Skip to content

Runs - Knowledge Base Acceleration 2014

BIT_Purdue-baseline

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-baseline
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 9d9dbd97d89e2d1e46efb80f24395a64
  • Run description: Entity title strings are used as surface form names, then any document containing one of the surface form names is ranked vital with confidence proportional to length of surface form name, and the longest sentence containing the longest surface form name is treated as a slot fill for all slot types for the given entity type.

BIT_Purdue-BinaryRank

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-BinaryRank
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 4f79a080b0d5173eaea8420dc256d93d
  • Run description: Use all the instances together to train a ranking model, then rank all the candidate documents.

BIT_Purdue-GlobalClassU

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-GlobalClassU
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 14859aa209e513775d5878dcdc5d5a1c
  • Run description: Use all the instances together to train Random Forest classification model, then classify all the candidate documents.

BIT_Purdue-GlobalClassV

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-GlobalClassV
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 4e27b649482ea531ba207d84bb7d565e
  • Run description: Use all the instances together to train Random Forest classification model( vital VS. Non-vital), then classify all the candidate documents.

BIT_Purdue-GlobalClassV1

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-GlobalClassV1
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 9f6bf1a8e21c669a1c95e239cad0ebac
  • Run description: A global classification model.

BIT_Purdue-GlobalRank

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-GlobalRank
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 96b916b8680a082a505aeebd94eed1dd
  • Run description: Use all the instances together to train a ranking model, then rank all the candidate documents.

BIT_Purdue-labeled

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-labeled
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: fe243e3cf51b358a9ba0055c1dac800f
  • Run description: query expansion run, use the related entities found in the vital labeled documents, and search the target entity and its related entities together against the built indices, the more entities are matched, the higher ranking of the candidate documents.

BIT_Purdue-profile

Participants | Proceedings | Input | Appendix

  • Run ID: BIT_Purdue-profile
  • Participant: BIT_Purdue
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e6d966441e42f2909c5a3d30cd5d2fba
  • Run description: a simple query expansion method, expand the exact query with the named entities found in the profile pages, rank the candidate documents according to their matched named entities

BUPT_PRIS-pris_baseline

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-pris_baseline
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 8092ad613d55f2752706f83f1eab0478
  • Run description: a run that use the results ES returns

BUPT_PRIS-pris_NN

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-pris_NN
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/1/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: afec17c3d4faa4fa4876422764b352ff
  • Run description: a run that use the results ES returns,and classified with NN

BUPT_PRIS-pris_rf

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-pris_rf
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/1/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 7d3d2aa19d80d0f5eef073d8459b2ee4
  • Run description: a run that use the results ES returns,and classfied with RF

BUPT_PRIS-pris_svm

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-pris_svm
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/28/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 93be3b2bd30c155c0b21691fa39df842
  • Run description: a run that use the results ES returns,and classify with SVM

BUPT_PRIS-ssf1

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-ssf1
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/8/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: fc72832cbfc691a552dc29100a7fd36a
  • Run description: This automatic extraction system contains three steps. First, With query expansion and corefence resolution, we can find relative sentences(to make the search faster, we built index using Elasticsearch). Second, we found rules by using KBP training data and bootstrapping method, and calculate the weight of each rules using logistic regression. Finally, we found slot answers by matching the rules. Specially, we manually picked up some seeds for those slottypes that KBP didn't contain to use bootstrapping method. And there are also some rules found manually such as PER_GENDER,PER_CONTACT,PER_EMAIL,ORG_CONTACT.

BUPT_PRIS-ssf2

Participants | Proceedings | Input | Appendix

  • Run ID: BUPT_PRIS-ssf2
  • Participant: BUPT_PRIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/8/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: a73c0b261d3c87ce12843ba6c34ee574
  • Run description: This automatic extraction system contains three steps. First, With query expansion and corefence resolution, we can find relative sentences(to make the search faster, we built index using Elasticsearch). Second, we found rules by using KBP training data and bootstrapping method, and calculate the weight of each rules using logistic regression. Finally, we found slot answers by matching the rules. Specially, we manually picked up some seeds for those slottypes that KBP didn't contain to use bootstrapping method. And there are also some rules found manually such as PER_GENDER,PER_CONTACT,PER_EMAIL,ORG_CONTACT.

ecnu-baseline_run

Participants | Input | Appendix

  • Run ID: ecnu-baseline_run
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e939f3d4d141f940204da4bd3fd304b9
  • Run description: Entity name strings are used to filter documents, each entity name is used as an query to do search in the indexed kba corpus. According to query result rank value, the higher value means the higher relevance. A threshold is used to identify the item relevance -1,0,1,2.

ecnu-idr_lda_1

Participants | Input | Appendix

  • Run ID: ecnu-idr_lda_1
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 0e79f1ec009395afbc254e67886fc625
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information in the training data are used to as query, then indri laguage models dirichlet is used, then according to query result, identify item relevance.

ecnu-idr_lda_2

Participants | Input | Appendix

  • Run ID: ecnu-idr_lda_2
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 2e23c1178d604cdd9befcdfe7c117939
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information and related entity (we assume that the number of the entity occurs in entity training data larger than a threshold, then it is related entity) in the training data are used to as query, then indri laguage models dirichlet is used, then according to query result, identify item relevance.

ecnu-idr_lda_3

Participants | Input | Appendix

  • Run ID: ecnu-idr_lda_3
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 67f9e2449bbe0770adc434ff70174b31
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information in the training data, external url are used to as query, then indri laguage models dirichlet is used, then according to query result, identify item relevance.

ecnu-idr_lda_4

Participants | Input | Appendix

  • Run ID: ecnu-idr_lda_4
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: b713a295594232f9010068c66818fce6
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information and related entity (we assume that the number of the entity occurs in entity training data larger than a threshold, then it is related entity) in the training data, external url are used to as query, then indri laguage models dirichlet is used, then according to query result, identify item relevance.

ecnu-rst_com_1

Participants | Input | Appendix

  • Run ID: ecnu-rst_com_1
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 72aedff6e6cab61cfb9fddf97a38eca1
  • Run description: Entity name strings are used to filter documents, each entity name is used as an query to do search in the indexed kba corpus. Linear combine language model lda and weighting models results. According to combine result rank value, the higher value means the higher relevance. A threshold is used to identify the item relevance -1,0,1,2.

ecnu-rst_com_2

Participants | Input | Appendix

  • Run ID: ecnu-rst_com_2
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 4888c68e13cff93e0aec6280bcafe4d8
  • Run description: Entity name strings are used to filter documents, each entity name is used as an query to do search in the indexed kba corpus. Linear combine language model lda, weighting models results and similarity based method results. According to combine result value, the higher value means the higher relevance. A threshold is used to identify the item relevance -1,0,1,2.

ecnu-sim_run

Participants | Input | Appendix

  • Run ID: ecnu-sim_run
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: b1f693c255d1f0dc5fb36cc0e0e203bf
  • Run description: Entity name strings are used to filter documents. Use entity training profile, according to each vital or useful streamitem clean text. Filter stop words, get top frequency words, and calculate cosine and jaccard similarity. then linear combine their results

ecnu-ssf_run

Participants | Input | Appendix

  • Run ID: ecnu-ssf_run
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: ff902f6306232be4f4621437c97308b4
  • Run description: Entity Canonical Name strings are used to filter documents, mentioned information in training data and profile information from external_profile are used as keys, then any document containing one of the keys is ranked vital with confidence proportional to length of keys, and the longest sentence containing the longest key is treated as a slot fill for all slot types for the given entity type.

ecnu-ter_wtms_1

Participants | Input | Appendix

  • Run ID: ecnu-ter_wtms_1
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: d21ae8f86762393cce68f1626f12d69c
  • Run description: Entity name strings are used to filter documents. Entity name mension information in training data are used to as query, then terrier weighting models BM25, PL2, TFIDF are used, then according to query result, identify item relevance.

ecnu-ter_wtms_2

Participants | Input | Appendix

  • Run ID: ecnu-ter_wtms_2
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 82e333995d851d14a29d2a8f472c89e3
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information and related entity (we assume that the number of the entity occurs in entity training data larger than a threshold, then it is related entity) in training data are used to as query, then terrier weighting models BM25, PL2, TFIDF are used, then according to query result, identify item relevance.

ecnu-ter_wtms_3

Participants | Input | Appendix

  • Run ID: ecnu-ter_wtms_3
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 74f06212893f72a093dac7b6945d5bd5
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information in training data, external profile are used to as query, then terrier weighting models BM25, PL2, TFIDF are used, then according to query result, identify item relevance.

ecnu-ter_wtms_4

Participants | Input | Appendix

  • Run ID: ecnu-ter_wtms_4
  • Participant: ecnu
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: a06f3eaa7bdcf0cf5698821bf917c499
  • Run description: Entity name strings are used to filter documents. Entity name, name mension information and related entity related entity (we assume that the number of the entity occurs in entity training data larger than a threshold, then it is related entity) in training data, external profile are used to as query, then terrier weighting models BM25, PL2, TFIDF are used, then according to query result, identify item relevance.

IRIT-alpha_100_0.25

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_100_0.25
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: fb4912c61006a0c6c073377e898d37c7
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_100_0.5

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_100_0.5
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: d38491f932521a2f3f05748bf169fc8c
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_100_0.75

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_100_0.75
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: bc83f70427b6f5777155235fd9100609
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_10_0.25

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_10_0.25
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: f1c32f28c0ad912f93f1986a96c43a5e
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_10_0.5

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_10_0.5
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: f5cd62b228ca29d75e98c81b0a541d71
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_10_0.75

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_10_0.75
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: b1a600e9c43660aeb9cb236236577e6f
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_50_0.25

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_50_0.25
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 7d2db350b2fa22b4b097e6f1c145dc1e
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_50_0.5

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_50_0.5
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 540e921d3c11f09177186a2b4747f944
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_50_0.75

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_50_0.75
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 1415ad513bff705cf1f1fb9c66d369c6
  • Run description: Linear combination of vital-training-based score and burst-relevance score

IRIT-alpha_50_0.75T

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-alpha_50_0.75T
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 9e5f9824379a85ad018d18c9b34fe7fc
  • Run description: detecting vitality combining the a vitality-training-score and bursty-relevance-new-dates score

IRIT-ULM_10

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-ULM_10
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 850995344a467413ca436696d2c70784
  • Run description: relevance-training-based score

IRIT-ULM_50

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-ULM_50
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: cd8bc96978c6d74703046789a35124de
  • Run description: relevance-training-based score

IRIT-ULMBuzz_50_0.5T

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-ULMBuzz_50_0.5T
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: feffb6c473679716f6f5bfcf17065b65
  • Run description: Ldetecting vitality using a bursty-new-dates score

IRIT-ULMBuzz_50_0.7T

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-ULMBuzz_50_0.7T
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: a20f935fc481f717f29954809f829e1e
  • Run description: detecting vitality using a bursty-new-dates score

IRIT-VLM_10

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VLM_10
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: fe203b70c1cff95b5864f164d443fadd
  • Run description: vital-language-model based score

IRIT-VLM_50

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VLM_50
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 869e681c21b495f79099d7b97438dbd7
  • Run description: vital-language-model-based score

IRIT-VULMBuz_50_0.5T

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VULMBuz_50_0.5T
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 738c4a40c2f5163e77f0c950bc509973
  • Run description: detecting vitality using a bursty-new-dates score

IRIT-VULMBuz_50_0.7T

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VULMBuz_50_0.7T
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 971a47d8408608313733975771dbf267
  • Run description: detecting vitality using a bursty-new-dates score

IRIT-VULMBuzz_10

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VULMBuzz_10
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: a0519cb2373d32144424f5d48a40f522
  • Run description: bursty vital-or-relevance language model

IRIT-VULMBuzz_50

Participants | Proceedings | Input | Appendix

  • Run ID: IRIT-VULMBuzz_50
  • Participant: IRIT
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 4550c2722bd8e0f1ee82d1fd7a968dd6
  • Run description: bursty vital-or-relevant language model

KobeU-ccr_03

Participants | Proceedings | Input | Appendix

  • Run ID: KobeU-ccr_03
  • Participant: KobeU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 947521b941a1799043e9773cea5302f0
  • Run description: For the documents which are extracted in system_id 'exact_match', run following two phases of filtering. First, classify rating_level of document as relevant (vital or useful) if having low similarity with the irrelevant documents. Next, classify rating_level of document as vital if having low similarity with the model generated from the entity page text of wikipedia and vital documents.

KobeU-ccr_08

Participants | Proceedings | Input | Appendix

  • Run ID: KobeU-ccr_08
  • Participant: KobeU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: c6180d160f1a65384d12734876aa5f0c
  • Run description: Opposite from system_id 'ccr_03', classify rating_level of document as vital if having 'high' similarity with the model generated from the entity page text of wikipedia and vital documents.

KobeU-exact_match

Participants | Proceedings | Input | Appendix

  • Run ID: KobeU-exact_match
  • Participant: KobeU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/5/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e47a88c4bedac0adb6745a82834c63e6
  • Run description: 'Wikipedia redirects' and 'canonical_name' are used as surface form names, then any document containing one of the surface form names is ranked vital. 'Wikipedia redirects' are got from enwiki-20120104-pages-articles.xml.xz.

LSIS-AF_NU_MCE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_NU_MCE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: f0ecde1698c642ca444916fffb38b696
  • Run description: Use the scores of every classifiers GNvsUV, UvsV, VvsOthers, UvsOthers and Single to express a confidence score according to all of those outputs

LSIS-AF_NU_SE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_NU_SE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 81cf35ffbbd7ad33fb46c867c2220ed6
  • Run description: Use score of a global classifier that choose between all four classes.

LSIS-AF_NU_TSE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_NU_TSE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 70464194095bd25611d6446ac2153ef2
  • Run description: Use the score of two classifiers GnvsUV and UvsV. if UV from first classifier then choose class from UvsV output. Else set as garbage.

LSIS-AF_NU_VOE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_NU_VOE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: d813456ad27d52547db23073db6c1a36
  • Run description: Use the score given by VitalvsOthers to determine whether a document is vital. Use SingleEvaluator otherwise.

LSIS-AF_UD_MCE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_UD_MCE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 699a9b4e7614e6b029b1b4c4e78edea6
  • Run description: Use the scores of every classifiers GNvsUV, UvsV, VvsOthers, UvsOthers and Single to express a confidence score according to all of those outputs

LSIS-AF_UD_SE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_UD_SE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 61a52a37405663d8c9d993c0f340f789
  • Run description: Use score of a global classifier that choose between all four classes.

LSIS-AF_UD_TSE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_UD_TSE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e92f7f09049770c54fbee9c5317715e5
  • Run description: Use the score of two classifiers GnvsUV and UvsV. if UV from first classifier then choose class from UvsV output. Else set as garbage.

LSIS-AF_UD_VOE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_UD_VOE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 2f0b1873ece99a9c6395508809ee9889
  • Run description: Use the score given by VitalvsOthers to determine whether a document is vital. Use SingleEvaluator otherwise.

LSIS-AF_US_MCE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_US_MCE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 2d73c999d16bc30c710cccfa9fd6dc15
  • Run description: Use the scores of every classifiers GNvsUV, UvsV, VvsOthers, UvsOthers and Single to express a confidence score according to all of those outputs

LSIS-AF_US_SE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_US_SE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e3df0d0311e242753a87a4c6475795cd
  • Run description: Use score of a global classifier that choose between all four classes.

LSIS-AF_US_TSE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_US_TSE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e9e3ff008f3ec46ca6268d3035bf64ac
  • Run description: Use the score of two classifiers GnvsUV and UvsV. if UV from first classifier then choose class from UvsV output. Else set as garbage.

LSIS-AF_US_VOE

Participants | Proceedings | Input | Appendix

  • Run ID: LSIS-AF_US_VOE
  • Participant: LSIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 7dd3ae3b2bf8dfe4528934b01b36dfc7
  • Run description: Use the score given by VitalvsOthers to determine whether a document is vital. Use SingleEvaluator otherwise.

MSR_KMG-TR_P_All_GA

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_P_All_GA
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 7ff4190ed5a3ea9e3165dc25398093b5
  • Run description: The docs are filtered by time range and event patterns, use all test docs, global adjusted, 0.95.

MSR_KMG-TR_P_All_GA_1

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_P_All_GA_1
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: c7b5dd01961bb285b52447c49df55665
  • Run description: The docs are filtered by time range and event patterns, use all test docs, global adjusted, 0.95.

MSR_KMG-TR_Pattern_All

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_Pattern_All
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: e3812df9ad612509ddd131c98cc9082e
  • Run description: The docs are filtered by time range and event patterns, use all test docs.

MSR_KMG-TR_PC_GA

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_PC_GA
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 1dc0991d1bab06bd4f35979360153089
  • Run description: The docs are filtered by time range and event patterns (cleaned), new time feature, use all test docs, global adjusted, 0.95, cutoff=140.

MSR_KMG-TR_PC_GA_1

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_PC_GA_1
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 30824ff9660969423d1e79118ea8116a
  • Run description: The docs are filtered by time range and event patterns (cleaned), new time feature, use all test docs, global adjusted, 0.95, cutoff=500.

MSR_KMG-TR_PC_GA_2

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_PC_GA_2
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 5b691707b81130d959f8d51d531aeef7
  • Run description: The docs are filtered by time range and event patterns (cleaned+", left"), new time feature, use all test docs, global adjusted, 0.95, cutoff=140.

MSR_KMG-TR_PC_GA_3

Participants | Input | Appendix

  • Run ID: MSR_KMG-TR_PC_GA_3
  • Participant: MSR_KMG
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 5a9c3eed7d0d27fab11ba7e2dd16ec85
  • Run description: The docs are filtered by time range and event patterns (cleaned+", left"), new time feature, use all test docs, global adjusted, 0.95, cutoff=500.

SCU-ssf_1

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_1
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 3b333c1eb7b160dd81f5677b83ad8337
  • Run description: No filters, same score for all results

SCU-ssf_10

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_10
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: bf75b1570d11328f499c51bee331e88f
  • Run description: Output all results for changing slots, majority results for non-changing slots, score on diff from min

SCU-ssf_11

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_11
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: a687b8eb2e75c4b596d3f60821295ba7
  • Run description: No filters on all results, score on diff from min

SCU-ssf_12

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_12
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: a30b517fa07accd66458afe1d41ff316
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on diff from min

SCU-ssf_13

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_13
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 62922c81fb08e1e82479f0a47812e8d9
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on diff from min, relevant rating

SCU-ssf_14

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_14
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: b36bd0ce66befc967b1c793c2c1b8864
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on diff from min, relevant rating, with checking for multivalue slots

SCU-ssf_2

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_2
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 27caf0ea2f3789b97e39c54c08fb9aac
  • Run description: No filters, same score for all results, added addtional slot

SCU-ssf_3

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_3
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: e99bdf19cdc3439558cd342c996f6cda
  • Run description: Output all results for changing slots, majority results for non-changing slots

SCU-ssf_4

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_4
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 84ba7f53222e1bce8c29f78db37d74e0
  • Run description: Output results of changing slots that has 80% of population, and top 1 for non-changing slots

SCU-ssf_5

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_5
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 30361ea6d01f836575f724d1aa44b9a2
  • Run description: Output all results for changing slots, majority results for non-changing slots, score on percentage

SCU-ssf_6

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_6
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 0d499d9fc365368ae88466f1579118f7
  • Run description: No filters on all results, score on percentage

SCU-ssf_7

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_7
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 6840d5ea87f78d2e1eada8426c0702dc
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on percentage

SCU-ssf_8

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_8
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: 62e6d211de5d37f980803811ea01a575
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on percentage, relevant rating

SCU-ssf_9

Participants | Proceedings | Input | Appendix

  • Run ID: SCU-ssf_9
  • Participant: SCU
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ssf-2014
  • MD5: f4756d9aec25f0d5159b5908b8b5b6da
  • Run description: Output results of changing slots that has 70% of population, and top 1 for non-changing slots, score on percentage, relevant rating, with checking for multivalue slots

uiucGSLIS-baseline_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-baseline_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/27/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 8cab9fd9939343d7b1d2a918af9d2cd1
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model, threshold estimated using maximum empirical F1

uiucGSLIS-baseline_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-baseline_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/27/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 660f32f771d6bd5d3838ea8072b0f1dc
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model, threshold estimated using maximum empirical F1

uiucGSLIS-length_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-length_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 68aa76da7ccbbc4e05acba690250c66b
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from high vital training qrels, documents scored using KL-divergence retrieval model, threshold estimated using maximum empirical F1, length prior

uiucGSLIS-length_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-length_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 20fb592261f684d8bfd80c1a9c42a6da
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model, threshold estimated using maximum empirical F1, length prior

uiucGSLIS-pdsrc_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-pdsrc_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 393a3687424fd2c52e87ef7aa29c23a3
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model combined with previous docs and source scores, threshold estimated using maximum empirical F1

uiucGSLIS-pdsrc_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-pdsrc_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: b081b6a6140a39b66681d8d81f6fb2b3
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model combined with previous docs and source scores, threshold estimated using maximum empirical F1

uiucGSLIS-pdverb_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-pdverb_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: f84c2500b07c85e4f2e3506c11c5e969
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model, score combined with previous number of documents and verb scores, threshold estimated using maximum empirical F1

uiucGSLIS-pdverb_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-pdverb_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: ccf8da9307d572a1eb2a19fd638cf565
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model combined with previous docs and verb scores, threshold estimated using maximum empirical F1

uiucGSLIS-prevdocs_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-prevdocs_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 068bf2d046a83cfd6d7a22a809243ca0
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model combined with previous docs score, threshold estimated using maximum empirical F1

uiucGSLIS-prevdocs_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-prevdocs_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: c34d312a56b62cace8bf21a8fd39b649
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model combined with previous docs score, threshold estimated using maximum empirical F1

uiucGSLIS-sourcelen_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-sourcelen_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 54e7811979a517ed032ace8ef86bcaf7
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model, score combined with source and length scores, threshold estimated using maximum empirical F1

uiucGSLIS-sourcelen_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-sourcelen_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 42c09ccc744551fa367d57a2cab891e1
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model combined with source and length scores, threshold estimated using maximum empirical F1

uiucGSLIS-verbsource_rm3

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-verbsource_rm3
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 70f6f749d883f70103449de98f97f241
  • Run description: Entity titles are used in 2-word unordered window for the initial query, RM3 query model from true vital training set, documents scored using KL-divergence retrieval model combined with verb and source scores, threshold estimated using maximum empirical F1

uiucGSLIS-verbsource_sf

Participants | Proceedings | Input | Appendix

  • Run ID: uiucGSLIS-verbsource_sf
  • Participant: uiucGSLIS
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/29/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: dcf3850c5bacf1630a186b9b7297d118
  • Run description: Entity titles are used in 2-word unordered window for the initial query, documents scored using KL-divergence retrieval model combined with verb model and source scores, threshold estimated using maximum empirical F1

UW-basic_multitask

Participants | Proceedings | Input | Appendix

  • Run ID: UW-basic_multitask
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/28/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 21e91f3ca4b858aa3777255c9d8afd6e
  • Run description: Same as basic_single run but U-V classifier uses multitask learning. U-V trained with 25 + 25x71 features. Used doc-level and doc-entity features

UW-basic_single

Participants | Proceedings | Input | Appendix

  • Run ID: UW-basic_single
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/28/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 487bce4d12e36a5ca313b427298fd855
  • Run description: Two GBT classifiers with 25 features each. First one predicts relevant and non-relevant. The same R-NR classifier is used in all runs. Second one, distinguishes between useful and vital. Used doc-level and doc-entity features

UW-clu_dyn_a08_g04

Participants | Proceedings | Input | Appendix

  • Run ID: UW-clu_dyn_a08_g04
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/4/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 9be6ec81e5f3847e700d047366523940
  • Run description: Same as mean_stat run but setting alpha to 0.8 and gamma to 0.4. It uses clustering features for verbs and nouns. U-V classifier is an Extremely Randomized Tree ensemble. Used 150 weaker classifiers

UW-clu_dyn_nv_e

Participants | Proceedings | Input | Appendix

  • Run ID: UW-clu_dyn_nv_e
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 426c1de86c91d9d95bcc990ea3e2bddb
  • Run description: Similar to clu_dyn_a08_g04 run. Setting alpha to 0.8, gamma decrease to 200 and gamma increase to 0.9. It uses clustering features for verbs and nouns with exponential decay and timeliness per entity. U-V classifier is an Extremely Randomized Tree ensemble. Used 150 weaker classifiers of depth 100

UW-clu_stat_a08

Participants | Proceedings | Input | Appendix

  • Run ID: UW-clu_stat_a08
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/2/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 342a9c86b42426fb1da739467f47c037
  • Run description: Same as mean_stat run but setting alpha (threshold for cluster assignment) to 0.8. It uses clustering features for verbs and nouns. U-V classifier is an Extremely Randomized Tree ensemble. Gamma set to 1

UW-embedding_comb

Participants | Proceedings | Input | Appendix

  • Run ID: UW-embedding_comb
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/28/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 1b3904297b4fc3c150fbd643af52c042
  • Run description: Same as basic_multitask run, but U-V classifier is trained with word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns and verbs that surround full or partial matches of the entities names

UW-embedding_pos

Participants | Proceedings | Input | Appendix

  • Run ID: UW-embedding_pos
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/28/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 06ac88b40aeca7cce8d9ad075f1b7ffe
  • Run description: Same as embedding_combined run, but using two separate embeddings for nouns and verbs computed with Google word2vec tool

UW-f_basic_multi

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_basic_multi
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 1884f827352b6ee0f775803412be13b0
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features but with multitask learning

UW-f_basic_single

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_basic_single
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 031503c6b0f5921cca3a8c6da2a1b492
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features

UW-f_clust_dyn

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_clust_dyn
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: fc6c3feeaba3916090559670895c699d
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (one embedding per word type, i.e. one for nouns, one for verbs and one for proper nouns). Used clustering features. Alpha is set to 0.8. The decreasing gamma is set to 1, and the increasing gamma is set to 0.1. Alpha is the threshold between 0 and 1 used to decide cluster assignment. Gamma controls the increment/decrement factor of the timeliness feature

UW-f_clust_stat

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_clust_stat
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: cc7d10e19e29d0b0680e62bf4c10aa8f
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (one embedding per word type, i.e. one for nouns, one for verbs and one for proper nouns). Used clustering features. Alpha is set to 0.8. The decreasing gamma is set to 0, and the increasing gamma is set to 1. Alpha is the threshold between 0 and 1 used to decide cluster assignment. Gamma controls the increment/decrement factor of the timeliness feature

UW-f_emb_comb

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_emb_comb
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/10/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 539581d24409efab2f0d02fc766bf73b
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (all combined into a single embedding)

UW-f_emb_pos

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_emb_pos
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 6e073ddf1d0418b9cd28696f9154838f
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (one embedding per word type, i.e. one for nouns, one for verbs and one for proper nouns)

UW-f_mean_dyn

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_mean_dyn
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: ee1baa089bedc7cba9c36a525a5c772e
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (one embedding per word type, i.e. one for nouns, one for verbs and one for proper nouns). Used clustering features. Alpha is set to 1. The decreasing gamma is set to 1, and the increasing gamma is set to 0.1. Alpha is the threshold between 0 and 1 used to decide cluster assignment. Gamma controls the increment/decrement factor of the timeliness feature

UW-f_mean_stat

Participants | Proceedings | Input | Appendix

  • Run ID: UW-f_mean_stat
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/11/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 7b0a0bafdf153754fd36001ffabee377
  • Run description: Two extremely randomized classifiers in cascade. First predicts relevant/non-relevant docs. Second predicts useful/vital. Used 25 basic doc-level and doc-entity features. Also used word embeddings computed with Google word2vec tool, using the pre-trained Google news dataset file, with the nouns, verbs and proper nouns that surround full matches of the entities names (one embedding per word type, i.e. one for nouns, one for verbs and one for proper nouns). Used clustering features. Alpha is set to 1. The decreasing gamma is set to 0, and the increasing gamma is set to 1. Alpha is the threshold between 0 and 1 used to decide cluster assignment. Gamma controls the increment/decrement factor of the timeliness feature

UW-mean_dyn_g06

Participants | Proceedings | Input | Appendix

  • Run ID: UW-mean_dyn_g06
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/4/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: dcea325e029372676051e3d7c77a366f
  • Run description: Same as mean_stat run but setting gamma (timeliness feature increment/decrement factor) to 0.6. It uses clustering features for verbs and nouns. U-V classifier is an Extremely Randomized Tree ensemble. Alpha set to 1. Used 150 weaker classifiers for the ensemble

UW-mean_stat

Participants | Proceedings | Input | Appendix

  • Run ID: UW-mean_stat
  • Participant: UW
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/2/2014
  • Type: automatic
  • Task: kba-ccr-2014
  • MD5: 67a035a0a161443d05b415bcde6c6454
  • Run description: Same as embedding_pos run but it also uses clustering features for verbs and nouns. U-V classifier is now an Extremely Randomized Tree ensemble. Both alpha (threshold for cluster assignment) and gamma (timeliness feature increment/decrement factor) are set to 1

WHU_IRGroup-baseline

Participants | Proceedings | Input | Appendix

  • Run ID: WHU_IRGroup-baseline
  • Participant: WHU_IRGroup
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 8/20/2014
  • Type: manual
  • Task: kba-ccr-2014
  • MD5: ff63c93aaf40b55095bb807787bd69d8
  • Run description: this is only a demo system.

WHU_IRGroup-BM_TF

Participants | Proceedings | Input | Appendix

  • Run ID: WHU_IRGroup-BM_TF
  • Participant: WHU_IRGroup
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: manual
  • Task: kba-ccr-2014
  • MD5: 9d2702eee669a045cf2d5598cab62176
  • Run description: This is the baseline system with BM temporal feature. BM temporal feature indicate the temporal feature used by BIT & MSRA in KBA 2013.

WHU_IRGroup-BM_TF_3

Participants | Proceedings | Input | Appendix

  • Run ID: WHU_IRGroup-BM_TF_3
  • Participant: WHU_IRGroup
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: manual
  • Task: kba-ccr-2014
  • MD5: f1d3de9ab48a86c08c43de27eaee07be
  • Run description: This is the baseline system with BM temporal feature. BM temporal feature indicate the temporal feature used by BIT & MSRA in KBA 2013.

WHU_IRGroup-CUSTOM_TF_FIXED

Participants | Proceedings | Input | Appendix

  • Run ID: WHU_IRGroup-CUSTOM_TF_FIXED
  • Participant: WHU_IRGroup
  • Track: Knowledge Base Acceleration
  • Year: 2014
  • Submission: 9/9/2014
  • Type: manual
  • Task: kba-ccr-2014
  • MD5: ca7c40c79dbcb800f0d5b91025ef6d2c
  • Run description: this is a system using our custom temporal features using fixed window size.