Knowledge and Media 2012 Lecture 10: Research proposal QA

RESEARCH PROPOSAL QAKM 2012 Lecture 10

Friday, November 30, 12

OVERVIEW

Research Proposal

Finding your topic

Defining your research question

Writing it up

Research Poster: Communicating your idea visually

Peer Review: Providing positive feedback

Lightning Talk: Condense your idea

Logistics


RESEARCH PROPOSAL


FINDING YOUR TOPIC

Which topics in the course did you like?

Which problem should be solved?

Think out of the box, what have you seen in the literature in other lectures that may be of use here?

Sleep on it.

Am I still excited about it? OK, go to step 2


INSPIRATION


INSPIRATION


INSPIRATION

&URZGOLQNLQJ$XWKRU��7KRPDV�3ORHJHU��.H\ZRUGV��:HE�RI�'DWD��FURZGVRXUFLQJ��,QWURGXFWLRQ7KLV� GRFXPHQW� LV� D� SURSRVDO� IRU� D� UHVHDUFK� SURMHFW� WKDW� DLPV� WR� HYDOXDWH� WKH� EHQHILWV� RI� XVLQJ�FURZGVRXUFLQJ�RQ�WKH�:HE�RI�'DWD��)LUVW��ZH�ZLOO�VXFFLQFWO\�LQWURGXFH�WKH�:HE�RI�'DWD�DQG�GHVFULEH�WKH� SUREOHP� WKDW�ZH� H[SHFW� WR� VROYH��1H[W��ZH�ZLOO� LQWURGXFH� FURZGVRXUFLQJ� DQG� GHVFULEH�ZK\�ZH�H[SHFW�WKDW�LW�ZLOO�VROYH�WKH�SUREOHP�WKDW�ZH�KDYH�GHVFULEHG��:H�ZLOO�FRQFOXGH�ZLWK�GHVFULELQJ�KRZ�WKH�UHVHDUFK�SURMHFW�VKRXOG�HYDOXDWH�RXU�SURSRVHG�VROXWLRQ��,QWURGXFLQJ�7KH�:HE�RI�'DWD7KH�:HE�RI�'DWD�LV�EXLOW�DURXQG�WZR�FRUH�SULQFLSOHV��7KH�ILUVW�SULQFLSOH�LV�PDNLQJ�GDWD�DYDLODEOH�RQ�WKH�:HE��7KH�VHFRQG�SULQFLSOH�LV�FUHDWLQJ�OLQNV��HQDEOLQJ�KXPDQV�RU�PDFKLQHV�WR�QDYLJDWH�WKH�:HE�RI�'DWD��%HUQHUV�/HH��8QOLNH�OLQNV�EHWZHHQ�UHVRXUFHV�RQ�WKH�WUDGLWLRQDO�:HE�RI�'RFXPHQWV��OLQNV�EHWZHHQ�UHVRXUFHV�RQ�WKH�:HE�RI�'DWD�DUH�W\SHG�OLQNV��7KLV�PHDQV�WKDW�WKH�OLQNV�DOORZ�XV�WR�VSHFLI\�PHDQLQJIXO�UHODWLRQVKLSV�EHWZHHQ�UHVRXUFHV��7KLV� GLIIHUHQFH� LQ� OLQNV� RQ� WKH�:HE� RI�'RFXPHQWV� DQG� WKH�:HE�RI�'DWD� LV� LOOXVWUDWHG� LQ� )LJXUH� ��ZKLFK�LV�UHSURGXFHG�IURP�.RLYXQHQ��0LOOHU��2Q�WKH�OHIW��ZH�VHH�DQ�H[DPSOH�RI�WKH�:HE�RI�'RFXPHQWV��ZLWK�XQW\SHG�OLQNV��2Q�WKH�ULJKW��ZH�VHH�DQ�H[DPSOH�RI�WKH�:HE�RI�GDWD��ZLWK�W\SHG�OLQNV�VXFK�DV�³GHSHQGV2Q´�RU�³FUHDWRU´�

��)LJXUH��/HIW��/LQNV�LQ�WKH�:HE�RI�'RFXPHQWV��5LJKW��/LQNV�LQ�WKH�:HE�RI�'DWD�

�7KH�W\SHG�OLQNV�WKDW�IRUP�WKH�:HE�RI�'DWD�DUH�W\SLFDOO\�FUHDWHG�E\�WZR�GLIIHUHQW�PHWKRGV��DV�GHILQHG�E\�%L]HU� HW� DO�� 7KH� ILUVW� LV� WKH�PDQXDO�PHWKRG�� LQ�ZKLFK� D� SHUVRQ�PDQXDOO\� LGHQWLILHV� WZR�UHVRXUFHV� WR� OLQN��DQG� WKHQ�FUHDWHV� WKH� OLQN��%HFDXVH� WKLV�DSSURDFK�REYLRXVO\�GRHV�QRW� VFDOH�ZHOO� WR�ODUJH�VHWV�RI�UHVRXUFHV��DXWRPDWHG�PHWKRGV�IRU�OLQNLQJ�UHVRXUFHV�ZHUH�GHYHORSHG��$Q�H[DPSOH�RI�VXFK�DQ�DXWRPDWHG�PHWKRG� LV� VHHQ� LQ� WKH� µ5')�%RRN�0DVKXS¶� WKDW�ZDV� �GHYHORSHG�E\�%L]HU��*DX��ZKLFK�XVHV�D�UHODWLYHO\�VLPSOH�SDWWHUQ�PDWFKLQJ�DOJRULWKP�WKDW�XVHV�,6%1�QXPEHUV� WR� FUHDWH� OLQNV�EHWZHHQ� UHVRXUFHV� DERXW� WKH� VDPH�ERRN��$Q�H[DPSOH�RI� � D�PRUH� DGYDQFHG�


DEFINING YOUR RQ

Dig into the literature, has my problem been researched before?

If so, what techniques have been used to deal with it?

Is my proposed solution novel and viable?

No literature? Ask yourself if the problem you want to investigate is relevant.


WRITING IT UP

Make sure the proposal is self-contained, i.e., any peer reviewer should understand your main problem and proposed solution by just reading your document

Use examples, or figures to explain your proposal

Don’t forget any parts (literature etc.)


YOUR RESEARCH POSTER


VISUALISING YOUR IDEA

A picture says more than a thousand words

Come up with a catchy example

Don’t paste text from your proposal into your poster!


Knowledge & Media Conference 2011December 12th VU University Amsterdam

J u i c i n g t h e L O D C l o u d w i t h W o r d N e t

Ben A. StudentVU University Amsterdam

[email protected]

Use WordNet to suggest new linksin the LOD Cloud

Use a validation metricto determine the

relevance of new links

The number of data sets that link to 1, 2, 3, 4, 5, 6 to 10 or more than 10 other data sets

Though at first glance it may seem as if there are many connections between data sources in the LOD Cloud, a more detailed look will show that most data sources are connected to only one or two other data sources. This also follows from the LOD Cloud statistics. More than 50% of the data sources in the LOD Cloud link to no more than two other

sources, and more than 66% of them link to no more than three other sources.

Use WordNet as a semantic and relational knowledge base to analyze the subjects,

predicates and objects of existing triples in the LOD Cloud and propose new links

between data items based on the linguistic relations defined in WordNet. Nouns, verbs, adjectives and adverbs are grouped into sets

of cognitive synonyms called synsets, each expressing a distinct concept. Synsets are

interlinked by means of conceptual-semantic and lexical relations.

WordNet contains 3 major relation types that could be utilized: Synonymy relations; relations between words that have similar meaning, e.g. ‘forest’ is synonymous to ‘wood’. Hyponymy relations; relations

between words that are sub concepts or super concepts of each other, e.g. ‘taxi’ is a sub concept of ‘car’, which in turn is a sub concept of ‘vehicle’. Meronymy relations;

relations that define if words are sub concepts, e.g. ‘bumper’ is a part of ‘car’.

Derive identifying terms from existing RDF Triples

▼Match these terms against synsets in

WordNet▼

Use synonymy hyponymy and meronymy relations

▼Suggest links based on

distance in the linguistic WordNet relation and matching percentage

▼Use a filter for

domain specific applications


PEER REVIEW


PROVIDING POSITIVE FEEDBACK

Meant to help each other in improving the proposal

Read critically, but fairly

Provide detailed as well as high level comments to aid the author whose work you are reviewing


LIGHTNING TALK


CONDENSING YOUR IDEA

Explain the core of your idea in one minute

Don’t try to summarise your entire proposal

Create a single slide to communicate your idea


Try-on eyewearSerious gaming for opticians


Music discovery and recommendations using the Semantic Webby Justin van Wees

Problem statement Research question• Enormous collections of music are available

online• To !nd new, possibly interseting music,

users can: - Read reviews - Listen to lots of tracks - ... or use colleborative !ltering services like:

20+ million songs

Colleborative !ltering methods have several disadvantages:• compares on (very few) high level

metadeta properties• content-based properties are

ignored• prone to a popularity bias; makes

it unlikely for artists located in the ‘Long Tail’ to be ever recommend

• recommendations are not transparent

The Top–737 artists accumulate 50% of total playcounts (Celma and Cano, 2008).

Can we create a system that generates personalized music recommendations by using Semantic Web technologies and currently available Linked Open Data?

We wan to:• help users discover new music that

!ts personal taste• combine collaborative !ltering data,

expert-based data and high-level content based features

• provide meaningful feedback on why items are suggested (Cohen and Fan, 2000)

• intergrate with a (popular) existing service

Methods• collect music related linked data and map it to the

Music Ontology (Raimond et al., 2007)• build and evaluate recommendation methods• determine what information on recommendations is useful to

the end-user

References Casey, M., Veltkamp, R., Goto, M., Leman, M., Rhodes, C., and Slaney, M. (2008). Content-based music information retrieval: current direc-tions and future challenges. Proceedings of the IEEE, 96(4):668–696. Celma, O. and Cano, P. (2008). From hits to niches?: or how popular artists can bias music recommendation and discovery. In Proceedings of the 2nd KDD Workshop on Large-Scale Recommender Systems and the Net"ix Prize Competition, page 5. ACM. Cohen, W. and Fan, W. (2000). Web-collaborative !ltering: Recommending music by crawling the web. Computer Networks, 33(1):685–698. Raimond, Y., Abdallah, S., Sandler, M., and Giasson, F. (2007). The music ontology. In Proceedings of the International Conference on Music Information Retrieval, pages 417– 422. Citeseer. http://en.wikipedia.org/wiki/ITunes_Store#Music, http://en.wikipedia.org/wiki/Spotify http://dbtune.org/

Text


Crowdsourcing for documentation and revitalization of endangered languages

Language embeds knowledge…

documenting

sharing

in the hands of the crowd


LOGISTICS


SUBMITTING TO EASYCHAIR


REVIEWING


LIGHTNING TALK SLIDE

Submit a PDF file with one single slide to the dropbox, named <LASTNAME>_slide.pdf

Deadline: Friday 7 December 23:59 CET.

Make sure the slide is in landscape mode and has at dimensions 1024x768 or greater with same proportions


FINAL VERSION

Process reviewers’ comments and lightning talk comments

Explain your improvements in a response letter

Deadline: Sunday 23 December 23:59 CET

Resubmit using Easychair


QUESTIONS?


Knowledge and Media 2012 Lecture 10: Research proposal QA

Documents

data items

linked data

data sourcesuse

data setsrelations

new music

collaborative filtering

lod cloud link

songs expertbased data