Runs - Novelty 2002¶

CIIR02tfkl¶

Run ID: CIIR02tfkl
Participant: umass
Track: Novelty
Year: 2002
Submission: 8/31/2002
Run description: We use a tf*idf weighting scheme to identify relevant sentences and KL divergence to identify novel sentences.

CIIR02tfnew¶

Run ID: CIIR02tfnew
Participant: umass
Track: Novelty
Year: 2002
Submission: 8/31/2002
Run description: We use a tf*idf weighting scheme to identify relevant sentences and a count of new words to identify novel sentences.

cmu02t300rAs¶

Run ID: cmu02t300rAs
Participant: cmu_lti
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Run cmu02t300rAs uses TFIDF weighting with pseudo-relevance feedback to find highly relevant sentences, using a single threshold rule, followed by a very simple translation model based on shallow parsing and Wordnet to remove redundant sentences (across all documents).

cmu02t300rBw¶

Run ID: cmu02t300rBw
Participant: cmu_lti
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Run cmu02t300rBw uses TFIDF weighting with pseudo-relevance feedback to find relevant sentences, paired with a simple discriminative classifier based on surface and semantic features, followed by a very simple translation model based on shallow parsing and Wordnet to remove redundant sentences.

cmu02t300rCb¶

Run ID: cmu02t300rCb
Participant: cmu_lti
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Run cmu02t300rCb is a baseline, using TFIDF weighting with pseudo-relevance feedback to find highly relevant sentences, and performing a trivial novelty step that accepts all sentences.

cmu02t300rCv¶

Run ID: cmu02t300rCv
Participant: cmu_lti
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Run cmu02t300rCv uses TFIDF weighting with pseudo-relevance feedback to find highly relevant sentences, then includes neighbor sentences based on additional criteria, followed by a very simple translation model based on shallow parsing and Wordnet to remove redundant sentences.

cmu02t300rCw¶

Run ID: cmu02t300rCw
Participant: cmu_lti
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Run cmu02t300rCw uses TFIDF weighting with pseudo-relevance feedback to find highly relevant sentences, then includes neighbor sentences based on additional criteria, followed by a simple translation model with shallow parsing and Wordnet to remove redundant sentences (lower threshold, within docs only).

dumbrun¶

Run ID: dumbrun
Participant: irit
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: for relevant sentences, we use topic terms weighting and match them against every sentence. For new sentences we compute individual sentence-to-sentence similarity and global sentences-to-sentence similarity.

fdut11n1¶

Results | Participants | Input | Summary | Appendix

Run ID: fdut11n1
Participant: Fudan
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: hmm based relevance decision

fdut11n2¶

Results | Participants | Input | Summary | Appendix

Run ID: fdut11n2
Participant: Fudan
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: hmm based relevance decision

fdut11n3¶

Results | Participants | Input | Summary | Appendix

Run ID: fdut11n3
Participant: Fudan
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: simple method

novcolcl35¶

Run ID: novcolcl35
Participant: columbia_novelty
Track: Novelty
Year: 2002
Submission: 9/2/2002
Run description: Sentences (or in some runs smaller units) were identified by the existence of topic concepts, which are similar to expanded queried built with the help of WordNet and co-occurrences in a background corpus, and in addition, an attempt was made to segement the articles into sections of higher concentration of topic concepts, and to determine the kind of document set and set the system parameters accordingly, while novelty was decided by the degree of overlap of these topic concepts.

novcolcl85¶

Run ID: novcolcl85
Participant: columbia_novelty
Track: Novelty
Year: 2002
Submission: 9/2/2002
Run description: Sentences (or in some runs smaller units) were identified by the existence of topic concepts, which are similar to expanded queried built with the help of WordNet and co-occurrences in a background corpus, and in addition, an attempt was made to segement the articles

novcolclfx¶

Run ID: novcolclfx
Participant: columbia_novelty
Track: Novelty
Year: 2002
Submission: 9/2/2002
Run description: Sentences (or in some runs smaller units) were identified by the existence of topic concepts, which are similar to expanded queried built with the help of WordNet and co-occurrences in a background corpus, and in addition, an attempt was made to segement the art

novcolmerg¶

Run ID: novcolmerg
Participant: columbia_novelty
Track: Novelty
Year: 2002
Submission: 9/2/2002
Run description: Sentences (or in some runs smaller units) were identified by the existence of topic concepts, which are similar to expanded queried built with the help of WordNet and co-occurrences in a background corpus, and in addition, an attempt was made to segement the art

novcolsent¶

Run ID: novcolsent
Participant: columbia_novelty
Track: Novelty
Year: 2002
Submission: 9/2/2002
Run description: Sentences (or in some runs smaller units) were identified by the existence of topic concepts, which are similar to expanded queried built with the help of WordNet and co-occurrences in a background corpus, and in addition, an attempt was made to segement the

nttcslabnvp¶

Run ID: nttcslabnvp
Participant: nttcom_kazawa
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Machine learning (SVM)

nttcslabnvr2¶

Run ID: nttcslabnvr2
Participant: nttcom_kazawa
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Machine learining with SVM (RBF kernels)

ntu1¶

Run ID: ntu1
Participant: NTU
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: use sliding windows to exact relevant sentence, and do the text segmentation, and compare similarity between each sentence and each segment.

ntu2¶

Run ID: ntu2
Participant: NTU
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: use sliding windows to exact relevant sentence, and do the text segmentation, and compare similarity between each sentence and each segment.

ntu3¶

Run ID: ntu3
Participant: NTU
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: use sliding windows to exact relevant sentence, and do the text segmentation, and compare similarity between each sentence and each segment.

pircs2N01¶

Run ID: pircs2N01
Participant: cuny
Track: Novelty
Year: 2002
Submission: 9/1/2002
Run description: Sentence retrieved by pircs using trec8 dictionary, relevent senctence threshold by rsv >=1.25, novelty threshold by |sent1 ^ sent2|/|sent1 V sent2| < .3

pircs2N02¶

Run ID: pircs2N02
Participant: cuny
Track: Novelty
Year: 2002
Submission: 9/1/2002
Run description: Sentence retrieved by pircs using trec8 dictionary, relevant senctence threshold by rsv >=1.25, novelty threshold by |sent1 ^ sent2|/|sent1 V sent2| <= .5

pircs2N03¶

Run ID: pircs2N03
Participant: cuny
Track: Novelty
Year: 2002
Submission: 9/1/2002
Run description: Sentence retrieved by pircs using trec8 dictionary, relevant senctence threshold by rsv >=1.5, novelty threshold by |sent1 ^ sent2|/|sent1 V sent2| <= .3

pircs2N04¶

Run ID: pircs2N04
Participant: cuny
Track: Novelty
Year: 2002
Submission: 9/1/2002
Run description: Sentence retrieved by pircs using trec8 dictionary, relevant sentence threshold by rsv >=1.5, novelty threshold by |sent1 ^ sent2|/|sent1 V sent2| <= .5

pircs2N05¶

Run ID: pircs2N05
Participant: cuny
Track: Novelty
Year: 2002
Submission: 9/1/2002
Run description: Sentence retrieved by pircs using trec8 dictionary, relevant sentence threshold by rsv >=1.5, novelty threshold by |sent1 ^ sent2|/|sent1 V sent2| <= .3

ss1¶

Run ID: ss1
Participant: streamsage
Track: Novelty
Year: 2002
Submission: 8/27/2002
Run description: Query expansion title terms plus multi-word units from news story corpus, and noun phrases from desc.

thunv1¶

Run ID: thunv1
Participant: tsinghua
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Query expansion based on local term co-occurrence and document expansion based on wordNet, with topic classification according to documents.

thunv2¶

Run ID: thunv2
Participant: tsinghua
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Query expansion based on local term co-occurrence and document expansion based on wordNet, with topic classification according to similarities between topic fields.

thunv3¶

Run ID: thunv3
Participant: tsinghua
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Query expansion based on local term co-occurrence and document expansion based on wordNet, and results combination.

thunv4¶

Run ID: thunv4
Participant: tsinghua
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Query expansion based on local term co-occurrence and Lin Dekang's dependency word similarity DB.

thunv5¶

Run ID: thunv5
Participant: tsinghua
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Dcument expansion based on coordinate terms to eliminate duplicated information with probabilistic based system.

UAmsT11ntcom¶

Run ID: UAmsT11ntcom
Participant: uva
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Retrieval part based on a combined run; simple frequency-based method for determining novelty

UAmsT11ntlem¶

Run ID: UAmsT11ntlem
Participant: uva
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: TreeTagger-based lemmatization; simple frequency-based method for determining novelty

UAmsT11ntste¶

Run ID: UAmsT11ntste
Participant: uva
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: Porter stemming; simple frequency-based method for determining novelty

UIowa02Nov4¶

Results | Participants | Input | Summary | Appendix

Run ID: UIowa02Nov4
Participant: uiowa
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: new named entities and nounphrases, threshold tuned for balance of recall and precision

UIowa02Nov5¶

Results | Participants | Input | Summary | Appendix

Run ID: UIowa02Nov5
Participant: uiowa
Track: Novelty
Year: 2002
Submission: 9/3/2002
Run description: new named entities and nounphrases, threshold tuned for precision

umich1¶

Run ID: umich1
Participant: umich
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: Adapt the MEAD summarizer to identify novel sentences.

UMIch2¶

Run ID: UMIch2
Participant: umich
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: Adapt the MEAD summarizer to identify novel sentences.

UMich3¶

Run ID: UMich3
Participant: umich
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: Adapt the MEAD summarizer to identify novel sentences.

UMICH4¶

Run ID: UMICH4
Participant: umich
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: Adapt the MEAD summarizer to identify novel sentences.

UMich5¶

Run ID: UMich5
Participant: umich
Track: Novelty
Year: 2002
Submission: 9/4/2002
Run description: Adapt the MEAD summarizer to identify novel sentences.