A question worth asking about ArabicNLP conference

[Ali Al-laith]

Hello Everyone,

The numbers tell an interesting story:

• 2021: 43 submissions → 29 accepted (67%)  
• 2022: 68 → 36 (52%)  
• 2023: 80 → 38 (47%)  
• 2024: 68 → 31 (42.6%)  
• 2025: 95 → 39 (41%)  
• 2026: 193 → ~42 (22%)

In six years, submissions jumped from 43 to 193, while the acceptance rate dropped from 67% to 22%.

Also worth noting: a small number of frequent contributors appear on many accepted papers each year. This is not a problem with them as they are well-known researchers in NLP and clearly doing important work. But when acceptance becomes this tight, we should ask whether the current format gives enough space for the growing community.

Maybe it is time to think about making ArabicNLP an independent conference, not just co-located with EMNLP. With a few dedicated days and more tracks, we could accept more papers without lowering standards. Also introduce a limit on number of submissions that one can be an author on.

The goal is not to make things easier. It is to let the community grow without making the main venue harder and harder to get into.

Thank you,

Ali Al-Laith

[Mandar Marathe]

Hi,

I am new to this community, so please forgive my ignorance.

To me, the jump from 2025 to 2026 is an outlier. It is abnormal. Before thinking about a plan, we should therefore thematically analyse those extra 100 papers which were submitted but didn't make the grade.

Are there any common factors? Trends? Shared characteristics? I suspect that insight will give us a better understanding of what to do about it.

Best wishes,

Mandar.

[Nizar Habash]

Dear Ali and colleagues,

Thanks for starting this thread.

Having co-organized the first Workshop on Arabic NLP in 2014, the series that eventually became ArabicNLP, I am very proud of the growth in submissions, the conference’s reputation, and our Arabic NLP community as a whole. This progress reflects years of commitment and hard work by many people. ArabicNLP is no longer simply a niche workshop or a fallback venue; it has earned visibility and respect across the field. Arabic NLP Workshop even appears in the top 20 venues in Computational Linguistics on Google Scholar.

The current ArabicNLP 2026 organizers (of whom I am not one) have taken an important step toward raising the conference’s reputation through a more selective program. This may risk discouraging some submissions, but I expect that it will instead attract more and stronger work by establishing ArabicNLP as a competitive, high-quality venue.  For the moment, aspirationally, I hope we can develop a sustained comparable reputation to the Conference on Computational Natural Language Learning (CoNLL) - this year they had 257 submissions and 18% acceptance rate.

BTW, regarding the question about size... EMNLP imposes no limit on the number of papers ArabicNLP may accept. That decision belongs entirely to the program chairs and reflects their strategy for the conference. The shared tasks account for a lot of additional papers (last year 139 shared task papers and only 40 conference papers) - I expect more this year.  This is a dimension that provides substantial additional space for participation and high-quality community work.

Now whether ArabicNLP should become independent is a separate question. Running a standalone conference requires significant funding, planning, volunteer effort, audience-building, and long-term reputation development. Co-location with EMNLP or ACL (we are, after all, part of the ACL family) also keeps Arabic NLP connected to the broader NLP community. Attendees can discover and join ArabicNLP sessions, while our participants benefit from engaging with the larger field.

Best,
Nizar 
(SIGARAB President Hat on)

--
You received this message because you are subscribed to the Google Groups "SIGARAB: Special Interest Group on Arabic Natural Language Processing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sigarab+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/sigarab/88ff9462-e15c-4f21-9d3a-507764decaf5n%40googlegroups.com.

--

Nizar Habash

Professor of Computer Science

New York University Abu Dhabi
https://www.nizarhabash.com/

Follow me on X!

[Nizar Habash]

Hi Mandar - 

A few broader observations may help put the jump in context:

• The increase is consistent with the overall growth in submissions across NLP.
• Since COVID, virtual participation has become more common, allowing more researchers to submit without first securing travel funding.
• A smaller venue still needs a minimum number of papers for a viable program. Consequently, fewer submissions can naturally result in a higher acceptance rate.
• Individual years are also affected by competing venues, e.g., our sister venue OSACT, which meets every other year.

WANLP 2021 was particularly unusual BTW because it took place only six months after WANLP 2020, amid the disruptions surrounding COVID. The timing was unfortunate, and managing it was quite stressful.

The earlier numbers were reasonably stable (mostly):

• WANLP 2020: 20/44 (45%)
• WANLP 2019: 22/45 (49%)
• WANLP 2017: 22/47 (47%)
• WANLP 2015: 15/23 (65%) << an unusual year in China, before virtual participation, with roughly half the usual submissions
• WANLP 2014: 20/40 (50%)

My own reading is optimistic: the community has grown, and Arabic NLP has become increasingly attractive. I also suspect that the growing number of institutes in the Arab world prioritizing Arabic research is contributing to this pattern. At the same time, many more Arabic NLP papers are being submitted to major top-tier conferences, so the growth extends well beyond ArabicNLP itself.

We are currently planning ArabicNLP 2027 and will make an announcement at ArabicNLP 2026. It should be interesting to see where the community goes next. :-)

Best

Nizar

--

You received this message because you are subscribed to the Google Groups "SIGARAB: Special Interest Group on Arabic Natural Language Processing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sigarab+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/sigarab/102a9423-a6c1-4920-bf05-989578bc1065n%40googlegroups.com.

[Hend Al-Khalifa]

Hi Mandar,

To answer your question (as one of the PC chairs for ArabicNLP 2026), the common factors turned out not to be thematic at all, but methodological. You can find a summary of the recurring concerns from the (meta)reviews, grouped into categories with what went wrong and how to avoid theme, here in both Arabic and English. 

Sharing it in case it is useful for authors preparing for the next cycle of ArabicNLP.

Regards,

Hend

[A M]

Dear Ali and all,

I'm coming to ArabicNLP from an adjacent community. I'll be honest here, because I truly care about the Arabic language in general and not only at the research level.

What worries me is how the review process can be so stochastic for a "small" community in terms of the number of submissions, at least based on my experience submitting and reviewing this year and last year.

In my opinion, there is serious work that needs to be done to better control the reviewing process. In other small communities, I still see that the reviewing process is well controlled.

For example, the meta-review this year (which is a good initiative) also followed the same random pattern. In one paper, the meta-reviewer did their job, and despite two unfavorable reviews, they checked the rebuttal and the quality of the reviews and offered the right judgment by increasing the score of the paper instead of just taking the average (I'm saying it was the right choice because I was the minority reviewer).
In another paper, it's very clear that the reviews and the rebuttal were fed to an LLM to summarize them and give the average score.

Another issue is that I feel that many reviewers approach submissions with a rather closed-minded mentality. For a new "task" or "dataset," they apply the same standards that they use for the types of work they are accustomed to. Complaining about the size of the data is not necessarily valid if the nature of the data is different. For example, 9 samples of TikTok videos are different in scope and content from 9 samples of full movies.

I know that I'm a relatively young scientist and maybe didn't see enough :-) However, given that the community is still within a controlled size, it's quite sad to see that it's suffering from the same problem of fluctuations in the quality of the reviews, which can be demotivating and discouraging.

Best regards,
Ali

[Preslav Nakov]

Note that WANLP was a workshop in 2021 and 2022 and turned into a conference only in 2023.

50% acceptance rate is typical for workshops, while for conferences a good acceptance rate is 18-25% or so.

--

You received this message because you are subscribed to the Google Groups "SIGARAB: Special Interest Group on Arabic Natural Language Processing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sigarab+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/sigarab/88ff9462-e15c-4f21-9d3a-507764decaf5n%40googlegroups.com.

[Omar Najar]

Dear colleagues,

Very interesting discussion. I think the real story here is not that ArabicNLP is becoming harder to get into, but that Arabic NLP itself is becoming much bigger.

A ~22% acceptance rate can be a healthy sign of a maturing and competitive conference. At the same time, as the community grows, our reviewing process, shared tasks, tracks, and opportunities for emerging researchers should grow with it.

Selectivity builds reputation, but a strong research community also needs pathways for new people and new ideas to grow.

So perhaps the bigger question is not simply “How many papers should we accept?” but “How do we scale the community without losing quality?”

Overall, I think this is a very positive challenge to have, and an exciting sign of where Arabic NLP is heading.

Best,
Omer

To view this discussion visit https://groups.google.com/d/msgid/sigarab/CAGdv6h-3BDokhccW-w0QoB%2B%3D%2BTZJc0ZC3cTe%2B3BujibSt65%2BkQ%40mail.gmail.com.