Pheme (research project)
| Phemes | |
|---|---|
 Logo of the project |
|
| Client: | European Commission |
| Management: | Kalina Bontcheva |
| Location: | University of Sheffield |
| Homepage: | http://www.pheme.eu/ |
Pheme is a 36-month EU funded research project aimed at analyzing the trustworthiness of Internet information. The term pheme is a neologism ; it refers to memes whose characteristics include truthfulness. Also, the name alludes to the ancient goddess of fame and rumor, Pheme in Greek and Fama in Latin .
Background and goals
The project started on January 1, 2014. The total costs amount to EUR 4,269,938, the EU funding from the 7th Research Framework Program to EUR 2,916,000.
Phemes are "rumor memes " as they often appear on social media . These can be targeted false reports, but also simply misunderstandings. Once they spread through services like Twitter or Facebook , it can be very difficult to determine their veracity. Up to now this has only been possible in resource-intensive manual procedures.
In addition to the three typical challenges of big data - data volume, data diversity, speed - there is also a fourth with social media , namely determining the truthfulness. The aim of the Pheme project is to analyze content in real time and to assess the correctness of the information it contains. When information spreads through a social network , each individual decides whether he wants to pass it on or not; the personal assessment of the correctness also plays a role. Pheme analyzes the language used, the dissemination of information through a network, and the spatial and temporal context of the information to develop a real-time polygraph for social media. Emergency services and disaster control, for example, can benefit from this: if they use social media for their rapid warning and response systems, Pheme can help them to recognize a potential false alarm . The project results will be directly applicable to medical information systems and digital journalism. In addition, many of the associated algorithms are to be published as open source .
method
Scientists from different fields work together in the interdisciplinary project: language technologies, web science, analysis of social networks and information visualization. They deal with four types of rumors: speculation, controversy, misinformation, and disinformation. In order to assess the trustworthiness of information, first analyze the linguistic ( semantic and syntactic ) characteristics of the text in question; The next step is a comparison with other sources classified as particularly trustworthy; and finally the distribution channels on the Internet are examined. The results are visualized with the help of a dashboard that is created in cooperation with the media analysis company webLyzard. In addition to the automated algorithms, a manually analyzed data set is also used.
One of the important things in the project is the automatic assessment of the reliability of sources. For example, a news tweet from the BBC would have more weight than a tweet from an unknown source. The system is also intended to identify both computer-generated spam and users who specialize in misinformation.
Two case studies will be carried out: one examines information from the health sector, where false reports can be particularly harmful; the focus of the others is information used by journalists. The medical case study deals, among other things, with so-called rumor intelligence, i.e. the gathering of information from rumors such as the spread of news of the swine flu outbreak. The journalistic case study deals with the credibility of user-generated content.
partner
The project is carried out in partnership between several universities and companies. The universities involved are Sheffield University (where the GATE text analysis platform used by Pheme was developed), Warwick University and King's College London in Great Britain, Saarland University in Germany and Modul University Vienna in Austria. Four companies are also involved: Atos (Madrid, Spain), iHub (Nairobi, Kenya), Ontotext (Sofia, Bulgaria) and swissinfo (Bern, Switzerland).
Individual evidence
- ^ Scientists develop a lie detector for tweets . Telegraph . Retrieved February 20, 2014.
- ↑ a b c d e About PHEME . Retrieved May 4, 2015.
- ↑ Computing Veracity Across Media, Languages, and Social Networks . Retrieved May 4, 2015.
- ^ Bulgaria: ICT Investment for SME Growth . Retrieved May 4, 2015.
- ^ Scientists Plan Lie-Detector For Tweets . Sky News. Retrieved February 20, 2014.
- ^ Robert Power, et al .: Finding Fires with Twitter . In: Proceedings Australasian Language Technology Association Workshop 2013 . 2013.
- ^ Paul S. Earle, et al .: Twitter earthquake detection: Earthquake monitoring in a social world . In: Annals of Geophysics . 2012.
- ↑ Researchers working on social media 'lie detector' . Arab News . Retrieved February 20, 2014.
- ↑ a b Researchers want to automatically identify social media lies . Retrieved May 4, 2015.
- ↑ Pheme: International research project is working on a lie detector for social media content . Retrieved May 4, 2015.
- ↑ Researchers are working on a lie detector to sniff out false tweets . Engadget . Retrieved February 20, 2014.
- ↑ Lies in 140 characters: researchers expose Twitter cheaters . Retrieved May 4, 2014.
- ^ EU project to build lie detector for social media . Retrieved February 20, 2014.
- ^ Lie detector on the way to test social media rumors . BBC News . Retrieved February 20, 2014.
- ↑ BuzzWord: rumint so RUMINT . Retrieved May 5, 2015.
- ^ Consortium . Retrieved May 5, 2015.
literature
- Leon Derczynski and Kalina Bontcheva. Passive-Aggressive Sequence Labeling with Discriminative Post-Editing for Recognizing Person Entities in Tweets. Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics , Volume 2: Short Papers. Göteborg: Association for Computational Linguistics, 2014. pp. 69–73.