Runs - Entity 2010¶

B64¶

Results | Participants | Proceedings | Input | Summary

Run ID: B64
Participant: HPI
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 0daeed79f5f26944ba3468d5c0ec3ce5
Run description: - The narrative query is pre-processed using the Stanford Part-of-Speech-Tagger - The source entity is queried to Freebase and the most popular synonyms are retrieved - A key word query is constructed consisting source entity, alternative names and verbs and nouns from the narrative (plus additional rewriting wrt. advanced search engines features) - Candidate documents are retrieved from a commercial web search engine (Life Web: BING, we took in this run 64 documents) - Candidate target entities are scored as described above - For home page finding, we collect advanced features of the web search engine (such as "find related documents" or "homepage URL occurs in anchors of other Web-pages") - As set of features is applied to score potential homepages, the features are aggregated using weights which are trained using a genetic algorithm - Finally the homepage URLs are mapped to Clueweb-IDs

bitDSHPRun¶

Results | Participants | Proceedings | Input | Summary

Run ID: bitDSHPRun
Participant: BIT
Track: Entity
Year: 2010
Submission: 10/2/2010
Type: automatic
Task: main
MD5: e7f27d6f1a41aea6f1f76bdef022d643
Run description: Reconstruct logical sitemap of the source entity's site and locate the most relevant pages to extract related entity with the target type.And employ different sources such as inlink,google,realname and different homepage features which differ person and product types from that of organization and location.

bitDSRRun¶

Results | Participants | Proceedings | Input | Summary

Run ID: bitDSRRun
Participant: BIT
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 95cc59b11ee6d94f19be3f7bdb939358
Run description: To reconstruct logical sitemap of the target entities and locate the most relevant pages to extract list or table entities. Search Google and realnames with entity names to help homepage finding.

bitRFRun¶

Results | Participants | Proceedings | Input | Summary

Run ID: bitRFRun
Participant: BIT
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: manual
Task: main
MD5: 3f4af220f770c4503edbc6326566acb9
Run description: To reconstruct logical sitemap of the target entities and locate the most relevant pages to extract list or table entities. Search Google and realnames with entity names to help homepage finding.

CARDENSMBLE¶

Results | Participants | Input | Summary

Run ID: CARDENSMBLE
Participant: CARD_UALR
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: 1fd7f12b5b6fb10f2ebff92c264400f2
Run description: entity tagging, indexing, entity entity graph, Ensemble approach, homepage finding

CARDFPR¶

Results | Participants | Input | Summary

Run ID: CARDFPR
Participant: CARD_UALR
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: c8d975133d9f29d254ff995cb285f78d
Run description: entity tagging, indexing, entity entity graph, pagerank approach, homepage finding

CARDHITS¶

Results | Participants | Input | Summary

Run ID: CARDHITS
Participant: CARD_UALR
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: 5c8606cfd7d256c8d6ba1c80172e0224
Run description: entity tagging, indexing, document entity graph, HITS approach, homepage finding

CARDSGFCS¶

Results | Participants | Input | Summary

Run ID: CARDSGFCS
Participant: CARD_UALR
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: f8aa85849447f66f749db9b87e1a5284
Run description: entity tagging, indexing, entity entity graph, cumulative similarity approach, homepage finding

Comp¶

Results | Participants | Proceedings | Input | Summary

Run ID: Comp
Participant: LIA_UAPV
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: d4f842b97a412b2bfd6fd3c0bdd335ba
Run description: For all ours runs, we start by downloading one hundred web pages with Yahoo! web search engine. We divide this web pages in passages of one or three sentences. Those passages are indexed by Indri and we keep the five hundred top ranked passages. Named entities are found by means of Stanford NER. We rank them by employing our own scoring function and search homepages on the web from new specific queries . Lastly, we map the url to the corresponding (if exists) clueweb id. For this baseline run, we used a density measure of question words around a candidate named entity which is called compacity.

Div¶

Results | Participants | Proceedings | Input | Summary

Run ID: Div
Participant: LIA_UAPV
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 5b0cc9cd07aee35c0b3a782fae1a094b
Run description: To filter the named entities with a little more accuracy than just the type (person, place, ...) we try to extract from the topic a more specific type expected (teammate, champion sports, ...) with grammatical rules . We then try to determine the lexical context of this type and whether the candidate named entities share it. For this we calculate here the probability distribution of n-grams in a corpus of web page on the subject. A first corpus is recovered on the internet searching for the type (eg, subjecting the word "teammate" in a search engine and recovered to 100 pages) which will be calculated the probability of occurrence of n-grams. For each candidate named entity such procedure will be performed (we shall then submit the text of the named entity in the search engine). Our assumption here is that the higher the probability distribution of an candidate named entity is close to the one of the expected type, so there are more chances of it being such. This distance is calculated using the Kullback-Leibler divergence. More important it is for a named entity, more it will be penalized in its rankings.

EntityHP¶

Results | Participants | Proceedings | Input | Summary

Run ID: EntityHP
Participant: PITTSIS
Track: Entity
Year: 2010
Submission: 9/29/2010
Type: manual
Task: main
MD5: 49637bec30286efeef58d8dc201dae9d
Run description: test

EntityHP1¶

Results | Participants | Proceedings | Input | Summary

Run ID: EntityHP1
Participant: PITTSIS
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 4f425699a158276f3640565b59606b2e
Run description: With groundtruth entity & Homepage Finding

FduWimET1¶

Results | Participants | Proceedings | Input | Summary

Run ID: FduWimET1
Participant: FDWIM2010
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: automatic
Task: main
MD5: e4893f6d508e2a516ce00247088f464f
Run description: Entity ranking according to multiple keywords Deep mining of authority pages Feature-based algorithm in entity homepage detection External resource: Google

FduWimET2¶

Results | Participants | Proceedings | Input | Summary

Run ID: FduWimET2
Participant: FDWIM2010
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: automatic
Task: main
MD5: b187eb45dd0714a505c1b2393a5b92cd
Run description: Combine corpus-based association rules and search engine in the entities Entity ranking according to multiple keywords Deep mining of authority pages Feature-based algorithm in entity homepage detection External resource: Google

FduWimET3¶

Results | Participants | Proceedings | Input | Summary

Run ID: FduWimET3
Participant: FDWIM2010
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: manual
Task: main
MD5: a0f382cc1cdd4f788f65d6703e78d190
Run description: Combine corpus-based association rules and search engine in the entities Entity ranking according to multiple keywords Deep mining of authority pages Feature-based algorithm in entity homepage detection Construct the queries for document retrieval manually External resource: Google

FduWimET4¶

Results | Participants | Proceedings | Input | Summary

Run ID: FduWimET4
Participant: FDWIM2010
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: automatic
Task: main
MD5: a21bb131827ec612c0273d6270bbc1bb
Run description: Combine corpus-based association rules and search engine in the entities Entity ranking according to multiple keywords Deep mining of authority pages Feature-based algorithm in entity homepage detection External resource: Google

G16¶

Results | Participants | Proceedings | Input | Summary

Run ID: G16
Participant: HPI
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 37e0916005611a4c2812dbf681d3c826
Run description: - The narrative query is pre-processed using the Stanford Part-of-Speech-Tagger - The source entity is queried to Freebase and the most popular synonyms are retrieved - A key word query is constructed consisting source entity, alternative names and verbs and nouns from the narrative (plus additional rewriting wrt. advanced search engines features) - Candidate documents are retrieved from a commercial web search engine (Life Web: Google, we took in this run 16 documents) - Candidate target entities are scored as described above - For home page finding, we collect advanced features of the web search engine (such as "find related documents" or "homepage URL occurs in anchors of other Web-pages") - As set of features is applied to score potential homepages, the features are aggregated using weights which are trained using a genetic algorithm - Finally the homepage URLs are mapped to Clueweb-IDs

G64¶

Results | Participants | Proceedings | Input | Summary

Run ID: G64
Participant: HPI
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 032b36b2fdb75fe472c2758c3dac36f6
Run description: - The narrative query is pre-processed using the Stanford Part-of-Speech-Tagger - The source entity is queried to Freebase and the most popular synonyms are retrieved - A key word query is constructed consisting source entity, alternative names and verbs and nouns from the narrative (plus additional rewriting wrt. advanced search engines features) - Candidate documents are retrieved from a commercial web search engine (Life Web: Google, we took in this run 64 documents) - Candidate target entities are scored as described above - For home page finding, we collect advanced features of the web search engine (such as "find related documents" or "homepage URL occurs in anchors of other Web-pages") - As set of features is applied to score potential homepages, the features are aggregated using weights which are trained using a genetic algorithm - Finally the homepage URLs are mapped to Clueweb-IDs

ICTNETRun1¶

Results | Participants | Proceedings | Input | Summary

Run ID: ICTNETRun1
Participant: ICTNET
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: 0d94b2bd88e28144ff9e4ada4372f2ea

Run description: Procedure: 1. Extend the keyword set with WordNet 2. Search for the relevant documents in the dataSetA. 3. Extract the target named entities using the stanford NER tools 4. Extract the table list which may include some target named entities. Here we use html tag (such as or

) and some heuristic methods to recognize the tables which we want 5. compute the initial ranking value of the candidae entities using some classic models such as language model 6. Reinforce the ranking value of the candidate entities with the help of table list. 7. Search for the homepages of candidate entities using google External Resources: 1. Extend the keyword set with WordNet. 2. Get the urls of homepage of candidate entity using google

ilpsA500¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsA500
Participant: UAms
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: e0ade9666cf43cf895a5281ae60e4da7
Run description: freebase, dbpedia, stanford tagger, bing

ilpsM50¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsM50
Participant: UAms
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 793be828d1dde0513be65fb90225f4b5
Run description: stanford tagger, freebase, dbpedia, bing

ilpsM50agfil¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsM50agfil
Participant: UAms
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: manual
Task: main
MD5: 3ad7c8e2fc9e61b17358bfda9b1da824
Run description: stanford tagger, freebase, bing ,dbpedia

ilpsM50var¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsM50var
Participant: UAms
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: f35a0817192c96400248cece4ddd7291
Run description: stanford tagger, dbpedia, freebase, bing

ilpsSetOL¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsSetOL
Participant: UAms
Track: Entity
Year: 2010
Submission: 10/19/2010
Type: automatic
Task: elc
MD5: 8872d701d63868392d6c2ec879999d37
Run description: Candidate entities are ranked according to set overlap with the predicates and objects linked to by the example entities. Candidates are entities that share common objects with example entities.

ilpsSetOLnar¶

Results | Participants | Proceedings | Input | Summary

Run ID: ilpsSetOLnar
Participant: UAms
Track: Entity
Year: 2010
Submission: 10/20/2010
Type: automatic
Task: elc
MD5: 1ec522bec8e42ccb955b1547e2dec4df
Run description: filtered with category, source entity and narrative

KMR1PU¶

Results | Participants | Proceedings | Input | Summary

Run ID: KMR1PU
Participant: Purdue_IR
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 66aa6625adfb1b20ccb8dcf842c5720e
Run description: We have special treatment on table and list data. We use Google to retrieve a set of candidate homepages and train classifiers to identify the homepages.

KMR3PU¶

Results | Participants | Proceedings | Input | Summary

Run ID: KMR3PU
Participant: Purdue_IR
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 1e25ec6435dbe33250c2a296818d7cdc
Run description: We utilize the structures of tables and lists to find the related entities. We use Google to retrieve a set of candidate answer pages as well as candidate homepages. We train logistic regression models on the four types of entities respectively.

KMR5PU¶

Results | Participants | Proceedings | Input | Summary

Run ID: KMR5PU
Participant: Purdue_IR
Track: Entity
Year: 2010
Submission: 10/17/2010
Type: automatic
Task: elc
MD5: f8fb240f52eff9843c447af065536a42
Run description: We build entity profiles only from BTC-2009 and index them by Indri. The retrieved entities are filtered based on the target entity types inferred from queries.

LearnDPI¶

Results | Participants | Proceedings | Input | Summary

Run ID: LearnDPI
Participant: LIA_UAPV
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 22920c2a4b63b4b06965fa54384caf74
Run description: In this final run, we try to use all the measures we have (divergence, "compacit", passages score, idf, ...) to determine the rank of a named entity. For this we determine a class "Yes" which that means the named entity is correct, answering to the request and a class "No". For train, we build a corpora from the lists questions from TREC QA 2007 and 2006. Lastly we rank candidates answers according to their belonging to the "Yes" class.

LiraSealClwb¶

Results | Participants | Proceedings | Input | Summary

Run ID: LiraSealClwb
Participant: CMU_LIRA
Track: Entity
Year: 2010
Submission: 10/17/2010
Type: automatic
Task: elc
MD5: 804b4d8e3c9ae0aeff518903e16f21bb
Run description: 1. I used a relation and list extraction technique called SEAL (Set Expander for Any Language) 2. I used the Clueweb index built in CMU : http://boston.lti.cs.cmu.edu:8085/clueweb09/search/cata_english/lemur.cgi 3. I also used the billion tripple index built by Krisztian Balog http://zookst1.science.uva.nl:8888/btc-webapp-0.1/btc2009

LiraSealgoog¶

Results | Participants | Proceedings | Input | Summary

Run ID: LiraSealgoog
Participant: CMU_LIRA
Track: Entity
Year: 2010
Submission: 10/17/2010
Type: automatic
Task: elc
MD5: ffc4d7141a5615995e892c1f33db8c39
Run description: 1. I used a relation and list extraction technique called SEAL (Set Expander for Any Language) 2. I used Google search engine to query the web. 3. I also used the billion tripple index built by Krisztian Balog http://zookst1.science.uva.nl:8888/btc-webapp-0.1/btc2009

PRIS1¶

Results | Participants | Proceedings | Input | Summary

Run ID: PRIS1
Participant: PRIS
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: manual
Task: main
MD5: eba08c0e1fb21e49710e4a65a25a6bc2
Run description: Document-Centered Model by Indri, giving priority to entities extracted manually

PRIS2¶

Results | Participants | Proceedings | Input | Summary

Run ID: PRIS2
Participant: PRIS
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: manual
Task: main
MD5: 9d13add7ff457e6dcc9b1b2e6341808b
Run description: Document-Centered Model by Indri

PRIS3¶

Results | Participants | Proceedings | Input | Summary

Run ID: PRIS3
Participant: PRIS
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: manual
Task: main
MD5: 46123f7748b3e9f2d3f623052554daf7
Run description: Entity-Centered Model by Indri

PRIS4¶

Results | Participants | Proceedings | Input | Summary

Run ID: PRIS4
Participant: PRIS
Track: Entity
Year: 2010
Submission: 9/28/2010
Type: manual
Task: main
MD5: 1f462727d944c1317b91386827df9b0f
Run description: Entity-Centered Model by Lemur (Okapi)

RanksDivComp¶

Results | Participants | Proceedings | Input | Summary

Run ID: RanksDivComp
Participant: LIA_UAPV
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 399657cf84da905a8bd8c8e6d5c6540d
Run description: In this run we merge the measure of divergence with the compacity for a given named entity by the f-measure of it ranks for each measures.

SIEL10RUN1¶

Results | Participants | Input | Summary

Run ID: SIEL10RUN1
Participant: siel10
Track: Entity
Year: 2010
Submission: 10/3/2010
Type: manual
Task: main
MD5: 3f9acc5e8af3d9083a3728ee338ff5c2
Run description: We had submitted 2 runs already, under the organization SIEL_IIITH, but it was a misunderstanding as it was an account of the previous year. So, we now submit these results with this year's (siel10) account. The features of this run are : 1) Used wikipedia data to find required entities appearing in certain patterns like lists, tables etc. 2) Used homepages of source entities to crawl them upto a certain depth (since we do not have the clueweb category A dataset) and find the required entites and if possible, links to their homepages, which appear in specific patterns. 3) Combine the results and find their homepages by searching in urls present in the url-docid dataset. 4) Use the clueweb url-docid dataset to retrieve the clueweb id's of each url, and present them.

SIELRUN1¶

Results | Participants | Input | Summary

Run ID: SIELRUN1
Participant: SIEL_IIITH
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 194b3a307af277f012d744df3dfa8385
Run description: The approach followed is: 1. Used wikipedia dump for finding results by matching the words in given query. 2. Using homepages given in a query, we crawled a domain (since we do not have the actual dataset - Category A) at a particular depth to find required pages by matching output entity types and the words in the query. 3. combine the result of 1 and 2 to find common entities 4. We find the homepages of required entites using the clueweb url-docid dataset.

SIELRUN2¶

Results | Participants | Input | Summary

Run ID: SIELRUN2
Participant: SIEL_IIITH
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 327ca02671734a370d52d65eca3f6ffc
Run description: Same as run 1 - changed the ranks

Supp¶

Results | Participants | Proceedings | Input | Summary

Run ID: Supp
Participant: NiCT
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 28626bbcf27a40f93533c4d9f902076d
Run description: Use MINPAR to parse queries and supporting sentences. Use YAGO BOSS for retrieval.

SuppHome¶

Results | Participants | Proceedings | Input | Summary

Run ID: SuppHome
Participant: NiCT
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 292298ddddffebe19f44cfc942bd1150
Run description: Use MINPAR to parse queries and supporting sentences. Use Homepage for similarity calculation. Use YAGO BOSS for retrieval.

SuppHomeIsA¶

Results | Participants | Proceedings | Input | Summary

Run ID: SuppHomeIsA
Participant: NiCT
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 412daf64f076404059c6483240f892cd
Run description: Use YAGO data to learn patterns for isA relations. Use MINPAR to parse queries and supporting sentences. Use Homepage for similarity calculation. Use YAGO BOSS for retrieval.

SuppIsA¶

Results | Participants | Proceedings | Input | Summary

Run ID: SuppIsA
Participant: NiCT
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 7595f4583d7d9cbf59ad84c5a6cf6060
Run description: Use YAGO data to learn patterns for isA relations. Use MINPAR to parse queries and supporting sentences. Use YAGO BOSS for retrieval.

UAbaseanchB¶

Results | Participants | Proceedings | Input | Summary

Run ID: UAbaseanchB
Participant: UAmsterdam
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: aa17674a7f1c410017ef8324bfb555ee
Run description: Wikipedia is used as a pivot. We search Wikipedia for the narrative first, for each retrieved Wikipedia page we use the title to search an anchor text index of Category B.

UAbaselinkA¶

Results | Participants | Proceedings | Input | Summary

Run ID: UAbaselinkA
Participant: UAmsterdam
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: dd5e02602feb804c483f10c96b060a3b
Run description: Wikipedia is used as a pivot. We search Wikipedia for the narrative first, for each retrieved Wikipedia page we follow the external links to search in Category A.

UAcatscombB¶

Results | Participants | Proceedings | Input | Summary

Run ID: UAcatscombB
Participant: UAmsterdam
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: manual
Task: main
MD5: dc039065138b824bd5e1c0fc67fae745
Run description: Wikipedia is used as a pivot. We search Wikipedia for the narrative first, using manually assigned topic categories. For each retrieved Wikipedia page we follow the external links to search in Category B, if no link is found, we search an anchor text index for the page title.

UAcatslinkA¶

Results | Participants | Proceedings | Input | Summary

Run ID: UAcatslinkA
Participant: UAmsterdam
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: manual
Task: main
MD5: 1e796eaaea4b8bdfbfe30e015ad2c313
Run description: Wikipedia is used as a pivot. We search Wikipedia for the narrative first, using manually assigned topic categories. For each retrieved Wikipedia page we follow the external links to search in Category A.

UWAT1¶

Results | Participants | Proceedings | Input | Summary

Run ID: UWAT1
Participant: UWaterlooEng
Track: Entity
Year: 2010
Submission: 10/3/2010
Type: automatic
Task: main
MD5: a395ff3e7c9871ee88339fe19c3ac82b
Run description: Top 50 web pages were retrieved using a search engine in response to queries generated from title and narrative. Snippets surrounding query term occurrences were processed using a named entity tagger. Candidate entities were ranked by tf*idf.

UWAT2¶

Results | Participants | Proceedings | Input | Summary

Run ID: UWAT2
Participant: UWaterlooEng
Track: Entity
Year: 2010
Submission: 10/3/2010
Type: automatic
Task: main
MD5: 4a8ea6e843f7fb506461ce382ae429e2
Run description: Top 50 web pages were retrieved using a search engine in response to queries generated from title and narrative. Snippets surrounding query term occurrences were processed using a named entity tagger. The selected named entities were then ranked and pruned by distributional similarity to seed entities. Seed entities were selected by extracting hyponyms of entity categories specified in the narrative.

UWEntTI¶

Results | Participants | Proceedings | Input | Summary

Run ID: UWEntTI
Participant: UWaterlooEng
Track: Entity
Year: 2010
Submission: 10/1/2010
Type: automatic
Task: main
MD5: 9e0e0f106b6b2db87ee8bdd6dec1d898
Run description: We used Yahoo! for retrieving the initial sets of documents from the Web and for finding homepages of entities.

ValueDoc¶

Results | Participants | Proceedings | Input | Summary

Run ID: ValueDoc
Participant: PITTSIS
Track: Entity
Year: 2010
Submission: 9/29/2010
Type: manual
Task: main
MD5: c69c5e2c762aec00aab5e0acfee9c1d2
Run description: test

Y64¶

Results | Participants | Proceedings | Input | Summary

Run ID: Y64
Participant: HPI
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: 3ffcc87a37b1bbd1d5e9e8ec7ef31488
Run description: - The narrative query is pre-processed using the Stanford Part-of-Speech-Tagger - The source entity is queried to Freebase and the most popular synonyms are retrieved - A key word query is constructed consisting source entity, alternative names and verbs and nouns from the narrative (plus additional rewriting wrt. advanced search engines features) - Candidate documents are retrieved from a commercial web search engine (Life Web: Yahoo!, we took in this run 64 documents) - Candidate target entities are scored as described above - For home page finding, we collect advanced features of the web search engine (such as "find related documents" or "homepage URL occurs in anchors of other Web-pages") - As set of features is applied to score potential homepages, the features are aggregated using weights which are trained using a genetic algorithm - Finally the homepage URLs are mapped to Clueweb-IDs

YahooEnHP¶

Results | Participants | Proceedings | Input | Summary

Run ID: YahooEnHP
Participant: PITTSIS
Track: Entity
Year: 2010
Submission: 9/30/2010
Type: automatic
Task: main
MD5: fbfab8c729436d3300d4e6540d35a446
Run description: Search on Yahoo, Entity Extraction & Homepage Finding