wiki:WAC-X

Version 1 (modified by Roland Schäfer, 4 years ago) (diff)

--

10th Web as Corpus Workshop (WAC-X) and EmpiriST Shared Task

The World Wide Web has become increasingly popular as a source of linguistic data, not only within the NLP communities, but also with theoretical linguists facing problems of data sparseness or data di­versity. Accordingly, web corpora continue to gain importance, given their size and diversity in terms of genres/text types. The field is still new, though, and a number of issues in web corpus construction need much additional research, both fundamental and applied. These issues range from questions of corpus design (e.g., corpus composition assessment, sampling strategies and their relation to crawling algorithms, handling of duplicated material) to more technical aspects (e.g., efficient implementation of individual post-processing steps in document cleansing and linguistic annotation, or large-scale paral­lelization to achieve web-scale corpus construction). Similarly, the systematic evaluation of web cor­pora, for example in the form of task-based comparisons to traditional corpora, has only recently shifted into focus. For almost a decade, the ACL SIGWAC (http://www.sigwac.org.uk/), and especially the highly suc­cessful Web as Corpus (WAC) workshops have served as a platform for researchers interested in com­pilation, processing and application of web-derived corpora. Past workshops were co-located with ma­jor conferences on computational linguistics and/or corpus linguistics (such as EACL, NAACL, LREC, WWW, Corpus Linguistics). As in previous years, the 10th Web as Corpus workshop (WAC-X) invites contributions pertaining to all aspects of web corpus creation, including but not restricted to

  • data collection (both for large web corpora and smaller custom web corpora)
  • cleaning/handling of noise
  • duplicate removal/document filtering
  • linguistic post-processing (including non-standard data)
  • automatic generation of meta data (including register, genre, etc.)
  • corpus evaluation (quality of text and annotations, comparison to other corpora, etc.)

Furthermore, aspects of usability and availability of web-derived corpora are highly relevant in the con­text of WAC-X

  • development of interfaces
  • visualization techniques
  • tools for statistical analysis of very large (e.g., web-derived) corpora
  • long-term archiving
  • documentation and standardization
  • legal issues

Finally, reports of the use of web corpora in language technology and linguistics are welcome, for ex­ample information extraction & opinion mining

  • language modeling, distributional semantics
  • machine translation
  • linguistic studies of web-specific forms of communication
  • linguistic studies of rare phenomena
  • web-specific lexicography, grammaticography, and language documentation

EmpiriST 2015 shared task

The EmpiriST 2015 shared task aims to encourage the developers of NLP applications to adapt their tools and resources to the processing of German discourse in genres of computer-mediated communica­tion (CMC), including both dialogical (chat, SMS, social networks, etc.) and monological (web pages, blogs, etc.) texts. Since there has been relatively little work in this area for German so far, the shared task focuses on tokenization and part-of-speech tagging as the core annotation steps required by virtu­ally all NLP applications. While we have a particular interest in robust tools that can be applied to dia­logical CMC and web corpora alike, participants are allowed to use different systems for the two sub­sets or submit results for one subset only. A substantial number of teams from German-speaking countries have already expressed their interest to participate in EmpiriST 2015. Knowledge of German is not essential for participation, though, since there are sufficient amounts of manually annotated training data (at least 10,000 tokens) and key docu­ments are provided in English.

The final workshop of EmpiriST 2015 will be co-located with WAC-X. It will include a detailed pre­sentation of the task and results, a poster session with all participating systems, oral presentations of se­lected systems, and a plenary discussion about the challenges of CMC in general as well as German CMC genres in particular.

Attachments (14)