Elamur Research⌕
☰
home / papers / 2016 / Discovering Potential Terminological Relationships from Twitter's Timed Content

Discovering Potential Terminological Relationships from Twitter's Timed Content

Paper · 2016 · Mohammad Daoud and Daoud Daoud

Summary

The paper looks at finding relations between new, unstandardized terms in social media. Tweets are short and noisy, and new Arabic coinages appear faster than terminologists can record them. The authors give up on textual context and compare usage histories instead. Each candidate term ("preterm") gets a daily frequency timeline from an Arabic social-media monitoring platform, which collects tweets through the Twitter API, stems and tokenizes them, and indexes them by term. Pairs of timelines are compared with Dynamic Time Warping (DTW) using a normalized Euclidean distance. DTW rewards matching shapes and peaks even when one curve is shifted by a day or has a different scale. The results are stored as weighted, unconfirmed "preterminological" relationships that experts can refine later. On 630 popular preterms from May 2016, about 270 correct relationships were found at a precision of 0.61. Many of the linked pairs never appeared together in any tweet. This makes temporal behaviour a usable signal for growing lexical resources where text is scarce.

Key points

  • The data are Arabic tweets from May 2016, covering 630 timelines for the most frequent preterms over a 30-day window.
  • The similarity measure is standard DTW, an algorithm first developed for speech recognition, with a normalized Euclidean distance.
  • Two judges rated the top 1108 proposed pairs, with Cohen's Kappa of 0.93. A pair counted as correct if it was event-related or had a lexical relation such as synonymy, acronym, hyponymy, hypernymy or antonymy.
  • Precision depends on the threshold:
    • At th = 0.85, precision is 0.92.
    • At th = 0.65, there are 200 correct links.
    • At th = 0.6, precision is 0.61 with 270 correct links.
    • Precision drops off below about 0.5.
  • The main error source was key terms from two unrelated events that happened in the same period. The authors expect timelines longer than 30 days to reduce these errors.

Interesting ideas

  • I1. Relatedness can come from usage over time instead of co-occurrence in text, so two terms that never share a post can still be linked if their frequency curves rise and fall together.
  • I2. DTW compares the shape of two sequences while tolerating small time shifts and differences in magnitude, which suits signals that lag or lead each other.
  • I3. The paper stores "preterminology": weighted links whose relation type is left open (synonym, acronym, hyponym and so on), kept separate from vetted terminology until someone refines them.
  • I4. The authors list four reasons two timelines can coincide: collocation, simultaneous but unrelated events, one event described by different concepts, and one concept under different words. This list doubles as a ready taxonomy of false positives.
  • I5. Evaluation reports a full precision-versus-threshold curve alongside absolute counts of correct links, with two raters and a kappa score, instead of a single number.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): Mova could build per-entity mention timelines over a dated corpus and use them to propose candidate knowledge-graph edges between new names, with each edge explainable by its aligned peaks.
  • I2 (world): the same alignment could match the event sequences of two retellings of a fable, where events are stretched or slightly reordered, and the warping path itself explains the match.
  • I3 (learn): Mova could keep auto-induced relations in a separate unconfirmed tier with scores, and let an LLM teacher or the agreement of several witnesses promote them into the trusted ontology.
  • I4 (learn): pairs from simultaneous but unrelated events make natural negative controls when evaluating any co-occurrence-based relation inducer.
  • I5 (learn): reporting threshold sweeps for confidence-scored outputs such as silver parses or induced repair rules would show where reliability collapses and help choose cut-offs.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Princess Sumaya University for TechnologyAmman · Jordan7 works · 63 citations🎓American University of MadabaMadaba · Jordan2 works

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 3,066 works

Cite

Cite as Mohammad Daoud and Daoud Daoud. 2016. Discovering Potential Terminological Relationships from Twitter's Timed Content. Proceedings of the 5th Workshop on Cognitive Aspects of the Lexicon (CogALex - V).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{daoud2016discovering,
  title = {{Discovering Potential Terminological Relationships from Twitter's Timed Content}},
  author = {Mohammad Daoud and Daoud Daoud},
  booktitle = {Proceedings of the 5th Workshop on Cognitive Aspects of the Lexicon (CogALex - V)},
  year = {2016},
  url = {https://aclanthology.org/W16-5319/}
}

References

36 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Agrawal, R., et al. (1993). Efficient Similarity Search In Sequence Databases. Proceedings of FODO. Illinois, USA. find
  2. Becker, H., et al. (2011). "Beyond trending topics: Real-world event identification on Twitter." In Proceedings of the Fifth International Conference on Weblogs and Social Media. find
  3. Berndt, D. J. and J. Clifford (1994). Using Dynamic Time Warping to Find Patterns in Time Series. KDD workshop, Seattle, WA. find
  4. Bothos, E., et al. (2010). "Using social media to predict future events with agent-based markets." Cabre, M. T. and J. C. Sager (1999). "Terminology: Theory, methods, and applications." J. Benjamins Publishing: xii, 247 p. ↗
  5. Cabré, T. and R. Nazar (2011). Towards a new approach to the study of neology. Neology and Specialised Translation 4th Joint Seminar Organised by the CVC and Termisti. find
  6. Chen, H., et al. (2012). "Business Intelligence and Analytics: From Big Data to Big Impact." MIS quarterly 36(4): 1165-1188. ↗
  7. Cohen, J. (1960). "A coefficient of agreement for nominal scales." Educational and Psychological Measurement 20(1): 37-46. ↗
  8. Cornolti, M., et al. (2013). A framework for benchmarking entity-annotation systems. Proceedings of the 22nd international conference on World Wide Web, International World Wide Web Conferences Steering Committee. ↗
  9. Daoud, D., et al. (2015). "Time-Sensitive Arabic Multiword Expressions Extraction from Social Networks." IN-TERNATIONAL JOURNAL OF SPEECH TECHNOLOGY. ↗
  10. Daoud, M. (2010). "Utilisation de ressources non conventionnelles et de méthodes contributives pour combler le fossé terminologique entre les langues en développant des" préterminologies" multilingues." Daoud, M., et al. (2009). Constructing Multilingual Preterminological Graphs using various online-community resources. The Eighth International Symposium on Natural Language Processing (SNLP09), Thailand, Bangkok, IEEE. find
  11. Daoud, M., et al. (2009). "Constructing multilingual preterminological graphs using various online-community resources". the Eighth International Symposium on Natural Language Processing (SNLP2009), Thailand: pp. 116 - 121. find
  12. Daoud, M., et al. (2010). "Building Specialized Multilingual Lexical Graphs Using Community Resources." Lecture Notes in Computer Science - Resource Discovery Volume 6162: pp 94-109. find
  13. Daoud, M., et al. (2010). "Passive and Active Contribution to Multilingual Lexical Resources through Online Cultural Activities". NLPKE10, Beijing, China: 4 p. find
  14. Davidson, L. M. (1998). "Knowledge extraction technology for terminology." Federmann, C., et al. (2012). Multilingual terminology acquisition for ontology-based information extraction. Proceedings of the 10th Terminology and Knowledge Engineering Conference (TKE 2012). ↗
  15. Gallego Hernández, D. and S. Herrero Díaz (2014). "Terminology and French-Spanish business translation: evaluating terminology resources for the translation of accounting documents." MonTI : Monografías de Traducción e Interpretación 6. ↗
  16. Gaussier, É. and J.-M. Langé (1997). "Some methods for the extraction of bilingual terminology." 1997). New Methods in Language Processing: 145-153. find
  17. Giannakidou, E., et al. (2014). Towards a Framework for Social Semiotic Mining. Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14), ACM. find
  18. Grinev, M., et al. (2011). "Analytics for the RealTime Web." Proceedings of the VLDB Endowment Volume 4: pp 1391-1394. find
  19. Hartley, A. and C. Paris (1997). "Multilingual Document Production: from Support for Translating to Support for Authoring." Machine Translation 12 (1-2), pp. 109-29. internetlivestats (2015). "Internet Live Stats - Internet Usage & Social Media Statistics." Retrieved 1/2/2015, 2015, from http://www.internetlivestats.com/one-second/. ↗
  20. Kageura, K. (2002). "The Dynamics of Terminology: A descriptive theory of term formation and terminological growth", Terminology and Lexicography Research and Practice 5, 322 p. find
  21. Kandias, M., et al. (2013). Proactive insider threat detection through social media: The YouTube case. Proceedings of the 12th ACM workshop on Workshop on privacy in the electronic society, ACM. find
  22. Keogh, E. and C. Ratanamahatana (2004). "Exact indexing of dynamic time warping." Knowledge and Information Systems: 358-386. ↗
  23. Kim, Y. G., et al. (2005). "Terminology construction workflow for Korean-English patent MT". MT Summit X. Phuket, Thailand: 5 p. find
  24. Kwak, H., et al. (2010). What is Twitter, a Social Network or a News Media? The 19th International World Wide Web (WWW) Conference, Raleigh NC (USA). find
  25. Peñas, A., et al. (2001). Corpus-based terminology extraction applied to information access. Proceedings of Corpus Linguistics, Citeseer. find
  26. Roche, C., et al. (2009). Ontoterminology: A new paradigm for terminology. International Conference on Knowledge Engineering and Ontology Development. find
  27. Sager, J. C. (1990). A Practical Course in Terminology Processing Amesterdam/Philadelphia, John Benjamins Publishing Company, 266 p. find
  28. Sakoe, H. and S. Chiba (1978). "Dynamic programming algorithm optimization for spoken word recognition." IEEE Transactions on Acoustics, Speech and Signal Processing: 43- 49. ↗
  29. Salvador, S. and P. Chan (2007). "Toward accurate dynamic time warping in linear time and space." Intelligent Data Analysis: 561 - 580. ↗
  30. Speriosu, M., et al. (2011). "Twitter polarity classification with label propagation over lexical links and the follower graph." EMNLP '11 Proceedings of the First Workshop on Unsupervised Learning in NLP: 53-63 find
  31. Suonuuti, H. (1997). " Guide to Terminology." Nordterm 8. Helsinki:TSK. find
  32. Twitter (2015). "Twitter." Retrieved 1/2/2015, 2015, from twitter.com. find
  33. Uherčík, T., et al. (0210). "Utilizing Microblogs for Web Page Relevant Term Acquisition." SOFSEM 2013: Theory and Practice of Computer Science - Lecture Notes in Computer Science Volume 7741: pp 457-468. find
  34. Vasconcellos, M., et al. (2001). "Terminology and machine translation." Handbook of terminology management 2: 697-723. find
  35. Yang, Y. T., et al. (2013). "Mining social media and web searches for disease detection." Journal of public health research 2(1): 17. ↗
  36. Zhao, W. X., et al. (2011). "Topical keyphrase extraction from Twitter." HLT '11 Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies - Volume 1 379-388. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/daoud-2016-discovering-potential-terminological-relationships-twitter/