Uniform Anaphora Resolution and Evaluation

Anonymous

Abstract: Despite the increasing attention on tackling anaphora resolution in an end-to-end multitask learning fashion, the state of the research topic is still unsatisfactory in that most works focus only on a subset of relations (either bridging or coreference), lacking generalizability and granularity for more complicated anaphoric relations. Moreover, the evaluations are still a mix of diverse metrics for different subtasks. We leverage a multitask learning framework from the Relation Extraction field which can be extended to perform fine-grained anaphora resolution and introduce a heterogeneous graph representation to evaluate coreference and other anaphoric relations using one uniform metric. All the data and source code will be publicly available.

Paper Type: short

Research Area: Resources and Evaluation

Contribution Types: Model analysis & interpretability, NLP engineering experiment

Languages Studied: English

Preprint Status: We are considering releasing a non-anonymous preprint in the next two months (i.e., during the reviewing process).

A1: yes

A1 Elaboration For Yes Or No: Section 9

A2: n/a

A3: yes

A3 Elaboration For Yes Or No: Abstract, Section 1

B: yes

B1: yes

B1 Elaboration For Yes Or No: Section 2, Section 5

B2: yes

B2 Elaboration For Yes Or No: Section 2, Section 5

B3: yes

B3 Elaboration For Yes Or No: Section 2, Section 5

B4: n/a

B4 Elaboration For Yes Or No: No sensitive information.

B5: yes

B5 Elaboration For Yes Or No: Section 5

B6: yes

B6 Elaboration For Yes Or No: Section 4, Section 6

C: yes

C1: yes

C1 Elaboration For Yes Or No: Section 6

C2: yes

C2 Elaboration For Yes Or No: Section 6

C3: yes

C3 Elaboration For Yes Or No: Section 7

C4: yes

C4 Elaboration For Yes Or No: Section 6

D: no

D1: n/a

D2: n/a

D3: n/a

D4: n/a

D5: n/a

E: no

E1: n/a