Semi-Manual Annotation of Topics and Genres in Web Corpora : The Cheap and Fast Way
| Authors | |
|---|---|
| Year of publication | 2022 |
| Type | Article in Proceedings |
| Conference | Proceedings of the Sixteenth Workshop on Recent Advances in Slavonic Natural Languages Processing, RASLAN 2022 |
| MU Faculty or unit | |
| Citation | |
| Web | |
| Keywords | web corpus; text corpus; topic; genre; text annotation |
| Description | In this paper we present a cheap and fast semi-manual approach to annotation of topics and genres in web corpora. The main feature of our method is assigning the same topic or genre label to all web pages coming from websites most represented in the corpus. We assume that web pages within a site share the topic of the whole domain. According to the evaluation of texts coming from sites that were manually assigned a topic label, our hypothesis holds in 92 % of cases. In other words, the noise in these semi-manually labelled web pages is just 8 %. That is clean enough to train a classifier of texts from websites not seen in the process. The procedure of fast manual topic and genre labelling of web domains is described in this paper. Recommendations for training a topic or genre classifier using semi-manually labelled texts from large websites follow. |
| Related projects: |