Venues, Volumes, and Events - acl-org/acl-anthology GitHub Wiki
- A volume is a grouping of papers published as a unit in a particular year, with a title and editors. The papers may be conference/workshop papers or journal articles. A conference may publish its proceedings across several volumes (e.g., long papers, short papers, industry track, demo track, tutorial abstracts, student research workshop). All metadata for a volume and its papers is stored as XML.
- A venue is a name of a journal, conference, or workshop abstracted away from the occurrence (a venue may recur over multiple iterations). The list of venues is declared in venues.json. In the XML, venues label volumes. Each volume belongs to at least one venue—possibly multiple, as when two conferences or workshops share a joint proceedings. Workshops should list the special "WS" venue in addition to the more specific venue. A workshop colocated with a main conference generally should NOT list the main conference venue. Findings is considered a distinct venue from ACL, EMNLP, etc.
- A SIG is an official special interest group of the ACL. The list of SIGs is declared in sigs.json. In the XML, SIGs label volumes: a volume may list one or more associated SIGs.
- A collection is a grouping of volumes stored together in an XML file with a common base ID (e.g.,
2026.aclorP96). The collection ID specifies the year and, usually, the venue ID (note that old-styleWencompassed workshops generally). - An event is a grouping of proceedings volumes for meetings that happened in a common time and place. (Journals do not have events.) Event IDs are denoted with a hyphen:
acl-2026. By default, an event is generated as a combination of each venue ID and year. A collection XML file may specify information about colocated volumes from another collection that should be included under the same event. This causes some volumes to be associated with multiple events, e.g. workshop volumes belonging to the event of the workshop and to the event of the overall conference. Similarly, Findings volumes belong to both the year's overall Findings event and the event for the particular associated conference.
NOTE: some of the details below need to be updated. For example, we have transitioned YAML data to JSON format.
Venues in the Anthology are conferences and journals where papers are published. Each venue has its own venue page. Examples include the following:
- Main conferences within ACL like ACL itself, NAACL, and EMNLP, and the journal Computational Linguistics
- Workshops like SSST
- Non-ACL events like LREC
Venues have first-order representation in the Anthology in data/yaml/venues.yaml. Here is an example entry for ACL:
acl:
acronym: ACL
is_acl: true
is_toplevel: true
name: Annual Meeting of the Association for Computational Linguistics
oldstyle_letter: PThe top-level key is the venue's slug. This is used to create its venue name and, since 2020, in the Anthology ID. Prior to 2020, a single letter was used for Anthology IDs for some venues, e.g., P as indicated here for ACL. (The history here is that "P" originally stood for "Proceedings", when there was only ACL to worry about). The default letter was W ("workshop"), a catch-all that placed all venues without a top-level letter into a single undifferentiated category. Only one venue can be associated with an "oldstyle letter"; if you add it to two or more venues, it will be used by the last venue it was added to.
acronym is the human-written acronym, which can contain hyphens and mixed capitalization, and name is the venue name. This name should be the name of the venue, and not of an event held by that venue.
There are a number of binary variables, which default to False. is_acl denotes whether this is an ACL venue (conference, journal, or publication). Unfortunately, this is not always clear or true across years, but generally an event is an ACL event if it is managed by ACL or its regional variants or if it is associated with one of the ACL Special Interest Groups.
is_toplevel determines whether the event is displayed on the front page of the Anthology.
The raw data for each venue is stored in XML files, one per year. For example, ACL in 2020 is 2020.acl.xml; in 2019 and before, it was Pyy.xml (where yy is the year).
Venues issue volumes. Prior to 2020, volumes were identified by a two-digit number for W and a single-digit number for all other letters. Volumes are listed underneath https://www.aclweb.org/anthology/volumes. Example volume names were P19-5 and W19-52.
Modern volume names are not restricted in this way, but can be named with arbitrary strings under [a-z0-9]+. Examples include 2020.acl-main and 2020.autosimtrans-1.
There is a one-to-one association between volumes and venues through their Anthology ID, but in case of the oldstyle "W"-volumes, this venue is always the generic "workshop" venue. To associate these volumes with more specific venues, we need the joint.yaml file described in the next section.
An event is a meeting of a venue in a particular year, e.g., EMNLP 2020. Events are not represented as first order objects, but are inferred from the file representation for venues, so there is one event per venue per year. There can be, however, an <event> block inside each collection file (e.g., 2025.acl.xml) that lists both metadata for an event (location, dates, website URL) and a set of volumes that were presented at the event.
Events can be joint events. This term unfortunately has two meanings:
- Two venues might issue joint proceedings or volume; for example, in 2015, ACL and IJCNLP held a joint conference in Beijing.
- Two venues might be co-located; for example, when a workshop is attached to a larger conference.
Volumes are associated with venues using the <venue> tag in the volume's <meta> block. This is used to determine (a) what venue pages a volume will be displayed on and (b) what event pages a volume will be displayed on. A joint venue is handled by adding multiple <venue> tags such that the volume appears under both venues. The Anthology ID itself must choose one of the venues for its name. For example, in 2024, MWE and UDW held a joint workshop. The Anthology ID used mwe, but multiple venue tags cause the volume to be displayed under both venues.
An interesting use of the "venue" tag is the ws venue. Any workshop volume is assigned this venue in addition to its own venue. This causes the volume to appear under both venues. For example, the 2025 volume of the AfricaNLP is tagged with both <venue>africanlp</venue> and <venue>ws</venue>. The ws venue allows easy browsing of all workshop volumes introduced in a year.
Second, events can be colocated. This is done using a <colocated> block under the <collection> tag in the XML file. The parent venue will list all the workshops that appeared with it. So, for EMNLP 2020, we use a block like this in 2020.emnlp.xml:
<collection id="2020.emnlp">
<!-- elided -->
<event id="emnlp-2020">
<colocated>
<volume-id>2020.findings-emnlp</volume-id>
<volume-id>2020.alw-1</volume-id>
<volume-id>2020.blackboxnlp-1</volume-id>
<volume-id>2020.clinicalnlp-1</volume-id>
<volume-id>2020.cmcl-1</volume-id>
<volume-id>2020.codi-1</volume-id>
<volume-id>2020.deelio-1</volume-id>
<volume-id>2020.eval4nlp-1</volume-id>
<volume-id>2020.insights-1</volume-id>
<volume-id>2020.intexsempar-1</volume-id>
<volume-id>2020.louhi-1</volume-id>
<volume-id>2020.nlpbt-1</volume-id>
<volume-id>2020.nlpcovid19-2</volume-id>
<volume-id>2020.nlpcss-1</volume-id>
<volume-id>2020.nlposs-1</volume-id>
<volume-id>2020.privatenlp-1</volume-id>
<volume-id>2020.scai-1</volume-id>
<volume-id>2020.sdp-1</volume-id>
<volume-id>2020.sigtyp-1</volume-id>
<volume-id>2020.splu-1</volume-id>
<volume-id>2020.spnlp-1</volume-id>
<volume-id>2020.sustainlp-1</volume-id>
<volume-id>2020.wmt-1</volume-id>
<volume-id>2020.wnut-1</volume-id>This causes all of these volumes to appear on the event page for EMNLP 2020, but not on the EMNLP 2020 venue pages. The venue page just lists volumes emanated under that venue ID, whereas the event page lists volumes that were associated by virtue of being presented at a shared event.