September 28, 2026 at 1:54 AM
Fiza Husain, Ankit Pandey, Yash Singh
Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR
https://arxiv.org/abs/2609.20828
Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR
https://arxiv.org/abs/2609.20828
September 22, 2026 at 10:48 PM
Fiza Husain, Ankit Pandey, Yash Singh
Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR
https://arxiv.org/abs/2609.20828
Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR
https://arxiv.org/abs/2609.20828
Fiza Husain, Ankit Pandey, Yash Singh: Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR https://arxiv.org/abs/2609.20828 https://arxiv.org/pdf/2609.20828 https://arxiv.org/html/2609.20828
September 21, 2026 at 6:39 AM
Fiza Husain, Ankit Pandey, Yash Singh: Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR https://arxiv.org/abs/2609.20828 https://arxiv.org/pdf/2609.20828 https://arxiv.org/html/2609.20828
Incidentally, one way Comic Sans shines is in its legibility: the way it's designed, it has a quality called "disfluency". That is, the letterforms are a bit irregular & choppy, which makes it a bit harder to read, which - ironically - leads to better retention of information. Weird, but it works.
September 10, 2026 at 4:58 PM
Incidentally, one way Comic Sans shines is in its legibility: the way it's designed, it has a quality called "disfluency". That is, the letterforms are a bit irregular & choppy, which makes it a bit harder to read, which - ironically - leads to better retention of information. Weird, but it works.
Does disfluency affect social judgments and decisions? Evidence from spontaneous speech https://pubmed.ncbi.nlm.nih.gov/42691919/
September 4, 2026 at 4:50 PM
Does disfluency affect social judgments and decisions? Evidence from spontaneous speech https://pubmed.ncbi.nlm.nih.gov/42691919/
I never listen to you with this sort of acute attention, but it’s interesting to hear someone repeat words (“on on on on on”), which is an less frequent pattern of disfluency
Everyone is digging their dialect maps but they might love to know what sort of disfluent speaker they are
Everyone is digging their dialect maps but they might love to know what sort of disfluent speaker they are
September 4, 2026 at 3:33 PM
I never listen to you with this sort of acute attention, but it’s interesting to hear someone repeat words (“on on on on on”), which is an less frequent pattern of disfluency
Everyone is digging their dialect maps but they might love to know what sort of disfluent speaker they are
Everyone is digging their dialect maps but they might love to know what sort of disfluent speaker they are
Does disfluency affect social judgments and decisions? Evidence from spontaneous speech https://pubmed.ncbi.nlm.nih.gov/42691919/
September 4, 2026 at 5:42 AM
Does disfluency affect social judgments and decisions? Evidence from spontaneous speech https://pubmed.ncbi.nlm.nih.gov/42691919/
Half the disfluency pain usually comes from condition_on_previous_text feeding hallucinated repeats back into context. Turning that off plus raising compression_ratio_threshold catches most false-start loops without a full finetune. Worth testing before you commit to retraining.
September 2, 2026 at 7:07 PM
Half the disfluency pain usually comes from condition_on_previous_text feeding hallucinated repeats back into context. Turning that off plus raising compression_ratio_threshold catches most false-start loops without a full finetune. Worth testing before you commit to retraining.
Like, the person will go uh and ah and "you know, the..." and pause, fighting for their life to recall the noun, and *then* when it pops back up in their mind: how do heavily inflected languages handle this disfluency?
August 31, 2026 at 6:12 PM
Like, the person will go uh and ah and "you know, the..." and pause, fighting for their life to recall the noun, and *then* when it pops back up in their mind: how do heavily inflected languages handle this disfluency?
Masking isn't the answer though. I can only hide my disfluency for so long. It's better to be open about it and find people who are accepting, who will love me for who I am. And who will have the patience to listen to me. +
August 31, 2026 at 5:38 AM
Masking isn't the answer though. I can only hide my disfluency for so long. It's better to be open about it and find people who are accepting, who will love me for who I am. And who will have the patience to listen to me. +
Gemini 3.5 Transcribe: Google's new speech-to-text model handles background noise & disfluency, outputs polished text with speaker timestamps. 5.04% WER, 70% latency gain vs Chirp 3. Powers Gemini app, Android, Chrome. https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-
August 28, 2026 at 9:54 AM
Gemini 3.5 Transcribe: Google's new speech-to-text model handles background noise & disfluency, outputs polished text with speaker timestamps. 5.04% WER, 70% latency gain vs Chirp 3. Powers Gemini app, Android, Chrome. https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-
Google DeepMind's new Gemini 3.5 Transcribe cuts transcription latency by 70% and achieves a 4.0% WER in streaming mode.
The model is already live on Gboard and Gemini macOS, supporting 85+ languages with native disfluency cleanup. A major step for real-time voice agents.
The model is already live on Gboard and Gemini macOS, supporting 85+ languages with native disfluency cleanup. A major step for real-time voice agents.
Gemini 3.5 Transcribe: Latency drops by 70% for intelligent dictation
Google DeepMind's new Gemini 3.5 Transcribe cuts transcription latency by 70% and achieves a 4.0% WER in streaming mode.
The model is already live on Gboard and Gemini macOS, supporting 85+ languages
www.alextech.ai
August 27, 2026 at 6:19 AM
Google DeepMind's new Gemini 3.5 Transcribe cuts transcription latency by 70% and achieves a 4.0% WER in streaming mode.
The model is already live on Gboard and Gemini macOS, supporting 85+ languages with native disfluency cleanup. A major step for real-time voice agents.
The model is already live on Gboard and Gemini macOS, supporting 85+ languages with native disfluency cleanup. A major step for real-time voice agents.
you're completely right, but I'm not sure how much of it is patience per se, I think the cognitive disfluency of being someone who is paying attention to [gestures vaguely at everything] & then having to talk to people who profoundly aren't is just crazy making.
August 25, 2026 at 11:40 AM
you're completely right, but I'm not sure how much of it is patience per se, I think the cognitive disfluency of being someone who is paying attention to [gestures vaguely at everything] & then having to talk to people who profoundly aren't is just crazy making.
Yes, is really sad that the cancer has spread to his bones.😢
I would say your comment that he was “such a poor speaker” should be explained because he lived with a stutter & a life-long speech disfluency since childhood.
I think he did amazingly well, considering.
I would say your comment that he was “such a poor speaker” should be explained because he lived with a stutter & a life-long speech disfluency since childhood.
I think he did amazingly well, considering.
August 9, 2026 at 12:02 AM
Yes, is really sad that the cancer has spread to his bones.😢
I would say your comment that he was “such a poor speaker” should be explained because he lived with a stutter & a life-long speech disfluency since childhood.
I think he did amazingly well, considering.
I would say your comment that he was “such a poor speaker” should be explained because he lived with a stutter & a life-long speech disfluency since childhood.
I think he did amazingly well, considering.
Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations
Zizhao Hu et al.
#arXiv #cs.AI
Zizhao Hu et al.
#arXiv #cs.AI
Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations
Human input reaches language models by typing or speaking, and each channel leaves a distinct signature: orthographic noise for keyboards; for voice, disfluency from conventional transcription and restructuring from AI-backed dictation tools. How do they impact an LLM's performance? In this paper w…
arxiv.org
August 5, 2026 at 6:39 AM
Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations
Zizhao Hu et al.
#arXiv #cs.AI
Zizhao Hu et al.
#arXiv #cs.AI
Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations
Human input reaches language models by typing or speaking, and each channel leaves a distinct signature: orthographic noise for keyboards; for voice, disfluency from conventional transcription and restructuring from AI-backed dictation tools. How do they impact an LLM's performance? In this paper w…
arxiv.org
August 5, 2026 at 5:12 AM
This was difficult. My birth year was a strange and dysfunctional time. Apparently, genetics was absolutely exploding. Anyway:
A-OK
Antidepressant
Disfluency
Diazepam
Grok
A-OK
Antidepressant
Disfluency
Diazepam
Grok
August 1, 2026 at 3:32 PM
This was difficult. My birth year was a strange and dysfunctional time. Apparently, genetics was absolutely exploding. Anyway:
A-OK
Antidepressant
Disfluency
Diazepam
Grok
A-OK
Antidepressant
Disfluency
Diazepam
Grok
Plus the whole “rolls off of the tongue” thing just isn’t true if you have certain kinds of speech disfluency (as I do)
July 27, 2026 at 1:05 AM
Plus the whole “rolls off of the tongue” thing just isn’t true if you have certain kinds of speech disfluency (as I do)
Pumpkin Spice Disfluency
July 23, 2026 at 6:16 PM
Pumpkin Spice Disfluency
🗣️ It was never my fault that I had disfluency and a thing for libraries. Masking required me to know all the words and write well.
July 22, 2026 at 2:36 AM
🗣️ It was never my fault that I had disfluency and a thing for libraries. Masking required me to know all the words and write well.
🎯
I won’t lie, I have some minor beef since it’s long been out of practical usage, but it’s definitely a word.
As a life long word nerd with disfluency, Wordle will chide me for using arcane, but accurate words….
Alas, computer games aren’t fair and it’s a good brain massage.
I won’t lie, I have some minor beef since it’s long been out of practical usage, but it’s definitely a word.
As a life long word nerd with disfluency, Wordle will chide me for using arcane, but accurate words….
Alas, computer games aren’t fair and it’s a good brain massage.
July 16, 2026 at 5:21 PM
🎯
I won’t lie, I have some minor beef since it’s long been out of practical usage, but it’s definitely a word.
As a life long word nerd with disfluency, Wordle will chide me for using arcane, but accurate words….
Alas, computer games aren’t fair and it’s a good brain massage.
I won’t lie, I have some minor beef since it’s long been out of practical usage, but it’s definitely a word.
As a life long word nerd with disfluency, Wordle will chide me for using arcane, but accurate words….
Alas, computer games aren’t fair and it’s a good brain massage.
Calls: 13th Workshop on Disfluency in Spontaneous Speech
Calls: 13th Workshop on Disfluency in Spontaneous Speech
Call for Papers:
We are pleased to announce the 13th edition of the DiSS workshop (Disfluency in Spontaneous Speech), to be held in Bolzano-Bozen (Italy) on February 4th-5th.
Established in 1999, the DiSS workshop series provides an interdisciplinary forum for research on disfluency and spontaneous speech, bringing together scholars from a wide range of disciplines. Submissions are encouraged from all fields concerned with disfluency, paralinguistics, and related phenomena, including, amon
dlvr.it
July 15, 2026 at 11:11 AM
Calls: 13th Workshop on Disfluency in Spontaneous Speech
Confs: 13th Workshop on Disfluency in Spontaneous Speech
Confs: 13th Workshop on Disfluency in Spontaneous Speech
We are pleased to announce the 13th edition of the DiSS workshop (Disfluency in Spontaneous Speech), which will take place in Bolzano-Bozen (Italy), on February 4th-5th.
Established in 1999, the DiSS workshop series provides an interdisciplinary forum for researchers interested in disfluency and spontaneous speech. It brings together researchers from a wide range of disciplines and welcomes studies on disfluency in spontaneous speech.
Submissions are encouraged from all fields that deal w
dlvr.it
July 14, 2026 at 10:14 AM
Confs: 13th Workshop on Disfluency in Spontaneous Speech
Calls: Speech Patterns and (Dis)fluency Markers: Insights from Spoken Corpora 2026
Calls: Speech Patterns and (Dis)fluency Markers: Insights from Spoken Corpora 2026
Call for Papers:
For centuries, linguistics focused on written language as “the norm”, and all features of language were studied based on written and often constructed artificial examples. Spoken corpora have enabled a shift in the way we view language, e.g. from a product-based to a process-based view on grammar, from an inherently-fluent view on communication to stipulating a fluency–disfluency continuum, from a verbally-centred communication to a multimodal construction of meaning.
Data
dlvr.it
July 9, 2026 at 12:11 PM
Calls: Speech Patterns and (Dis)fluency Markers: Insights from Spoken Corpora 2026
“Going round in circles on a book – finding it hard, feeling stuck – doesn’t seem challenging in a motivating way. It just feels bad.” #disfluency psyche.co/ideas/strugg...
psyche.co
June 24, 2026 at 9:05 PM
“Going round in circles on a book – finding it hard, feeling stuck – doesn’t seem challenging in a motivating way. It just feels bad.” #disfluency psyche.co/ideas/strugg...