MLEE (mneb format) — nested event extraction

The Multi-Level Event Extraction corpus (Pyysalo et al., Bioinformatics 2012) converted into the mneb json_structures event-extraction format. MLEE annotates biomedical events across multiple levels of biological organisation — from molecular through cellular and tissue to organism level — over PubMed abstracts on angiogenesis.

This dataset preserves event-as-argument nesting. A third of MLEE's argument links point at another event rather than an entity. Prior conversions (including TextEE's, and an earlier version of this one) drop those links; this one keeps them. See Nested arguments — any consumer of this dataset must handle the nested argument shape.

Char offsets are character-based and end-exclusive (input[start:end] == text). One record = one MLEE document.

Splits

The official MLEE partition is preserved.

Split	Records	With events	Events
train	131	131	3,297
validation	44	44	1,175
test	87	87	2,206
Total	262	262	6,678

TextEE instead discards the official boundary and makes five random re-splits, so no number reported on a TextEE split is directly comparable to this one.

Record format

{
  "input":  "<document text>",
  "output": {"json_structures": {"<event_type>": [<event>, ...]}},
  "schema": []
}

An event is {"trigger": {"text","start","end"}, "arguments": [<arg>, ...]}.

Nested arguments

An argument is one of two things. Either an entity span:

{"role": "Cause", "text": "anti-VEGF neutralizing antibody", "start": 1705, "end": 1736}

or a nested event:

{"role": "Theme", "event": {"type": "Positive_regulation",
                            "trigger": {"text": "stimulated", "start": 1611, "end": 1621},
                            "arguments": [ /* recursively the same union */ ]}}

The discriminator is the presence of the event key. A nested event carries its own type; a top-level event does not, because its type is the json_structures dict key.

Only root events — those that are not an argument of any other event — appear at the top level. 4,419 of the 6,678 events are roots. To enumerate every event (for event detection, where nesting is irrelevant because the trigger set is identical either way), walk the tree:

def walk(ev, etype):
    yield etype, ev
    for a in ev["arguments"]:
        if "event" in a:
            yield from walk(a["event"], a["event"]["type"])

A real example from PMID-10586954 — "The insulin-conditioned RPE cell media stimulated capillary endothelial cell proliferation, an effect that was completely blocked by anti-VEGF neutralizing antibody":

"Negative_regulation": [{
  "trigger": {"text": "blocked", "start": 1694, "end": 1701},
  "arguments": [
    {"role": "Theme",
     "event": {"type": "Positive_regulation",
               "trigger": {"text": "stimulated", "start": 1611, "end": 1621},
               "arguments": [
                 {"role": "Theme",
                  "event": {"type": "Cell_proliferation",
                            "trigger": {"text": "proliferation", "start": 1649, "end": 1662},
                            "arguments": [{"role": "Theme",
                                           "text": "capillary endothelial cell",
                                           "start": 1622, "end": 1648}]}}]}},
    {"role": "Cause", "text": "anti-VEGF neutralizing antibody",
     "start": 1705, "end": 1736}]}]

Flattened, this would say only that something was blocked by an antibody — losing that what was blocked is the stimulation of proliferation.

Statistics

• 29 event types, 14 role types (role strings kept verbatim, so Theme2, Participant2..4 and Instrument2 are not collapsed into their base role).
• 8,597 arguments = 5,765 entity-valued + 2,832 event-valued (32.9% of all argument links are event-to-event).
• 2,260 events (33.9%) take at least one event argument.
• Nesting depth histogram {1: 4416, 2: 1981, 3: 269, 4: 10} — max depth 4.
• Nesting is almost entirely driven by the three regulation types; the sole exception is Planned_process, which takes an event argument 9 times. Only Theme and Cause are ever event-valued.

Note on event counts. The raw standoff has 6,678 E lines, but two pairs are byte-identical duplicate annotations (PMID-16076702 E28/E29 and PMID-19540587 E11/E25 — same type, same trigger, same arguments), so there are 6,676 distinct events. Both copies are retained in the data; deduplicating by full subtree collapses them.

Full type/role inventory and nesting patterns: mlee_label_summary.md. Browsable rendering: mlee_vis.html (open directly; data embedded, no server needed).

How this was derived

Built from the MLEE-1.0.2-rev1 standoff release (standoff/full/*.{txt,ann}), with the official split membership taken from the filenames in standoff/{development/train, development/test, test/test}:

1. Each document's .ann (a1 + a2 merged: entity mentions, event triggers and E event lines) is parsed into a single text-bound annotation map.
2. Each E line becomes an event; Role:T… arguments become entity spans and Role:E… arguments become inline nested event objects.
3. Events referenced as an argument by another event are omitted from the top level, so the top level lists roots only.
4. Relation (R), equivalence (*) and attribute (A/M: Negation, Speculation) lines are not carried over. Event-type and role strings are kept verbatim.

Verification built into the converter: every emitted span is re-checked against the source text (input[start:end] == text, 0 failures), and the set of full-subtree signatures in the emitted trees is compared against the same set computed straight off the raw standoff — they are equal for every document, which proves the nesting survived the conversion losslessly.

Licence and terms of use

• Annotations are licensed under Creative Commons BY-NC-SA 3.0. This is a non-commercial, share-alike licence: derivative works must carry the same terms, and commercial use is not permitted. Please attribute by citing the paper below and linking to http://www.nactem.ac.uk/MLEE/.
• Abstracts are from PubMed, a database of the U.S. National Library of Medicine; see the NLM copyright information.

Citation

@article{Pyysalo12mlee,
  author  = {Sampo Pyysalo and Tomoko Ohta and Makoto Miwa and Han-Cheol Cho and
             Jun'ichi Tsujii and Sophia Ananiadou},
  title   = {Event extraction across multiple levels of biological organization},
  journal = {Bioinformatics},
  volume  = {28},
  number  = {18},
  pages   = {i575--i581},
  year    = {2012}
}