Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Artículo científico |
| Sprache: | en |
| Veröffentlicht: |
Instituto Politécnico Nacional
2013
|
| Schlagworte: | |
| Online-Zugang: | https://www.redalyc.org/articulo.oa?id=402640462009 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Inhaltsangabe:
- A POS Tagger for Social Media Texts Trained on Web Comments Melanie Neunerdt Michael Reyer Rudolf Mathar Computación of part German speech tagging opinion mining Using social media tools such as blogs and forumshave become more and more popular in recent years. Hence, ahuge collection of social media texts from different communitiesis available for accessing user opinions, e.g., for marketingstudies or acceptance research. Typically, methods from NaturalLanguage Processing are applied to social media texts toautomatically recognize user opinions. A fundamental componentof the linguistic pipeline in Natural Language Processingis Part-of-Speech tagging. Most state-of-the-art Part-of-Speechtaggers are trained on newspaper corpora, which differ in manyways from non-standardized social media text. Hence, applyingcommon taggers to such texts results in performance degradation.In this paper, we present extensions to a basic Markov modeltagger for the annotation of social media texts. Consideringthe German standard Stuttgart/T ¨ubinger TagSet (STTS), wedistinguish 54 tag classes. Applying our approach improves thetagging accuracy for social media texts considerably, when wetrain our model on a combination of annotated texts fromnewspapers and Web comments.Index Terms—Natural language processing, part-of-speechtagging, opinion mining, German. 2013 artículo científico 1870-9044 https://www.redalyc.org/articulo.oa?id=402640462009 en http://www.redalyc.org/revista.oa?id=4026 Polibits application/pdf Instituto Politécnico Nacional Polibits (México) Vol.48