Download presentation
Presentation is loading. Please wait.
Published byCornelia Floyd Modified over 8 years ago
1
14. June 2016 Mapping democracy Indira Ishmurzina miss.indi@yandex.ru
2
Outline Introduction Research questions: “The Nationalities of Issues: Rights Types” “For the ppl of Iran - #iranelection RT” Conclusion 2 Indira Ishmurzina
3
Introduction The rise of digital communication technologies has placed new emphasis on an old problem: information overload. Solution is use of search engines to filter information and to create ordered lists of relevant sources Mapping and visualising data sets are means by which information can be made more accessible and useful We look at Google results and see society, instead of Google Indira Ishmurzina3
4
Research questions: Where does “Google studies” end, and social research begin? What kinds of findings may be made by interpreting search engine results, especially the rankings of sites for particular queries? What kinds of findings can be made by comparing results across the many local versions of Google? (Google.de, Google.uk, Google.it) Indira Ishmurzina4
5
The Nationalities of Issues: „Right Types“ RESEARCH STRATEGY: Employ Google to show most prominent types of rights per country. METHOD: Query the term “rights” in the local languages in the local Google versions Manually read the results and make lists of the top ten distinctive rights types, leaving them in the order that Google provided. Google.com with query “rights” Google.de with query “rechte” Google.fr with query “droits” Google.ru with query “prava” Most significant rights types per country according to local Google results of the query for “rights” in the local languages. Indira Ishmurzina5
6
The Nationalities of Issues: „Right Types“ FINDINGS: “what matters to each language or culture” Countries could be said to have distinctive concerns, compared to other countries, as read from Google results. For example, “computer programmers’ rights” in Japan and the “right to oblivion” (the right to have personal data deleted) in Italy are unique to the respective countries. Indira Ishmurzina6
7
The Nationalities of Issues: Right Types“ 7 Information visualization Indira Ishmurzina
8
“For the ppl of Iran - #iranelection RT” RESEARCH STRATEGY: The question is, could the hundreds of thousands of tweets about the Iran election crisis be made into a comprehensible account of what has been happening on the ground as well as online? METHOD: Assemble the top-3 “retweets” per day, and order them chronologically. DATA BREAKDOWN (10-30 June 2009): Tweets tagged with #iranelection: 653,883 Unique number of Twitter users using #iranelection tag: 99,811 Indira Ishmurzina8
9
#iranelection RT Top 3 retweets per day 9 Indira Ishmurzina
10
“For the ppl of Iran - #iranelection RT” FINDINGS: The most significant retweets show the urgency and the emotion of those twenty days in June Tweets shows how tweeters respond to what is happening online and on the ground. Tweets reporting important websites Indira Ishmurzina10
11
Goal of the topic “The Nationalities of Issues: Rights Types” What is the problem addressed in the paper? What kinds of findings may be made by interpreting search engine results, especially the rankings of sites for particular queries? What kinds of findings can be made by comparing results across the many local versions of Google? (Google.de, Google.uk, Google.it) How does the solution look like? The study showed hierarchies of rights type per country. How is it evaluated? The goal was to show differences in what matters to each language or culture. “For the ppl of Iran - #iranelection” What is the problem addressed in the paper? The question is, could the hundreds of thousands of tweets about the Iran election crisis be made into a comprehensible account of what has been happening on the ground as well as online? How does the solution look like? The most retweeted tweets have been filtered and organised chronologically How is it evaluated? The most significant retweets show almost all events which happened in these days in Iran Indira Ishmurzina11
12
14. June 2016 AIDA-light: High-Throughput Named-Entity Disambiguation Indira Ishmurzina miss.indi@yandex.ru
13
Outline Introduction System architecture AIDA-light processing of sample sentence Conclusion 13 Indira Ishmurzina
14
Introduction The Web of Linked Data provides a wealth of data and knowledge sources that are richly interlinked at the entity level. To advance the Web of Linked Data, mapping ambiguous names in structured and unstructured contents onto knowledge bases would be a vital asset. The problem of mapping ambiguous entity onto a set of known target entities is referred to as Named Entity Disambiguation (NED). State-of-the-art NED methods face major compromises regarding run-time efficiency/scalability vs. accuracy. Indira Ishmurzina14
15
AIDA-light processing of sample sentence 15Indira Ishmurzina
16
16Indira Ishmurzina AIDA-light processing of sample sentence
17
17Indira Ishmurzina AIDA-light processing of sample sentence
18
Contributions AIDA-light makes a judicious choice of contextual features to compute pairwise mention-entity and entity-entity similarities. AIDA-light is the first approach that uses a thematic domain hierarchy to capture measures for domain-entity and entity-entity coherence. AIDA-light uses a new two-stage algorithm in which we determine the “easy and low-cost" mappings first. AIDA-light is a complete system for NED, which is orders of magnitude faster while achieving comparable output quality. Indira Ishmurzina18
19
Conclusion AIDA-light is a new NED system that combines high output quality with fast run-time. AIDA-light is geared for high-throughput usage at Web scale due to: two-stage algorithm, simple but expressive features, low footprint. Its architecture easily allows it to scale out the processing of a large corpus across the nodes of a cluster or distributed platform. AIDA-light runs well even on low-end nodes due to small memory consumption. Indira Ishmurzina19
20
Goal of the topic What is the problem addressed in the paper? This paper reconsiders to develop an NED system that achieves both high throughput and high output accuracy. How does the solution look like? AIDA-light uses a new kind of two-stage mapping algorithm. How is it evaluated? The accuracy of AIDA-light is competitive to the very best NED systems, while its run-time is comparable to or better than the performance of the fastest systems. Indira Ishmurzina20
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.