Pith. sign in

REVIEW

Inherent Biases in Reference based Evaluation for Grammatical Error Correction and Text Simplification

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1804.11254 v3 pith:2RI55W4K submitted 2018-04-30 cs.CL

Inherent Biases in Reference based Evaluation for Grammatical Error Correction and Text Simplification

classification cs.CL
keywords correctionbiaserrorevaluationgrammaticalreferencessimplificationsystems
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The prevalent use of too few references for evaluating text-to-text generation is known to bias estimates of their quality ({\it low coverage bias} or LCB). This paper shows that overcoming LCB in Grammatical Error Correction (GEC) evaluation cannot be attained by re-scaling or by increasing the number of references in any feasible range, contrary to previous suggestions. This is due to the long-tailed distribution of valid corrections for a sentence. Concretely, we show that LCB incentivizes GEC systems to avoid correcting even when they can generate a valid correction. Consequently, existing systems obtain comparable or superior performance compared to humans, by making few but targeted changes to the input. Similar effects on Text Simplification further support our claims.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.