Skip to content

Runs - Temporal Summarization 2014

1APSalRed

Participants | Input | Appendix

  • Run ID: 1APSalRed
  • Participant: cunlp
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: b6fec46cb9982daeaa6e011cdcb48c03
  • Run description: Run uses Gaussian process regression model to predict salience (Trained using last years nuggets) of input text. Salience is penalized based on semantic distance to previous updates. Penalized salience is input to AP clustering algorithm and along with sentence semantic similarities. Cluster exemplars are selected as updates.

2APSal

Participants | Input | Appendix

  • Run ID: 2APSal
  • Participant: cunlp
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: 9fc1fc8da7db1c9c77336d148d860b73
  • Run description: Run uses Gaussian process regression model to predict salience (Trained using last years nuggets) of input text. Salience is input to AP clustering algorithm and along with sentence semantic similarities. Cluster exemplars are selected as updates.

3AP

Participants | Input | Appendix

  • Run ID: 3AP
  • Participant: cunlp
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: f2c22552fff3c598b8f4c530f51c0236
  • Run description: AP clustering algorithm run hourly, all inputs are given uniform salience. Cluster exemplars are selected as updates.

Cluster1

Participants | Proceedings | Input | Appendix

  • Run ID: Cluster1
  • Participant: BUPT_PRIS
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 8/25/2014
  • Task: main
  • MD5: 7af9d6b5b5a84f61b34428cfa3a9c0eb
  • Run description: It uses the filtered subset of the KBA corpus.

Cluster2

Participants | Proceedings | Input | Appendix

  • Run ID: Cluster2
  • Participant: BUPT_PRIS
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 8/28/2014
  • Task: main
  • MD5: cb6305f551e0ed537ab618f9c7b14873
  • Run description: It uses the filtered subset of the KBA corpus.

Cluster3

Participants | Proceedings | Input | Appendix

  • Run ID: Cluster3
  • Participant: BUPT_PRIS
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/1/2014
  • Task: main
  • MD5: e660472fba4ed29c8d6bb5f02e7924f7
  • Run description: It uses the filtered subset of the KBA corpus

Cluster4

Participants | Proceedings | Input | Appendix

  • Run ID: Cluster4
  • Participant: BUPT_PRIS
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/2/2014
  • Task: main
  • MD5: 0419be2bb50a54f5ef0950dd75b73442
  • Run description: It uses the filtered subset of the KBA corpus

KW30H10NW300

Participants | Proceedings | Input | Appendix

  • Run ID: KW30H10NW300
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: fd4e6d62edebd630668aa50df555f983
  • Run description: - Learn top-30 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-10 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 5 minutes to check the novelty of updates.

KW30H5NW300

Participants | Proceedings | Input | Appendix

  • Run ID: KW30H5NW300
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: a1d294cf1b3f326d998729eccd643eea
  • Run description: - Learn top-30 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-5 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 5 minutes to check the novelty of updates.

KW30H5NW3600

Participants | Proceedings | Input | Appendix

  • Run ID: KW30H5NW3600
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: c06f2108ec2841136ac43ce624d42815
  • Run description: - Learn top-30 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-5 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 1-day to check the novelty of updates.

KW80H10NW300

Participants | Proceedings | Input | Appendix

  • Run ID: KW80H10NW300
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: da5e6cbe36663d600e948d38674b3e1e
  • Run description: - Learn top-80 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-10 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 5 minutes to check the novelty of updates.

KW80H5NW300

Participants | Proceedings | Input | Appendix

  • Run ID: KW80H5NW300
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: b5cfbd1dd3ee7b017a2e187266938b12
  • Run description: - Learn top-80 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-5 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 5 minutes to check the novelty of updates.

KW80H5NW3600

Participants | Proceedings | Input | Appendix

  • Run ID: KW80H5NW3600
  • Participant: IRIT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/3/2014
  • Task: main
  • MD5: 71de130379b449534d6a903c4761fa2e
  • Run description: - Learn top-80 event related terms from the events nuggets of trec-ts-2013. - Select from each hour the top-5 documents based on the TF of event query. - Ranking the updates based on the learned terms. - Using the cosine similarity with a time-window of 1-day to check the novelty of updates.

Q0

Participants | Proceedings | Input | Appendix

  • Run ID: Q0
  • Participant: BJUT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 8/24/2014
  • Task: main
  • MD5: 3da55a34bbf0da1c568fd7f9e700632b
  • Run description: First run using filtered subset of the KBA corpus

Q1

Participants | Proceedings | Input | Appendix

  • Run ID: Q1
  • Participant: BJUT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 8/25/2014
  • Task: main
  • MD5: 5a7ef4865b143a4f1c30350f9709fe70
  • Run description: Second run using filtered subset of the KBA corpus

Q2

Participants | Proceedings | Input | Appendix

  • Run ID: Q2
  • Participant: BJUT
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 8/25/2014
  • Task: main
  • MD5: 6444fe4cbb51c4eb112f8e614d0b1457
  • Run description: In order to get a system that could emit relevant and novel sentences to an topic,in this run, first we download TREC-TS_2014F(filtered subset of the KBA corpus) as our corpus; Second, make data decryption and parsing;Third,index the corpus and retrieval relevant sentences based on topics;Forth,cluster the sentences; Finally, select some sentences and sort them by time.

run1

Participants | Proceedings | Input | Appendix

  • Run ID: run1
  • Participant: ICTNET
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: 7f3b6160d81f1fd9bed58f4ce20c7242
  • Run description: I only use the filtered subset of the KBA corpus.

run2

Participants | Proceedings | Input | Appendix

  • Run ID: run2
  • Participant: ICTNET
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: b02ff1f4bc90b7f31a583c8c38384bf8
  • Run description: I only use the filtered subset of the KBA corpus.

run3

Participants | Proceedings | Input | Appendix

  • Run ID: run3
  • Participant: ICTNET
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: ab0c963ed5d8a23d80a6138f1db04f61
  • Run description: I only use the filtered subset of the KBA corpus.

run4

Participants | Proceedings | Input | Appendix

  • Run ID: run4
  • Participant: ICTNET
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: 85c46a428f3d2c2fb09f9fe8280c21e2
  • Run description: I only use the filtered subset of the KBA corpus.

uogTr2A

Participants | Proceedings | Input | Appendix

  • Run ID: uogTr2A
  • Participant: uogTr
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: 1924d774ce3a7bffd612d22fc03e8b2e
  • Run description: Multi-stage Real-time Filtering approach. Uses a sentence window of size 2 within each document to identify relevant sentences. Only sentences from the news stream are considered. Data Sources: - TREC-TS-2014 (News Streams Only) - Wikipedia - Freebase - DBpedia - 2008 Reuters Headline corpus

uogTr4A

Participants | Proceedings | Input | Appendix

  • Run ID: uogTr4A
  • Participant: uogTr
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: c8bea33f73e35bfcb527a43b0bdcaff5
  • Run description: Multi-stage Real-time Filtering approach. Uses a sentence window of size 4 within each document to identify relevant sentences. Only sentences from the news stream are considered. Data Sources: - TREC-TS-2014 (News Streams Only) - Wikipedia - Freebase - DBpedia - 2008 Reuters Headline corpus

uogTr4AC

Participants | Proceedings | Input | Appendix

  • Run ID: uogTr4AC
  • Participant: uogTr
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: 526cb6598b72a61d3907102f44c1f6e0
  • Run description: Multi-stage Real-time Filtering approach. Uses a sentence window of size 4 within each document to identify relevant sentences. Applies a machine learned classifier to identify topical and good quality sentences. Only sentences from the news stream are considered. Data Sources: - TREC-TS-2014 (News Streams Only) - Wikipedia - Freebase - DBpedia - 2008 Reuters Headline corpus

uogTr4ARas

Participants | Proceedings | Input | Appendix

  • Run ID: uogTr4ARas
  • Participant: uogTr
  • Track: Temporal Summarization
  • Year: 2014
  • Submission: 9/4/2014
  • Task: main
  • MD5: e6e2088eaa49486d5d308d402cc52caf
  • Run description: Multi-stage Rank then Select approach. Uses a sentence window of size 4 within each document to identify relevant sentences. Only sentences from the news stream are considered. Data Sources: - TREC-TS-2014 (News Streams Only) - Wikipedia - Freebase - DBpedia - 2008 Reuters Headline corpus