Nada

Integrated language tools for writing and document handling

Head of project:
Kerstin Severinson Eklundh, IPLab, Nada, KTH

Project group at Nada, KTH:
Viggo Kann, head of group
Johan Carlberger
Rickard Domeij
Ola Knutsson
Stefan Larsson
Annika Hansen-Eriksson
More information: Grammar checking (in Swedish)

Project group at Department of Linguistics, Gothenburg:
Robin Cooper , head of group
Sylvana Sofkova Hashemi
More information: Finite state grammar for finding grammatical errors in Swedish text.

Project group at the Swedish Language Council (Svenska Språknämnden):
Margareta Westman, head of group
Ola Karlsson
 
 The project is sponsored by the HSFR/NUTEK Language Technology programme .
 

Project overview

The aim of this project is to develop and integrate different language tools in an environment for writing support. The tools will include functions for grammar checking, linguistic search/editing, and help systems with language rules and facilities for sharing texts and comments over the web. The project builds on previous work on a grammar checking prototype called Granska (Domeij, Knutsson, Larsson, Rex & Severinson Eklundh, 1998) and on a spelling error detection and correction program called Stava (Kann, Domeij, Hollman & Tillenius, 1998; you can use Stava with a web interface). All functions in the new writing environment will have access to information about word classes and inflections of the words in a text and can use this linguistic information for more advanced processing, such as analysis of grammatical errors or search for linguistic objects.

The work in the project will focus on the following research areas which will be introduced below:
1. Grammar checking and proof reading.
2. Linguistic editing functions.
3. Language rules and help system.
4. Interface design and implementation.
 

Presentation of the project

The project presented itself (in Swedish) in the afternoon of 22 March 1999. More information can be found here.

Work shop on language technology

The project will arrange a work shop at KTH December 3-4 1999 in cooperation with Nordiska språkrådet. More information can be found here.

Grammar checking and proof reading

The Granska environment will be improved to give room for new checking functionality as well as improved efficiency. Combinations of rule-based and statistical methods will be explored and used for  word class disambiguation and error detection. In collaboration with the Department of Linguistics at the University of Gothenburg we intend to explore finite state methods for finding grammatical errors. Among the many challenging problems in Swedish grammar checking, we will focus on noun phrase agreement errors and split compounds. The efficient spell checker Stava, developed at Nada, will be integrated in the grammar checking function.
 

Linguistic search and editing functions

The linguistic information in Granska makes it possible to introcuce new linguistic functions which are of interest to writers as well as language learners. These include linguistic search, i.e. searching for linguistic units rather than strings of characters. For example, a writer may need to locate all verbs in a text in order to consider the tense choice, and possibly change a verb to present instead of past tense. The latter is an example of linguistic editing functions, which use the linguistic structure of the text to provide powerful tools for revision.

Within the project a linguistic search function will be implemented. The design and use of linguistic editing functions will be considered and explored.
 

Language rules and help system

It is of primary importance for writers to be able to locate language resources and help information without leaving the current task. As a collaboration between IPLab and the Swedish Language Council (Svenska språknämnden), we plan to create a working help system for writers, to be included in a networked environment. A particularly important part of the project is to make the well-known style guide Svenska skrivregler widely accessible in electronic form in WWW environments. The style guide will also be linked to the grammar checking function in Granska. The work will take advantage of existing work at IPLab and CID on designing web based style guides and shared help systems.
 

Interface design and implementation

The research profile at IPLab makes it natural to focus specifically on the interface of the writing environment, and how various design choices affect the writing process. The project will include empirical studies of the Granska prototype with users, in order to explore how the dialogue and system explanations should be designed with respect to different users' linguistic knowledge. Special focus will be placed on how the users can adapt the system to their purposes, and take active part in the linguistic analysis of the text when necessary. Another important issue is how to integrate the different system functions into a unified and flexible writing environment, preferably connected to the web.
 

Selected publications

Carlberger, J. & Kann, V. (1998) Implementing an efficient part-of-speech tagger. Software Practice and Experience 29, 815-832, 1999.  Postscript, PDF

Domeij, R.,  Knutsson, O., Larsson, S.,  Rex, Å., Severinson Eklundh, K. (1998) Granskaprojektet 1996-1997. IPLab-rapport. KTH, Nada, IPLab.

Kann, V., Domeij, R., Hollman, J., Tillenius, M. (1998) Implementation aspects and applications of a spelling correction algorithm. NADA report TRITA-NA-9813.  Postscript  PDF

Larsson, S. (1998). Interaktivitet och användbarhet vid datorstödd språkgranskning och
redigering i en integrerad skrivmiljö. Exjobbsrapport TRITA-NA-E9833 (IPLab-150), Nada.
 

^ Upp till Nadas ingångssida.


Sidansvarig: Rickard Domeij <domeij@nada.kth.se>
Senast ändrad 18 september 1999
Tekniskt stöd: <webmaster@nada.kth.se>