Runs - Real-time Summarization 2016¶

AmILPWSEBM¶

Participants | Input | Summary | Appendix

Run ID: AmILPWSEBM
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 1496ec57dc8f7366723bfc69d40eca69
Run description: The proposed approach first filters out all tweets that do not contain at least a minimum of either a predefined constant (K=2) or the number of words in the query title; then tweets that pass this filter are clustered incrementally into aspects. At the end of a day and for each topic, the approach selects a subset of tweets that cover as many aspects as possible within the specified length limit (100 tweets per day). Tweet selection problem is formulated as an Integer Linear Programming (ILP) problem that is solved through a standard branch-and-bound algorithm. We use ILP to optimize a global objective function for tweet selection, which is based on the relevance of tweets with respect to the query (user interest). We select tweets which receive the highest relevance score with respect to the user interest from each aspect cluster subject to a series of constraints related to redundancy, coverage and length limit. Extended Boolean model is used to estimate the relevance score of tweets in which the similarity between the query term and all the terms of an incoming tweet T is considered as the weight of query terms. The similarity between two words is measured by cosine similarity between their vector which are generated by wrd2vec model using a training data
Code: http://irit.fr

bjutdt¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: bjutdt
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: da5f3fc8f98140e7eaff722b8c873589
Run description: We had a web crawler to get daily news and website information.And the external resource is google.So we can update our query expansion every day and there is no human intervention.

bjutgbdt¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: bjutgbdt
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 0d9c342147aafa06099ce2be6c389c76
Run description: We had a web crawler to get daily news and website information.And the external resource is google.So we can update our query expansion every day and there is no human intervention.

BJUTmydt-04¶

Run ID: BJUTmydt-04
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

BJUTmydt-05¶

Run ID: BJUTmydt-05
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

BJUTmyrf-03¶

Run ID: BJUTmyrf-03
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

bjutrf¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: bjutrf
Participant: BJUT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 893b43cc65443de1e64d2be02fce8b1d
Run description: We had a web crawler to get daily news and website information.And the external resource is google.So we can update our query expansion every day and there is no human intervention.

CCNUNLPrun1¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: CCNUNLPrun1
Participant: CCNU2016NLP
Track: Real-time Summarization
Year: 2016
Submission: 8/11/2016
Task: b
MD5: ca7b8afdc0ea20208e5a4a396f260b7a
Run description: 1. Use Bing search api to extend the query. 2. Use JS-Divergence to calculate the relevance of the summaries and coming tweet. 3.Use TFIDF to calculate the novelty 4.resort summaries in Scenario B at the end of a day by BM25
Code: https://github.com/beichao1314/TREC2016

CCNUNLPrun1-06¶

Run ID: CCNUNLPrun1-06
Participant: CCNU2016NLP
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

CCNUNLPrun2¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: CCNUNLPrun2
Participant: CCNU2016NLP
Track: Real-time Summarization
Year: 2016
Submission: 8/11/2016
Task: b
MD5: c0fb0e503a57efc8126be1d3e49ed77e
Run description: 1.Use Bing search api to extend the query 1.Use JS-Divergence to calculate the relevance of the queries and tweets, summaries and coming tweet. 2.Use TFIDF to calculate the novelty 3.resort summaries in Scenario B at the end of a day by BM25
Code: https://github.com/beichao1314/TREC2016

CCNUNLPrun2-07¶

Run ID: CCNUNLPrun2-07
Participant: CCNU2016NLP
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

CLIP-A-1-08¶

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

Run ID: CLIP-A-1-08
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

CLIP-A-2-09¶

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

Run ID: CLIP-A-2-09
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

CLIP-A-3-10¶

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

Run ID: CLIP-A-3-10
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

CLIP-B-2015¶

Participants | Input | Summary | Appendix

Run ID: CLIP-B-2015
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 00bfeb096c08bd990e252ea1c3df3b2c
Run description: We trained a word2vec model on 1B old tweets.

CLIP-B-MAX¶

Participants | Input | Summary | Appendix

Run ID: CLIP-B-MAX
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: c4923d8bc00fc8bbfec9c4cf980481c2
Run description: We ran a search on Twitter to determine a score threshold for returning candidate tweets. We also trained a word2vec model on 1B old tweets.

CLIP-B-MIN¶

Participants | Input | Summary | Appendix

Run ID: CLIP-B-MIN
Participant: CLIP
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: b537a642ce016b64812a70a2b90f0210
Run description: We ran a search on Twitter to determine a score threshold for returning candidate tweets. We also trained a word2vec model on 1B old tweets.

Hamid-20¶

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

Run ID: Hamid-20
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

HLJIT_LM¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: HLJIT_LM
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: a42eb1d3e6f061b9596c6667982cbf4b
Run description: The language model is used in this run.

HLJIT_LM-19¶

Run ID: HLJIT_LM-19
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

HLJIT_LM_TIME¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: HLJIT_LM_TIME
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: d76bbadb7d562ff4da2ad5bbdfc44a29
Run description: The time-based document model is used to smooth the document.

HLJIT_LM_URL¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: HLJIT_LM_URL
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 01431ad725f3b1fdd88c47dd1a287127
Run description: The web pages linked from tweets are used. The retrieval model is use the model described in "A hyperlink-extended language model for microblog retrieval,International Journal of Database Theory and Application 8.6 (2015) pp.89-100".

iitbhu-15¶

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

Run ID: iitbhu-15
Participant: DPLAB_IITBHU
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

IritIrisSDA-22¶

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

Run ID: IritIrisSDA-22
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

IritIrisSDB¶

Participants | Input | Summary | Appendix

Run ID: IritIrisSDB
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: d892a68e0c1e63f7ed1029a7e98b0cda
Run description: The main aim of this run was the speed of the decision making : all answers are taken within few minutes. Many features, in addition of the content of the tweet, are taken into account to make a first score on the text content of the tweet, and a second one based on all the additionnal features. Python had been prefered to Java to increase the rapidity of the decision making even more, exept for the final ranking of the result (back to Java as the sampling).

iritRunBiAm-21¶

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

Run ID: iritRunBiAm-21
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

IRLAB¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: IRLAB
Participant: IRLAB_DA-IICT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: b
MD5: 7c3bc5cc567a062bc7dc3f99c0ffd946
Run description: Query Expansion has been done using word2vec model which is trained using tweets of past few months. Relevance score has been obtained using Okapi BM25. Novelty detection done using cosine similarity.

IRLAB2¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: IRLAB2
Participant: IRLAB_DA-IICT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: b
MD5: be8356a8175294c8a1cd8a8072c5713c
Run description: Query expansion has been done using word2vec model which is trained using tweets of past few months. Relevance score has been obtained using Okapi BM25. Novelty detection done using Jaccard similarity.

isikol_tag¶

Participants | Input | Summary | Appendix

Run ID: isikol_tag
Participant: ISIKol
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 49461a11ecf525ce33b1184d632f672c
Run description: This run uses parts of speech and named entity tagger to find key words from each query.

isikol_ti¶

Participants | Input | Summary | Appendix

Run ID: isikol_ti
Participant: ISIKol
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 49f41b3523e20c80f35f28a6cc59ed8f
Run description: This run uses tf-idf metric to find key words from each query.

MyBaseline-17¶

Run ID: MyBaseline-17
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

MyBaseline-18¶

Run ID: MyBaseline-18
Participant: HLJIT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

MyBaseline-24¶

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

Run ID: MyBaseline-24
Participant: ISIKol
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

nudt_biront¶

Participants | Input | Summary | Appendix

Run ID: nudt_biront
Participant: NUDTSNA
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: b3251ac68edbaab60cc654afde9ef427
Run description: We apply a recommendation framework based on Lucene. External terms extracted from Google search engine and Twitter using pro le title as query are well incorporated to enhance the understanding of a pro le's interest. We set each term's weight as its tf-idf value based on search result returned by Google and Twitter, then combine pro le title and external terms, we build our Lucene query and search hits satis ed our limit from twitter stream indexed. In the email digest task, based on the candidate tweets retrieved from the rst task, we select top-100 ranked twitter of each pro le per day as our result.

nudt_sna¶

Participants | Input | Summary | Appendix

Run ID: nudt_sna
Participant: NUDTSNA
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 6433a1ce0aec7b4ce7cde3590dc96346
Run description: Based on the search engine library Lucene, we implement a system for 2016 RTS track. With the Google search engine and the external corpora Wikipedia, we just use the title field of a profile for query expansion. Besides, we also explore the Twitter search engine to obtain useful hashtags for some profiles. Last but not least, we overwrite the similarity scoring function to improve the retrieval performance.

nudt_sna-28¶

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

Run ID: nudt_sna-28
Participant: NUDTSNA
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

nudt_sna-29¶

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

Run ID: nudt_sna-29
Participant: NUDTSNA
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

nudt_sna-30¶

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

Run ID: nudt_sna-30
Participant: NUDTSNA
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

PKUICSTRunB1¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PKUICSTRunB1
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 0cc7eedafed9a665038c55bb5aa08934
Run description: We use KL divergence with Jaccard smooth to obtain relevant tweets, and utilized a uniform novel threshold N = 0.73.
Code: https://github.com/yaolili/trec16/tree/master/src

PKUICSTRunB2¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PKUICSTRunB2
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 4d3f491107fbc39c740612e3ed804ac8
Run description: We utilized google web search to realize query expansion before the evaluation period, and utilized a uniform novel threshold N = 0.72.
Code: https://github.com/yaolili/trec16/tree/master/src

PKUICSTRunB3¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PKUICSTRunB3
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 0f875850ed578baa4fc42aaa145e494c
Run description: We utilized google web search to realize query expansion before the evaluation period, and adopted simhash to calculate novelty with a uniform novel threshold D = 42.
Code: https://github.com/yaolili/trec16/tree/master/src

PolyURunB1¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PolyURunB1
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/11/2016
Task: b
MD5: 82d32944a35db5e2d18532cd28496f8c
Run description: most salient features: add a word2vec information redundancy feature to train the SVM model external resources: use GoogleNews Corpus and Reuters Corpus to train the model. They are not timely with respect to the queries.

PolyURunB2¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PolyURunB2
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/11/2016
Task: b
MD5: ee5d3686bb9cd1c1d43193f75ea94736
Run description: the most salient features: use features to train SVM model (without word2vec information redundancy feature) external resources: Reuters Corpus

PolyURunB3¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PolyURunB3
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/11/2016
Task: b
MD5: 2a2d60ac6df5f42680c71672ff5ffaca
Run description: the most salient features: no ML model, just a very naive method external resources: Reuters Corpus

PRNABaseline-34¶

Run ID: PRNABaseline-34
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

PRNATaskA2-35¶

Run ID: PRNATaskA2-35
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

PRNATaskA3-36¶

Run ID: PRNATaskA3-36
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

PRNATaskB1¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PRNATaskB1
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 760f7460c6e9a9de9c029589561b139c
Run description: This run uses text from title, description, narrative as features for search, with dynamic push strategy.

PRNATaskB2¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PRNATaskB2
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 109973e451f6f8657e29213626731fdd
Run description: This run uses text from title, description, narrative as features for search, with different weighting on different types of search terms.

PRNATaskB3¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: PRNATaskB3
Participant: prna
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 0107b1cc2b494876af3a6f8b26cad538
Run description: This run combines the two algorithms from run1 and run2 with our algorithm for Scenario A. It uses text features from topic title, description and narrative as features for search with different weighting on features and push thresholds.

QUBaseline-37¶

Run ID: QUBaseline-37
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

QUDR8¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: QUDR8
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 5db59fdfc1bddf33562103edda5b450f
Run description: This run uses the title as a query and runs it at the end of each day against the index of tweets. Before evaluation period, we indexed tweets of 5 days to get initial statistics. It uses language modeling based on the Dirichlet smoothing method for computing the similarity between tweets and the query with a relevance threshold = 8 and novelty threshold of 0.6.

QUExpP-38¶

Run ID: QUExpP-38
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

QUExpT-39¶

Run ID: QUExpT-39
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

QUJM16¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: QUJM16
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 1f105546efad4383040285d947a90baa
Run description: This run uses the title as a query and runs it at the end of each day against the index of tweets. Before evaluation period, we indexed tweets of 5 days to get initial statistics. It uses language modeling based on the Jelinek-Mercer smoothing method for computing the similarity between tweets and the query with a relevance threshold = 16 and novelty threshold of 0.6.

QUJMDR24¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: QUJMDR24
Participant: QU
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: f7f76242b246047cd6e5d4c5d378ff61
Run description: This run uses the title as a query and runs it at the end of each day against the index of tweets. Before evaluation period, we indexed tweets of 5 days to get initial statistics. It uses language modeling based on the Jelinek-Mercer and Dirichlet smoothing methods for computing the similarity between tweets and the query with a relevance threshold of 24 and novelty threshold of 0.6.

QUT_RTS-40¶

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

Run ID: QUT_RTS-40
Participant: QUT_RTS
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

ru32-33¶

Run ID: ru32-33
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

run1-11¶

Run ID: run1-11
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

run1-31¶

Run ID: run1-31
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

run2-12¶

Run ID: run2-12
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

run2-32¶

Run ID: run2-32
Participant: PKUICST
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

run3-13¶

Run ID: run3-13
Participant: COMP2016
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

runA_daiict_irlab-23¶

Run ID: runA_daiict_irlab-23
Participant: IRLAB_DA-IICT
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

RunBIch¶

Participants | Input | Summary | Appendix

Run ID: RunBIch
Participant: IRIT
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: c5443b812aae9b3aa0a7adac56c70a82
Run description: In this run, we used some basic features to compute the relevance of tweets wrt the given profiles.

udelRunBM25-43¶

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

Run ID: udelRunBM25-43
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

udelRunBM25B¶

Participants | Input | Summary | Appendix

Run ID: udelRunBM25B
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Task: b
MD5: a40b8b3960f6ff831822eb4960bfbc1a
Run description: Trains classifier using top Google documents, tweets from Twitter and uses BM25 similarity.

udelRunTFIDF-44¶

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

Run ID: udelRunTFIDF-44
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

udelRunTFIDFB¶

Participants | Input | Summary | Appendix

Run ID: udelRunTFIDFB
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: c5eef88586c73ec5c59dfc0e6e3833f5
Run description: Trains classifier using top Google documents, tweets from Twitter.

udelRunTFIDFQ-45¶

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

Run ID: udelRunTFIDFQ-45
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Task: a

udelRunTFIDFQB¶

Participants | Input | Summary | Appendix

Run ID: udelRunTFIDFQB
Participant: udel
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Task: b
MD5: 463ad70f7cdb00d0f3213bbdefeaa7ec
Run description: Trains classifier using top Google documents, tweets from Twitter and relevant tweets from microblog 2015 qrels file.

UDInfo_TlmN¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: UDInfo_TlmN
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: e6ea105b85b38798212a66269d379832
Run description: Use search result snippets of Bing, Google, and Yahoo to expand the original queries. Query prediction techniques, such as clarity, as well as the estimated relevant language model differences were used to predict the score threshold for each day. Simple redundancy check using set difference between tweets.

UDInfo_TlmNlm¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: UDInfo_TlmNlm
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: d4203b144b495aae8ec9b4c16e7684bb
Run description: Use search result snippets of Bing, Google, and Yahoo to expand the original queries. Query prediction techniques, such as clarity, as well as the estimated relevant language model differences were used to predict the score threshold for each day and determining the redundancy threshold for each day.

UDInfo_TN¶

Participants | Proceedings | Input | Summary | Appendix

Run ID: UDInfo_TN
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 1ac4f6b03fb26a2dcebbdb69d428a673
Run description: Use search result snippets of Bing, Google, and Yahoo to expand the original queries. Query prediction techniques, such as clarity, was used to predict the score threshold for each day

UDInfoDFP-47¶

Run ID: UDInfoDFP-47
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

UDInfoSFP-46¶

Run ID: UDInfoSFP-46
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

UDInfoSPP-48¶

Run ID: UDInfoSPP-48
Participant: udel_fang
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

UmdHcilBaseline-49¶

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

Run ID: UmdHcilBaseline-49
Participant: umd_hcil
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

WaterlooBaseline-50¶

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

Run ID: WaterlooBaseline-50
Participant: WaterlooClarke
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

WaterlooBaseline-51¶

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

Run ID: WaterlooBaseline-51
Participant: WaterlooLin
Track: Real-time Summarization
Year: 2016
Submission: 8/13/2016
Type: automatic
Task: a

YoGoslingBSL¶

Participants | Input | Summary | Appendix

Run ID: YoGoslingBSL
Participant: WaterlooLin
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: de0ac209a6e243498c3df40c7c8a50d8
Run description: This run served as the track-wide baseline, which timely searches the index using only title term match scoring.
Code: https://github.com/YoGosling/Anserini

YoGoslingLMGTFY¶

Participants | Input | Summary | Appendix

Run ID: YoGoslingLMGTFY
Participant: WaterlooLin
Track: Real-time Summarization
Year: 2016
Submission: 8/12/2016
Type: automatic
Task: b
MD5: 9e9e29ade5bbcecc0938ac938409c9c7
Run description: Our method, based on the track wide baseline YoGosling, additionally queries the Google custom search API every day at the beginning, using the "title" term. We compute the KL-divergence statistics to generate a list of "relevant" phrase and perform phrase search with slopping distance tolerance.