2023Q3 Reports: SIGMORPHON
This is the 2022-2023 annual report for SIGMORPHON, ACL's special interest group for computational morphology, phonology, and phonetics.
Membership
The SIG membership is documented in the following table, observing 24% growth from 176 to 218 members in the past 5 years. In particular, the last 2 years have seen significant growth in the SIG.
| Year | Membership |
|---|---|
| 2023 | 218 |
| 2022 | 196 |
| 2021 | 175 |
| 2020 | 176 |
| 2019 | 176 |
| 2018 | 157 |
| 2017 | 167 |
| 2016 | 157 |
| 2015 | 150 |
| 2014 | 144 |
| 2013 | 142 |
| 2012 | 137 |
| 2011 | 120 |
| 2010 | 105 |
| 2009 | 96 |
| 2008 | 90 |
| 2007 | 73 |
| 2006 | 55 |
Workshop
The 20th SIG workshop was held at ACL 2023, organized by Garrett Nicolai, Eleanor Chodroff, Çağrı Çöltekin, and Fred Mailhot. Attendance was approximately 30 participants in person, with another 30 participating online. This is especially important now as our community is undergoing rapid growth.
The 2023 Workshop had 22 submissions (an increase of 22% over the previous year) from which the program committee selected 12 for publication (55% acceptance rate). There were two invited talks: Carmen Saldana, and Kyle Gorman. The workshop was held in a hybrid format, with both live attendees at NAACL, and virtual attendees via Zoom.
In 2023, SIGMORPHON held an open solicitation for shared tasks, selecting 3, with one eventually retracting due to time constraints. Each task was separately proposed by an organizing committee, and separately approved by the SIG Executive Committee, who worked with the organizing committee in some cases. This structure was inspired by other groups (SIGSEM and SIGNLL) that run multiple shared tasks, and we expect to keep it in future years.
Task 0
SIGMORPHON's eighth installment of its inflection generation shared task was divided into three parts.
In the first part, participants designed a model that learned to generate morphological inflections from a lemma and a set of morphosyntactic features of the target form.
In the second part, participants were tasked with generating reasonable inflectional responses to a Wug task in Korean.
In the third part, participants produce models that fail in ways similar to children - ie, the aim was to imitate child learning of morphology.
Task 1
For the first time, we presented a shared task on interlinear glossing, with two sub-tracks:
In the first, participants must automatically gloss lower-resource data using only the source sentence and potentially a matrix translation (ie, English or Spanish).
In the second, the participants also have access to a morphological segmentation of the source text.
Elections
2023 represented an election of new officers to the SIG, who are listed on the website: https://sigmorphon.github.io/. Officers are serving a 2-year term, with the next election cycle to begin at the end of 2024.