Re­sources

This site con­tains datasets, ap­pli­ca­tions, tools and other re­sources pub­licly pro­vid­ed by the KE Group.​

The Knowl­edge En­gi­neer­ing Group dis­solved at the end of 2021 due to the de­par­ture of Prof.​ Fürnkranz to Linz.​ Thanks to all for­mer mem­bers of the group, to all the stu­dents that were part of our group, to all par­tic­i­pants in our cours­es and lec­tures, and to all sup­port­ers.

Since Jan­uar 2022, this is just a mir­ror of the Knowl­edge En­gi­neer­ing site.

The fol­low­ing list gives a short de­scrip­tion of the avail­able re­sources:

Datasets

  • EUR-Lex text col­lec­tion
    The EUR-Lex text col­lec­tion pro­vides a large mult­la­bel clas­si­fi­ca­tion bench­mark with up to 4000 dif­fer­ent class­es.
  • Datasets for Grad­ed Mul­ti­l­abel Clas­si­fi­ca­tion
    The known BeLaE Dataset and two new datasets from med­i­cal text clas­si­fi­ca­tion and movie rat­ings.
  • In­ci­dent-Re­lat­ed Twit­ter Datasets
    These datasets com­prise la­beled tweets from 10 major cities in the En­glish-speak­ing world.​ The tweets were se­lect­ed and la­beled for the do­main of in­ci­dent de­tec­tion.
  • Med­i­cal Con­cept Em­bed­dings
    Con­cept vec­tor rep­re­sen­ta­tions learned from a large la­beled back­ground corpus.​ These were used for com­put­ing the se­man­tic sim­i­lar­i­ty be­tween terms from the med­i­cal do­main
  • DIP-SumEval: A Data Set of Human Sum­ma­ry Eval­u­a­tions
    A dataset con­tain­ing over 400 au­to­mat­i­cal­ly gen­er­at­ed sum­maries for 49 top­ics of an data set for mul­ti-doc­u­ment sum­ma­riza­tion, 1274 judge­ments ac­cord­ing to 11 text and sum­ma­ry qual­i­ty cri­te­ria on a Lik­ert-scale (1 to 5) per­formed by 26 trained an­no­ta­tors, and 43218 pair­wise judge­ments ac­cord­ing to 6 cri­te­ria per­formed by 64 crowd-work­ers.

On­tolo­gies

  • UI² On­tol­o­gy
    The UI² On­tol­o­gy is a for­mal on­tol­o­gy for de­scrib­ing user in­ter­faces, their com­po­nents, and the pos­si­ble in­ter­ac­tions with them.

Soft­ware

  • Com­put­er Poker Bots and the TUD poker frame­work
    A small repos­i­to­ry of (old) Com­put­er Poker Bots and our frame­work for de­vel­op­ing, com­par­ing bots and play­ing against them with a GUI
  • At­tach­ment Check­er
    A Thun­der­bird plu­g­in that learns to warn you when you for­get to at­tach a file to your mes­sage.
  • Clas­si­fi­ca­tion GUI
    A graph­i­cal user in­ter­face that al­lows to in­tu­itive­ly as­sign con­cepts from an on­tol­o­gy to a set of doc­u­ments in order to quick­ly and eas­i­ly de­vel­op a (mul­ti­l­abel) clas­si­fi­ca­tion dataset.
  • Pee­wit
    A light-weight meta-frame­work for ma­chine learn­ing ex­per­i­ments.
  • FeGeLOD
    A tool for gen­er­at­ing ma­chine-learn­ing fea­tures from Linked Open Data.
  • Ex­plain-a-LOD
    A tool for gen­er­at­ing pos­si­ble ex­pla­na­tions for statis­tics based on Linked Open Data.
  • SeCo
    A frame­work for Sep­a­rate-and-Con­quer Rule Learn­ing.
  • Per­cep­tro­ve­ment
    A high­ly mod­u­lar frame­work for the ef­fi­cient Per­cep­tron al­go­rithm con­tain­ing a great col­lec­tion of ef­fec­tive ex­ten­sions
  • MoB4LOD
    A frame­work for cre­at­ing cus­tomized brows­er ap­pli­ca­tions for Linked Open Data
  • JFreeWeb­Search
    A free (i.​e.,​ no reg­is­tra­tion and API key re­quired) Java li­brary to per­form search­es on the web
  • On­tol­o­gy Match­ing Tools
    The KE group has de­vel­oped a va­ri­ety of on­tol­o­gy match­ing tools.
  • Grad­ed Mul­ti­l­abel Clas­si­fi­ca­tion, Code and Data
    The code and data used for our paper about pair­wise grad­ed mul­ti­l­abel classification.​ In this set­ting, a label is not only pre­sent or ab­sent, but can have sev­er­al grades, e.​g.​ stars.
  • P³oodle
    A brows­er ex­ten­sion/add-on for per­son­al­ized pri­va­cy-pro­tect­ed web search.
  • Ai­TextML
    Learn con­tin­u­ous vec­tor rep­re­sen­ta­tions joint­ly for words, doc­u­ments, and labels.​ Use cor­po­ra with la­belled doc­u­ments and use also de­scrip­tions of labels.​ This en­ables also to do ze­ro-shot learn­ing, i.​e.,​ to pre­dict la­bels for which no doc­u­ments were ob­served dur­ing train­ing.

Com­put­ing

Kontakt

small ke-icon

Knowledge Engineering Group

Fachbereich Informatik
TU Darmstadt

S2|02 D203
Hochschulstrasse 10

D-64289 Darmstadt

Sekretariat:
Telefon-Symbol+49 6151 16-21811
Fax-Symbol +49 6151 16-21812
E-Mail-Symbol info@ke.tu-darmstadt.de

 
A A A | Drucken | Impressum | Sitemap | Suche | Mobile Version
zum Seitenanfangzum Seitenanfang