Elamur Research⌕
☰
home / papers / 2023 / Em Direção à Anotação Sintatica - UD de Tweets do Mercado Financeiro

Em Direção à Anotação Sintatica - UD de Tweets do Mercado Financeiro

Summary

Portuguese already has UD annotation guidelines for standard texts, but there have been no rules for the syntactic annotation of tweets. Consequently, no Portuguese tweebank exists either. The authors take a first step in this direction using DANTEStocks. This is a corpus of tweets about the financial market which until now had only semi-automatic UD morphological annotation. The researchers manually analyzed the first tenth of the corpus. They identified phenomena typical of user-generated content that are not covered by the general guidelines for Portuguese: URLs, hashtags, cashtags, mentions, emoticons, RT markers, laughter, and truncated words and sentences. For each phenomenon, they counted its frequency and proposed a rule for choosing the deprel. Each rule was either borrowed from tweebanks of other languages or developed specifically. The main principle is as follows: an element embedded in the sentence receives a relation according to its syntactic role. A detached element is attached to the root via parataxis with a subtype. The annotation examples were checked by three annotators. The paper contains no quantitative evaluation of parsing. Its value lies in ready-made rules on which the guidelines and the manual revision of the automatic annotation of the future first Portuguese tweebank will be built.

Key points

  • The DANTEStocks corpus. 4,048 tweets (~81 thousand tokens) from 2014, up to 140 characters, without normalization. They were collected using the tickers of 73 stocks from the IBovespa index. 405 tweets (10%) were taken for analysis.
  • Frequency. The most frequent phenomena are detached URLs (142) and hashtags embedded in the sentence (98). Cashtags are rare (28 in total). RT markers, mentions and truncations each number about 30. Laughter occurred 8 times, emoticons 6 times. The authors attribute this to the predominantly informational nature of the tweets.
  • Rules for deprel. Detached URLs, hashtags, cashtags and mentions receive parataxis with the subtypes :url, :hashtag, :cashtag, :mention. An embedded URL that depends on a verb receives obl:url. Emoticons and laughter are annotated as discourse. For RT, parataxis was chosen rather than discourse. A mention following RT depends on RT via nmod.
  • Truncation and a typical pattern. A truncated word whose full form could be recovered receives the function of the full word with a new subtype :wtrunc (for example, obl:wtrunc). Truncated sentences are for now resolved case by case. A typical pattern "ticker hashtag + complement + URL", in which the root is the hashtag, is also described.
  • Tools and plans. The examples were annotated in Arborator-Grew-NILC. Next, the rules are to be tested on another 10% of the corpus and formalized into full annotation guidelines.

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

Fields: Linguistics & languages

Topics: Universal Dependencies

Cite

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

@inproceedings{barbosa2023em,
  title = {{Em Direção à Anotação Sintatica - UD de Tweets do Mercado Financeiro}},
  author = {Barbosa, Bryan K. S. and Di-Felippo, Ariani},
  booktitle = {Proceedings of the 2nd Edition of the Universal Dependencies Brazilian Festival},
  year = {2023},
  doi = {10.5753/stil.2023.25490},
  url = {https://doi.org/10.5753/stil.2023.25490}
}

References

11 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Di-Felippo, A., Postali, C., Ceregatto, G., Gazana, L., Silva, E., Roman, N., and Pardo, T. (2021). Descrição preliminar do corpus dantestocks: Diretrizes de segmentação para anotação segundo universal dependencies. In Anais do XIII Simpósio Brasileiro de Tecnologia da Informação e da Lingu… find
  2. Duran, M. S. (2022). Manual de Anotação de Relações de Dependência – Versão Revisada e Estendida. find
  3. Duran, M. S., Oliveira, H., and Scandarolli, C. (2022). Que simples que nada: a anotação da palavra que em corpus de UD. In Proceedings of the Universal Dependencies Brazilian Festival, pages 1–11, Fortaleza, Brazil. Association for Computational Linguistics. find
  4. Guibon, G., Courtin, M., Gerdes, K., and Guillaume, B. (2020). When collaborative treebank curation meets graph grammars. In Proceedings of The 12th Language Resources and Evaluation Conference, pages 5293–5302, Marseille, France. European Language Resources Association. find
  5. Liu, Y., Zhu, Y., Che, W., Qin, B., Schneider, N., and Smith, N. A. (2018). Parsing tweets into Universal Dependencies. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 965… find our card
  6. Nivre, J. et al. (2020). Universal dependencies v2: An evergrowing multilingual treebank collection. Proceedings of the Twelfth International Conference on Language Resources and Evaluation (LREC 2020). find
  7. Sanguinetti, M., Bosco, C., Cassidy, L., Çetinoğlu, Ö., Cignarella, A. T., Lynn, T., Rehbein, I., Ruppenhofer, J., Seddah, D., and Zeldes, A. (2020). Treebanking user-generated content: A proposal for a unified representation in Universal Dependencies. In Proceedings of the Twelfth Language Resou… find
  8. Sanguinetti, M., Bosco, C., Cassidy, L., Özlem Çetinoğlu, Cignarella, A. T., Lynn, T., Rehbein, I., Ruppenhofer, J., Seddah, D., and Zeldes, A. (2022). Treebanking user-generated content: a UD based overview of guidelines, corpora and unified recommendations. Language Resources and Evaluation, 57… find
  9. Sanguinetti, M., Bosco, C., Mazzei, A., Lavelli, A., and Tamburini, F. (2017). Annotating Italian social media texts in Universal Dependencies. In Proceedings of the Fourth International Conference on Dependency Linguistics (Depling 2017), pages 229–239, Pisa,Italy. Linköping University Electronic… find
  10. Silva, F. J. V., Roman, N. T., and Carvalho, A. M. (2020). Stock market tweets annotated with emotions. Corpora, 15(3):343–354. find
  11. Silveira, N., Dozat, T., de Marneffe, M.-C., Bowman, S., Connor, M., Bauer, J., and Manning, C. D. (2014). A gold standard dependency corpus for English. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/barbosa-2023-em-direcao-anotacao-sintatica-ud/