Skip to content

Runs - Real-time Summarization 2017

adv_lirmm-Run1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: adv_lirmm-Run1
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: f7ea6c75692aaa98d6ab108f7fb78aa1
  • Run description: This run rely on a strict filtering function based on the cosine similarity measure. The approach is built on top on the distributed real-time computation system Apache Storm.
  • Code: https://bitbucket.org/amjedbj/eidorm

adv_lirmm-Run2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: adv_lirmm-Run2
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 10114cc8686fa9d85bdd791092175160
  • Run description: This run rely on a strict filtering function based on the cosine similarity measure. The approach is built on top on the distributed real-time computation system Apache Storm.
  • Code: https://bitbucket.org/amjedbj/eidorm

adv_lirmm-Run3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: adv_lirmm-Run3
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 629e6f5460f56b7edccf02b102583921
  • Run description: This run rely on a strict filtering function based on the cosine similarity measure. The approach is built on top on the distributed real-time computation system Apache Storm.
  • Code: https://bitbucket.org/amjedbj/eidorm

advanse_lirmm-Run1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: advanse_lirmm-Run1-A
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

advanse_lirmm-Run2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: advanse_lirmm-Run2-A
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

advanse_lirmm-Run3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: advanse_lirmm-Run3-A
  • Participant: advanse_lirmm
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

bjut_tmg

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: bjut_tmg
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 3bf68b7cddbe051d619e85087f3f8280
  • Run description: use Bing Search API to do query expansion

bjutg

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: bjutg
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Type: automatic
  • Task: b
  • MD5: 639c762f9dd9e3b6a72cda8f5a65e532
  • Run description: This run use search engines and Wikipedia. And those resources are not timely.

bjutgs

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: bjutgs
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Type: automatic
  • Task: b
  • MD5: 21c66e1367d04df58c1d589194636895
  • Run description: This run use search engines and Wikipedia. And those resources are not timely. And this run is receive more tweets than 'bjutg' because of the silent day.

BL1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: BL1-A
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

BL2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: BL2-A
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

BL3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: BL3-A
  • Participant: BJUT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

HLJIT_l2r

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: HLJIT_l2r
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Task: b
  • MD5: 03ac6796dccbb6f0749ed6698e15424a
  • Run description: We calculate the relevance between the topic and the tweet by learning to rank technique, the similarities between the expanded topic and the tweet are used as the features to train the learning to rank model. The query is expanded by using Relevance-based Language Model from the top 50 retrieval results returned by Google. The similarity of the query and the twitter is calculated measured by the expanded topic and the url.

HLJIT_rank_svm

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: HLJIT_rank_svm
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Task: b
  • MD5: 3babfdd596556b176993148090f0e16f
  • Run description: Using the algorithm of learning to rank to learn the model for each query. Feedback is exploited.

ICTNET-Run1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNET-Run1
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 1168a5ff80391e31a624da096d144b16
  • Run description: this run uses Wikipedia corpus to train the word2vec model as external resources
  • Code: https://github.com/stanpcf/rts

ICTNET-run1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: ICTNET-run1-A
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

ICTNET-Run2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNET-Run2
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 08340be125095ad008a543c565e8d01c
  • Run description: 1. this run used Wikipedia corpus to train the word2vec model as external resources 2. this run used google search api to expand the topics 3. this run uses the twitter data streamed over passed 3 days as corpus to train the TFIDF model
  • Code: https://github.com/stanpcf/rts

ICTNET-run2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: ICTNET-run2-A
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

ICTNET-Run3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNET-Run3
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: f3fc8eca881acd111f99227d8f6203b5
  • Run description: 1. this run used Wikipedia corpus to train the word2vec model 2. this run used google search api to expand the topics 3. this run used passed 3 days' twitter data as corpus to train the TFIDF model
  • Code: https://github.com/stanpcf/rts

ICTNET-run3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: ICTNET-run3-A
  • Participant: ICTNET
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

IRIT-Run1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: IRIT-Run1-A
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

IRIT-Run2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: IRIT-Run2-A
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

IRIT-Run3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: IRIT-Run3-A
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

IRIT-RunB1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: IRIT-RunB1
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: b
  • MD5: 1192d3fc16d99b31d7947412762d8fbc
  • Run description: In this run we first apply a simple filter that discards any tweets that fulfil one of the following conditions: text length less than five terms, contain more than one URL or three hashtags and the number of word overlap with the title and the description of the query is less than a minimum of either 3 words or the size of the title and the description of the query. Then tweets that pass a binary classifier are selected for a daily summary. The binary classifier was trained using TREC RTF 2015 data set. In this run, we use the live assessment feedback to retrain the binary classifier periodically (active learning) Tweets that pass the filtering stage are clustered and the summary generation was formulated as an optimization problem to select a subset of tweets that maximizes the global summary relevance and fulfills constraints related to non-redundancy, coverage, temporal diversity and summary length. Branch and bound algorithm is used to resolve the optimization problem.
  • Code: http://irit.fr

IRIT-RunB2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: IRIT-RunB2
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 35758dff79302cf911b01dd2157f9dc5
  • Run description: In this run we first apply a simple filter that discards any tweets that fulfil one of the following conditions: text length less than five terms, contain more than one URL or three hashtags and the number of word overlap with the title and the description of the query is less than a minimum of either 3 words or the size of the title and the description of the query. Then tweets that pass a binary classifier are selected for a daily summary. The binary classifier was trained using TREC RTF 2015 data set. The binary classifier was trained using TREC RTF 2015 data set. In this run, the live assessment feedback was not used. Tweets are clustered and the summary generation was formulated as an optimization problem to selects a subset of tweets that maximizes the global summary relevance and fulfills constraints related to non-redundancy, coverage, temporal diversity and summary length.
  • Code: http://irit.fr

IRIT-RunB3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: IRIT-RunB3
  • Participant: IRIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: b
  • MD5: 306e61d4e39d09f4e42073a5cc54260e
  • Run description: In this run we first apply a simple filter that discards any tweets that fulfil one of the following conditions: text length less than five terms, contain more than one URL or three hashtags and the number of word overlap with the title and the description of the query is less than a minimum of either 3 words or the size of the title and the description of the query. Then tweets that pass a binary classifier are selected for a daily summary. The binary classifier was trained using TREC RTF 2015 data set. In this run, we use the live assessment feedback to retrain the binary classifier periodically (active learning) To generate a daily summary, the top-10 weighted tweets are iteratively selected with the exclusion of those having a similarity above a predefined threshold (0.75) with the current summary (tweets already selected). To evaluate the similarity between two tweets we use word embedding based similarity function.
  • Code: http://irit.fr

IRLAB-DAIICT

Participants | Input | Summary | Appendix

  • Run ID: IRLAB-DAIICT
  • Participant: IRLAB_DAIICT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 12cf945c9e9ee35cd115c6c0e2f71f79
  • Run description: NE base query expansion using Twitter 2 week corpus.Language model with JM smoothing is used to calculate Similarities. Language model parameter estimated using grid search. Relevance threshold and silent day estimated using last year dataset. for summarization, Graph based method used.

IRLAB-LDRP2

Participants | Input | Summary | Appendix

  • Run ID: IRLAB-LDRP2
  • Participant: IRLAB_DAIICT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: fe43f376b6cf20c470bc10caba3ce063
  • Run description: NE base query expansion using Twitter 2 week corpus.Language model with Dirichlet smoothing is used to calculate Similarities. Language model parameter estimated using grid search. Relevance threshold and silent day estimated using last year dataset. for summarization, Graph based method used.

irlab-Run1-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: irlab-Run1-A
  • Participant: DA_IICT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

IRLAB_LDRP

Participants | Input | Summary | Appendix

  • Run ID: IRLAB_LDRP
  • Participant: IRLAB_DAIICT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: e8e8a0e914fdd9e0dae406d114ca93a3
  • Run description: NE base query expansion using Twitter's last 2 week corpus. Okapi BM 25 model is used to calculate Similarities. Relevance threshold and silent day estimated using last year dataset. for summarization, Graph/jaccard based method used.

IUB

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: IUB
  • Participant: SOIC
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: b769ac986e1cc0d41f53dbb78c1d21fd
  • Run description: use wordnet as extended resource

ldrp-Run2-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: ldrp-Run2-A
  • Participant: DA_IICT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

lm-jm-lambda0.5

Participants | Input | Summary | Appendix

  • Run ID: lm-jm-lambda0.5
  • Participant: ISIKol
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: d3643c97917b0a2295f38a3283038566
  • Run description: Tried to explore collection statistics' influence on Micro blog.

NOVASearchB1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NOVASearchB1
  • Participant: NOVASearch
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: 0e87f0f97eb04eed86ca682c68b1d150
  • Run description: Uses query expansion using external news sources and tweets are ranked using a learning to rank model

NOVASearchB2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NOVASearchB2
  • Participant: NOVASearch
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: fbbc0d3e45a3accf01bf6d4b853a0bb8
  • Run description: Uses query expansion using external news sources from verified accounts in Twitter

NOVASearchB3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NOVASearchB3
  • Participant: NOVASearch
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: 8d9c711a12eb62a0ca607d52164b740e
  • Run description: Queries were expanded by identifying entities in the narrative that aren't already in the Title. The URL in the tweets were expanded to retrieve more text.

pertopicburst-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: pertopicburst-A
  • Participant: umd-hcil
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PKUICSTRunA1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PKUICSTRunA1-A
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PKUICSTRunA2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PKUICSTRunA2-A
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PKUICSTRunA3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PKUICSTRunA3-A
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PKUICSTRunB1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PKUICSTRunB1
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/5/2017
  • Type: automatic
  • Task: b
  • MD5: 6f1aabc27002089b916be5105c888aac
  • Run description: language model, using dirichlet smoothing. two step threshold.

PKUICSTRunB2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PKUICSTRunB2
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/5/2017
  • Type: automatic
  • Task: b
  • MD5: ae9e7977399b7319a6c977e8f34e521f
  • Run description: vector space model with tf-idf. two step threshold.

PKUICSTRunB3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PKUICSTRunB3
  • Participant: PKUICST
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/5/2017
  • Type: automatic
  • Task: b
  • MD5: 4bcee037ef21b3c137a624439a21e1ac
  • Run description: mixed scoring algorithm. language model with dirichlet smoothing and vector space model with tf-idf. two step threshold. At the end of each day, use the previous day's tweets for query expansion. the tweets was collected by twitter-search api

PRNA-A1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PRNA-A1-A
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PRNA-A2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PRNA-A2-A
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PRNA-A3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: PRNA-A3-A
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

PRNA-B1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PRNA-B1
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: ddef38a8a751bc844a52e3773e43d04b
  • Run description: This run uses named entities and paraphrases as features.

PRNA-B2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PRNA-B2
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 57c44f6d3debd919b05e5d235108ea75
  • Run description: we train a supervised linear regression model per topic where the training data is automatically labeled based on how many title words and description words for a topic are present in a tweet (collected before the competition start time), assuming more matches correlate with higher relevance The model is applied to a test tweet to predict a relevance score, selected when above a threshold score.

PRNA-B3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: PRNA-B3
  • Participant: prna
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: 19e86b44fd2065842ba7da1d57dee93a
  • Run description: This run ranks tweets by measuring semantic relevance of a tweet for a given topic using a novel attention-based convolutional neural network model.

qFB_url

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: qFB_url
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/8/2017
  • Type: automatic
  • Task: b
  • MD5: 87a2625d97e67f77fa53d8505ef9d42a
  • Run description: The query is expanded by using Relevance-based Language Model from the top 50 retrieval results returned by Google. The similarity of the query and the twitter is calculated measured by the expanded topic and the url.

QUBaseline-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: QUBaseline-A
  • Participant: QU
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

QUExp-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: QUExp-A
  • Participant: QU
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

QUExpDyn-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: QUExpDyn-A
  • Participant: QU
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

retweet-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: retweet-A
  • Participant: umd-hcil
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

SHNU_run1

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: SHNU_run1
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: fd413a51aaf3f283340de11b447f427e
  • Run description: This is run1,the clientid is Voat1N7Fd0Wc .

SHNU_run1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: SHNU_run1-A
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

SHNU_run2

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: SHNU_run2
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: b412e88197e52155074caf9ea87a757e
  • Run description: This is run2,the clientid is A9ImWPyrAzxo .

SHNU_run2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: SHNU_run2-A
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

SHNU_run3

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: SHNU_run3
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: 31cf7d40c30d3a23ae9d4ac8b831d724
  • Run description: This is run3,the clientid is 9bFTfkK7pZa4 .

SHNU_run3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: SHNU_run3-A
  • Participant: S.T
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

SOIC-Run1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: SOIC-Run1-A
  • Participant: SOIC
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

testRun1-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: testRun1-A
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

testRun2-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: testRun2-A
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

testRun3-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: testRun3-A
  • Participant: HLJIT
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

udelRun081D-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: udelRun081D-A
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

udelRun081D-B

Participants | Input | Summary | Appendix

  • Run ID: udelRun081D-B
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: e3efdc0dfcb0f2fff675fa42ff64bfc5
  • Run description: A language modeling approach was used to filter tweets. Language models for this run were generated using the given information for every topic and top tweets associated (extracted using Twitter API) with those topics. These language models and their score thresholds were updated in real-time using the feedback from mobile assessors. Selected tweets were then clustered using cosine similarity for controlling redundancy.

udelRun081HT-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: udelRun081HT-A
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

udelRun081HT-B

Participants | Input | Summary | Appendix

  • Run ID: udelRun081HT-B
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: a877967383c52e73479a4ec26d1aa8bc
  • Run description: A language modeling approach was used to filter tweets. Language models for this run were generated using topic titles and top hashtags (extracted using Twitter API). Selected tweets were then clustered using cosine similarity for controlling redundancy. And mobile assessments were used to dynamically update score thresholds.

udelRun081HTD-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: udelRun081HTD-A
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

udelRun081HTD-B

Participants | Input | Summary | Appendix

  • Run ID: udelRun081HTD-B
  • Participant: udel
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/6/2017
  • Task: b
  • MD5: c3df9484b4e83fc6a8d3dc47c3b4173d
  • Run description: A language modeling approach was used to filter tweets. Language models for this run were generated using topic titles and top hashtags (extracted using Twitter API). These language models and their score thresholds were updated in real-time using the feedback from mobile assessors. Selected tweets were then clustered using cosine similarity for controlling redundancy.

UDInfoBL-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: UDInfoBL-A
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

UDInfoEXP-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: UDInfoEXP-A
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

UDInfoJac

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDInfoJac
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 4a4e523108f59df450092607d19d9eb0
  • Run description: Using query term relation features to detect silent days. The detector is trained on data from microblog 2011, microblog 2015, and RTS 2016. The F2EXP ranking function is used to retrieve tweets.

UDInfoSDWR-A

Participants | Proceedings | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: UDInfoSDWR-A
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

UDInfoW2VPre

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDInfoW2VPre
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 45d2815c62ddecceaf263a71b2787f0a
  • Run description: Using query term relation features to detect silent days. The detector is trained on data from microblog 2011, microblog 2015, and RTS 2016. The F2EXP ranking function is used to retrieve tweets. In order to check whether a tweet is redundant, the tweets are converted to vectors using word2vec vectors that were trained on a Google news data set. We then compute the cosine similarity of these vectors and set a static similarity threshold.

UDInfoW2VTWT

Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDInfoW2VTWT
  • Participant: udel_fang
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: b60ddc1b819aa9b73c5badf782a53177
  • Run description: Using query term relation features to detect silent days. The detector is trained on data from microblog 2011, microblog 2015, and RTS 2016. The F2EXP ranking function is used to retrieve tweets. In order to check whether a tweet is redundant, the tweets are converted to vectors using word2vec vectors that were trained on the tweets crawled during the 12-day period prior to the evaluation period (These tweets were crawled using stream api of Twitter). We then compute the cosine similarity of these vectors and set a static similarity threshold.

umc_hcil_ptbv1

Participants | Input | Summary | Appendix

  • Run ID: umc_hcil_ptbv1
  • Participant: umd-hcil
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 1d9b502672bcb3692c9f29f487b8208c
  • Run description: This run tracks topic-related tweets in real time, generates scores for the level of "burstiness" a token experiences, and returns tweets that include these bursty tokens. This system allows the flow of data to determine when a tweet should be pushed to the user. Bursts are also recorded per topic. Thresholds for burstiness and topic relevance were calculated from prior TREC RTS runs.
  • Code: https://github.com/cbuntain/UMD_HCIL_TREC2015

umc_hcil_rtv1

Participants | Input | Summary | Appendix

  • Run ID: umc_hcil_rtv1
  • Participant: umd-hcil
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Type: automatic
  • Task: b
  • MD5: 588303c7e79e81586e1fa904c9e65445
  • Run description: This run tracks topic-related tweets in real time and scores based on retweets. If a tweet has more than 100 retweets, it becomes a candidate for pushing to the user.
  • Code: https://github.com/cbuntain/UMD_HCIL_TREC2015

WuWien-Run1-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: WuWien-Run1-A
  • Participant: WUWien
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

WuWien-Run2-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: WuWien-Run2-A
  • Participant: WUWien
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a

WuWien-Run3-A

Participants | Input | Summary (Batch) | Summary (Mobile) | Appendix

  • Run ID: WuWien-Run3-A
  • Participant: WUWien
  • Track: Real-time Summarization
  • Year: 2017
  • Submission: 8/7/2017
  • Task: a