RTE7 - Ablation Tests

From ACL Wiki
Revision as of 07:40, 27 March 2012 by Amarchetti (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

The following table lists the results of the ablation tests submitted by participants to RTE7 .
The exploratory effort about knowledge resources, started in RTE5 and extended to tools in RTE-6, was proposed also in RTE-7.

In the table below, the first column contains the specific resources which have been ablated.
The second column lists the Team Run in the form [name_of_the_Team][number_of_the_submitted_run].[submission_task] (e.g. BIU1.2way, Boeing3.3way).
The third column presents the normalized difference between the accuracy of the complete system run and the accuracy of the ablation run (i.e. the output of the complete system without the ablated resource), showing the impact of the resource on the performance of the system.
The fourth column contains a brief description of the specific usage of the resource. It is based on the information provided both in the "readme" files submitted together with the ablation tests and in the system reports published in the RTE7 proceedings.
If the ablated resource is highlighted in yellow, it is a tool, otherwise is a knowledge resource.

Participants are kindly invited to check if all the inserted information is correct and complete.

Ablated Component Ablation Run[1] Resource impact - F1 Resource Usage Description
WordNet BIU2_abl-1 -0.05 Without WordNet, which is used as a lexical rulebase resource
Direct BIU2_abl-2 0.94 Without Bap (AKA "Direct"), which is used as a lexical rulebase resource
Wikipedia BIU2_abl-3 1.56 Without Wikipedia, which is used as a lexical rulebase resource
Coreference resolver BIU2_abl-4 0.69 Without any coreference resolution engine, instead of sing ArkRef to obtain coref information from the text, when preprocessing it
WordNet DFKI1_abl-1 -0.14 Features based on WordNet similarity measures (JWNL).
Named Entity Recognition DFKI1_abl-2 2.08 Features based on WordNet similarity measures (JWNL).
Wikipedia FBK_irst3_abl-2 -2.64 Ablating wikipedia LSA similarity scores.
Named Entity Recognition FBK_irst3_abl-3 -0.89 Ablating named entities matching module.
Paraphrase Table FBK_irst3_abl-4 -1.43 Ablating paraphrase matching module. The paraphrases were extracted from parallel corpora.
Acronym List IKOMA3_abl-1 -0.16 No acronyms of organization names extracted from the corpus.
CatVar IKOMA3_abl-2 0.84 No CatVar.
WordNet IKOMA3_abl-3 0.85 No WordNet.
WordNet JU_CSE_TAC1_abl-1 9.81 WordNet Ablated
Named Entity Recognition JU_CSE_TAC1_abl-2 7.97 NER Ablated
WordNet SINAI1_abl-1 -0.12 Resource ablated: lexical similarity module based on Personalized Page Rank vectors over WordNet 3.0
Wikipedia SJTU_CIT1_abl-1 8.89 we removed wikipedia resouce
VerbOcean SJTU_CIT1_abl-2 5.93 we removed verbocern resource
WordNet u_tokyo1_abl-1 0.83 Ablated resource is WordNet
WordNet u_tokyo2_abl-1 0.64 Ablated resource is WordNet
WordNet u_tokyo3_abl-1 0.99 Ablated resource is WordNet
UAIC Knowledge Resource UAIC20112_abl-1 0 Ablation of the BK (acronym database and world knowledge component)
Named Entity Recognition UAIC20112_abl-3 -8.29 Ablation of the NE resources.


  1. For further information about participants, click here: RTE Challenges - Data about participants

   Return to RTE Knowledge Resources