and no comments is suspicious. How can you tell it's real? Web archives serve as a good src of evidence that it's real. Their goal is to build an automated method for evidence that a tweet is authentic @tarannum44.bsky.social #ACMHT2025
@tarannum44.bsky.social from @webscidl.bsky.social is up next at #ACMHT2025 describing web archives usage for verifying attribution of Twitter screenshots with a compelling use case of one's love for pineapple pizza. 🍍🍕 @phonedudemln.bsky.social
September 16, 2025 at 9:45 PM
@tarannum44.bsky.social from @webscidl.bsky.social is up next at #ACMHT2025 describing web archives usage for verifying attribution of Twitter screenshots with a compelling use case of one's love for pineapple pizza. 🍍🍕 @phonedudemln.bsky.social
Dominik Soós from @webscidl.bsky.social asks "Can LLMs Beat Humans on Discerning Human-written and LLM-generated Science News?" Their contribs inc. benchmarks & findings, new dataset w/ 362 article triplets, & a novel method to improve perf on open-weight LLMs #ACMHT2025
dl.acm.org/doi/full/10....
dl.acm.org/doi/full/10....
September 16, 2025 at 8:31 PM
Dominik Soós from @webscidl.bsky.social asks "Can LLMs Beat Humans on Discerning Human-written and LLM-generated Science News?" Their contribs inc. benchmarks & findings, new dataset w/ 362 article triplets, & a novel method to improve perf on open-weight LLMs #ACMHT2025
dl.acm.org/doi/full/10....
dl.acm.org/doi/full/10....
In his #ACMHT2025 talk, Mariusz Pisarski asks the questions about Google's AI integration into their search UI stating that research shows that users largely don't click on results at all, but instead rely on the provided summaries.
Paper 🔗 dl.acm.org/doi/full/10....
Paper 🔗 dl.acm.org/doi/full/10....
September 16, 2025 at 8:12 PM
In his #ACMHT2025 talk, Mariusz Pisarski asks the questions about Google's AI integration into their search UI stating that research shows that users largely don't click on results at all, but instead rely on the provided summaries.
Paper 🔗 dl.acm.org/doi/full/10....
Paper 🔗 dl.acm.org/doi/full/10....
Hashing the first few letters and having a linear search is a perfectly reasonable data structure and works well for a collection of thousands, probably better than a full sort.
- @eastgate.bsky.social @ @human-ht.bsky.social #ACMHT2025
- @eastgate.bsky.social @ @human-ht.bsky.social #ACMHT2025
September 16, 2025 at 4:32 PM
Hashing the first few letters and having a linear search is a perfectly reasonable data structure and works well for a collection of thousands, probably better than a full sort.
- @eastgate.bsky.social @ @human-ht.bsky.social #ACMHT2025
- @eastgate.bsky.social @ @human-ht.bsky.social #ACMHT2025
@eastgate.bsky.social @ @human-ht.bsky.social:The problem w/ dealing w/ lots of stuff: there's always more to read & more questions to pursue. Even in antiquity, we have people that used thousands of srcs. Ancient collections assume readers want multiple tellings & read against them. #ACMHT2025
September 16, 2025 at 4:29 PM
@eastgate.bsky.social @ @human-ht.bsky.social:The problem w/ dealing w/ lots of stuff: there's always more to read & more questions to pursue. Even in antiquity, we have people that used thousands of srcs. Ancient collections assume readers want multiple tellings & read against them. #ACMHT2025
At @human-ht.bsky.social at #ACMHT2025: @eastgate.bsky.social describes historical compression: We want to compress messages, because every bit is expensive. When your bits are inscribed in stone, they are really expensive.
One tablet had very bad handwriting then tooth marks on the back side.
One tablet had very bad handwriting then tooth marks on the back side.
September 16, 2025 at 4:23 PM
At @human-ht.bsky.social at #ACMHT2025: @eastgate.bsky.social describes historical compression: We want to compress messages, because every bit is expensive. When your bits are inscribed in stone, they are really expensive.
One tablet had very bad handwriting then tooth marks on the back side.
One tablet had very bad handwriting then tooth marks on the back side.
Words per pope: Andrezej Adamski discusses the inter-pope sentiment of popes in time at the HUMAN workshop (@human-ht.bsky.social ) at #ACMHT2025. Takeaway: phrases track the media environment of each era.
September 16, 2025 at 3:23 PM
Words per pope: Andrezej Adamski discusses the inter-pope sentiment of popes in time at the HUMAN workshop (@human-ht.bsky.social ) at #ACMHT2025. Takeaway: phrases track the media environment of each era.
How would hypertext help for census date resolution? Cathy Marshall discusses in her @human-ht.bsky.social keynote at #ACMHT2025, "It's important to be able to back out records when Ruth isn't Paul's bio mother & that all her records should be deprecated. The cost for link-making is high and noisy."
September 16, 2025 at 2:50 PM
How would hypertext help for census date resolution? Cathy Marshall discusses in her @human-ht.bsky.social keynote at #ACMHT2025, "It's important to be able to back out records when Ruth isn't Paul's bio mother & that all her records should be deprecated. The cost for link-making is high and noisy."
At the HUMAN workshop ( @human-ht.bsky.social ) at Day 2 of ACM Hypertext, Cathy Marshall gives the keynote with a story about John Tukey looking data at large while she was close reading the data (i.e., observing microwaved Ivory soap) #ACMHT2025
September 16, 2025 at 2:23 PM
At the HUMAN workshop ( @human-ht.bsky.social ) at Day 2 of ACM Hypertext, Cathy Marshall gives the keynote with a story about John Tukey looking data at large while she was close reading the data (i.e., observing microwaved Ivory soap) #ACMHT2025
Enjoyed Alexander Petros proposing to bake in some features of HTMX into the HTML spec, namely to extend it to support more hypermedia controls. #ACMHT2025
Check out, "The Missing Mechanic: Behavioral Affordances as the Limiting Factor in Generalizing HTML Controls":
doi.org/10.1145/3720...
Check out, "The Missing Mechanic: Behavioral Affordances as the Limiting Factor in Generalizing HTML Controls":
doi.org/10.1145/3720...
September 15, 2025 at 9:48 PM
Enjoyed Alexander Petros proposing to bake in some features of HTMX into the HTML spec, namely to extend it to support more hypermedia controls. #ACMHT2025
Check out, "The Missing Mechanic: Behavioral Affordances as the Limiting Factor in Generalizing HTML Controls":
doi.org/10.1145/3720...
Check out, "The Missing Mechanic: Behavioral Affordances as the Limiting Factor in Generalizing HTML Controls":
doi.org/10.1145/3720...