Out of seven KLIP applications that were submitted in response to the call 2014, four
have been selected and are supported by ZüKL with 22'000 CHF.
Below you find a list with short descriptions.
Content:
The online portal for the ZECO (Zurich Electronic Corpus of Language Varieties) already has several functions for searching and filtering the corpus. However, so far it lacks functionality to find graphic variants of words, which is crucial because it contains texts in Swiss German, which does not have a standardised orthography. For example, the query miir 'to me' will match both miir and mììr but not mir and other potential variants of the same word.
The project will address this problem by assigning a normalised form to every word in the corpus. The normalised form will be internally represented but will not be visible to users. Queries will be converted into the normalised form before the search is performed. Once the matching words are found in this representation, the results will be displayed in the original spelling.
Enriching a lexical database for the Swiss-German Sign Language with Annotations
ID: KLIP-2014-A-2
Researchers: Penny Boyes Braem (Center for Sign Language Research, Basel), Tobias Haug (Interkantonale Hochschule für Heilpädagogik), Sarah Ebling (Institute of Computational Linguistics)
Content:
The Center for Sign Language Research (Basel) has been involved inthe compilation of a lexical database for the Swiss-German Sign Language (DSGS) since 1996. Presently the database holds about 9000 entries, each of which contains information on semantics, form and usage of one sign. The database and the associated software are hosted on the ZüKL server since the SNF-Project InterGaze.
Currently there are machine-readable notations of the form of a sign in the Hamburg Notation System (HamNoSys) for about 3000 entries. These notations are required for standardised form comparisons and for automatic processing of sign language. This KLIP will hire two speakers of DSGS in order to add 1500 annotations in HamNoSys to the database and to evaluate existing annotations. This will contribute to enriching the database and making it more user-friendly.
Follow up project: SNF Sinergia "SMILE: Scalable Multimodal sign language Technology for sIgn language Learning and assessmEnt" (CRSII2_160811 / 1), Project start 01/2016
Database interface for a corpus of diaries and letters of the late 18th century (Lady Mary Hamilton Archive)
Since 2012 Researchers and students of the English Department have digitalized, transliterated and transcribed manuscripts of diary entries and letters from the late 18th century (Lady Mary Hamilton Archive) according to TEI Standards and in XML. The corpus comprises now more than 25,000 words. The KLIP funding will enable us to create a database interface in collaboration with the Institute of Computational Linguistics, which allows to search within the digitalized material (incremental corpus).The database will be accessible to external users because the material is not protected by copyright and needs no anonymisation. The KLIP project is completed by a SNSF grant for editions aiming at the digitalization of the complete material.
Alignment of Swiss German dialect recordings with existing transcriptions
ID: KLIP-2014-B-2
Researchers: Hanna Ruch (URPP Language and Space), Anne Göhring (Institute of Computational Linguistics)
Content:
The linguistic departments of UZH host a great number of corpora that consist of transcriptions of spoken speech. However, these transcriptions are frequently not time-aligned. Time alignment is desirable for several reasons: it is the base for research focussing on phonetic aspects of spoken language, it makes it possible to easily examine non-transcribed aspects of the signals such as prosody, and it is useful for evaluating transcriptions and making selected audio stretches available to the broader public.
The Munich Automatic Segmentation System (MAuS) is a tool that automatically aligns transcriptions and audio recordings. However, so far MAuS has been tailored to process short recordings with few speakers whose voices do not overlap. The aim of the present KLIP is to enhance MAuS in such a way that longer recordings taken in a natural setting can be handled, too. The results will be of interest for all linguists working with spoken language corpora in Zurich and elsewhere.
