Till startsida
Sitemap

Current Ph.D. Students

 

Jenny Brusk

Department of Philosophy, Linguistics, and Theory of Science, University of Gothenburg
web: http://mainweb.hgo.se/

Academic background: M.A. in Computational Linguistics
Thesis Description:

The aim of this PhD project is to develop a new method for managing game dialogues using statecharts, "a visual
formalism for complex systems" (David Harel, 1987). The dialogue manager will be implemented using SCXML,
i.e. a new W3C standard which combines XML syntax with statechart semantics. Statecharts are in fact an
abstraction of finite state machines, allowing hierarchial as well as orthogonal states. It is thus possible to allow
simultaneous processes to run in parallel, which can be exemplified by a situation in which a game character picks
up an object while talking. Within the Synergy Project (Torbjörn Lager, Fredrik Kronlid and myself) we are
investigating how statecharts and SCXML can be used in dialogue system design as well as for game programming.
Dialogues in games are usually used for providing the player with background story, quests and motivations for
performing certain actions in the game, but also as a way to get to know the characters in the game. Most dialogues
in games are however canned and thus limited in expressivity and variation. One incentive for this research is to
allow more flexible dialogues, through natural language conversations.
A game dialogue manager should for example support
• Story construction
• Characterisation and character development
• Game progression
• Social interaction
 

Håkan Burden

Department of Computer Science, CTH
web:

Academic background: MA Computational Linguistics,University of Gothenburg, 2006
Thesis Description:

For the licentiate I will implement a large grammar for Swedish. The work consists of two major parts: 1)
Developing lexica with suffuicient information. Today there are no open-source Swedish lexica with syntactic
information such as which arguments a certain verb takes. Another approach is to take available lexica and find
ways of using them together with Grammatical Framework. 2) Extend the existing set of grammar rules in the
Swedish resource grammar given by Grammatical Framework. For the PhD I will then use the large grammar in
order to explore different approaches for robust parsing.

Karin Cavallin

Department of Philosophi, Linguistics, and Theory of Science, University of Gothenburg
web: http://www.flov.gu.se/om/personal/karin_cavallin/

Academic background: M.A. in Computational Linguistics, University of Gothenburg, 2003


Thesis Description:

My thesis aims at looking at how far you can get with a lexical semantic analysis if we use simpler, and more
linguistically motivated methods (Lexical Sets) than for instance LSI.
From Lexical Sets we will build an embryo of a generative lexicon (Pustejovsky) and from this look at how to
handle innovations, how to motivate lexical updates, alternativley deal with coercion.
In my thesis I will probably also explore how to relate Lexical Sets to the construction of a Swedish FrameNet.
There will also be a survey of the appropriateness of different statistical measures in constructing lexical sets.
As a starting point I will look at the verb ”läsa” (’read’) and its lexical set, and also at the lexical sets the nominal
objects with ”läsa” give rise to.

Dana Dannélls

The Department of Swedish Language, University of Gothenburg
web: http://www.svenska.gu.se/om-oss/personal/dana-dannells/

Academic background: Master of Arts in Computational Linguistics
Thesis Description:

The aim of this research is to clarify how existing language technology tools and methods can be utilized in order to adapt documents for a particular category of readers, with an emphasis on the adaptation for lay readers of documents originally produced by professionals for professionals.

This PhD project will explore methodologies that reduce linguistic complexity in syntax, vocabulary and rhetorical text structure. More specifically, this project will focus on empirical evaluation of approaches for generating simple and understandable documents for different categories of (non-professional) readers from professional text in an arbitrary domain. This involves investigation of a spectrum of language technology and information access methods, such as text understanding, text segmentation, topic detection, natural language generation and summarization techniques. We will identify and suggest optimal ways in which such natural language processing techniques can be brought to bear upon the problem of adapting the presentation of text content to a specific readership.

Automatic simplification of technical and other kinds of documents has been studied to some extent, but almost exclusively for English. In this research we are planning to study whether the proposed approaches are applicable to Swedish.

Grégoire Détrez

Department of Computer Science, CTH

Susanne Ekeklint

School of Mathematics and Systems Engineering, Växjö University
web: http://w3.msi.vxu.se/users/sek/

Academic background: M.A. in Computational Linguistics, University of Gothenburg, 2001
Thesis Description:

In general, my research interest lies in automatic methods for semantic analysis of text, and the machine learning methods that are suitable for such tasks. My interest in such tasks stems from my previous work on automatic illustration of narrative texts with multiple events, in particular physical events such as traffic accidents. My research group identified two pivot problems for automatic analysis of such texts: predicate-argument structures (“who does what to whom”) and temporal structure (i.e. the ordering of the events in time). Primarily, my own research has focused on the first of these; in particular, I have constructed automatic systems for frame-semantic analysis of predicate-argument relations according to FrameNet. I have also participated in projects that addressed the problem of temporal structure analysis.

Since the automatic systems that have been described in literature all rely (for training of statistical models) on large quantities of annotated data which are not available for small languages such as Swedish, I have also studied and created methods for automatic annotation of corpora, in particular by means of structure transfer from English via parallel texts. My licentiate thesis described a FrameNet-based semantic analyzer for Swedish, which was trained on automatically transferred structures. My current research focuses primarily on the following key problems:

  1. New methods and improved mathematical models for transfer of semantic structures from one language to another.
  2. Methods to measure and improve the quality of the automatically produced data, preferably methods involving a minimal amount of manual labor.
  3. 3. New methods and models for automatic semantic analysis of text, such as predicate-argument structures.

There are several interesting (and partly unexplored) machine learning aspects of the problems at hand, which are important to my work (in particular for the third point above):

  1. Prediction of complex structures, where there are strong dependencies between the predicted variables.
  2. Machine learning with incomplete feedback.
  3. Machine learning with noisy training data, where the rate of corruption differs between sections of the training data.

 

Ann-Marie Eklund

The Department of Swedish Language, University of Gothenburg

Johan Eklund

Swedish School of Library and Information Sciences, University College of Borås
web: http://www.adm.hb.se/~JEK/index_en.htm

Academic background: Master’s degree in Library and Information Science
Thesis Description:

Classification is a central activitity in libraries. The available information resources need to be categorized according to structured schemes, such as classification systems, to be accessible to the libraries’ users. However, classification is a time-consuming intellectual task and it is an increasingly urgent issue to find useful methods for (at least partly) automating this process. The major research focus of my thesis project is to investigate to what extent one of the most successful classification algorithms, the support vector machine (SVM), can induce a useful classification model given the limited data available in typical library catalogs. The records in the collections available for this project are manually classified according to the SAB classification scheme, which will be used as a point of comparison, and enriched with value-adding data such as reviews and tables of contents. Three major research questions will be investigated: Viewed as a nonlinear optimization problem, how can the training efficiency of SVM for this particular application area be maximized? Which method for multi-class classification gives the overall best performance? Is there any significant difference in classification performance between mono- and multi-disciplinary subject areas?

Laura Enflo

Department of Speech, Music and Hearing, KTH Master of Science in Engineering, Degree Programme in Electrical Engineering
web:

Academic background: Master of Science in Engineering


Thesis Description:

Multimodal Speech Synthesis
The main part of the thesis work is made in projects, currently a project about the voice source in HMM synthesis and previously in the project SIMULEKT.
The voice source project aims to implement a voice source in a way that it can be used for improving the voice quality and the controlling of voice specific parameters in HMM speech synthesis.
The primary goal of the SIMULEKT research project has been to produce more precise and thorough knowledge about phrase and utterance prosody in some major regional varieties of Swedish. The main project work has consisted of the analysis and description of this material regarding prosody as well as the simulation of prosodic patterns by means of speech synthesis. The accent types have been supposed to be geographically distributed: South, Göta, Svea, Gotland, Dala, North, and Finland Swedish. The significance of this project work has been both within basic research and within speech technology applications.

Jody Foo

Department of Computer and Information Science, Linköping University
web: http://www.ida.liu.se/~jodfo/

Academic background: Fil. Mag. in Cognitive Science
Thesis Description:Terminology work is mainly a manual task that involves time consuming research and investigations of concepts and linguistic usage. Because of the resources and effort required to build large terminologies, large terminologies
are rare today. Furthermore creating bilingually or multilingually harmonized terminologies requires careful analysis
of concepts in two or more languages. My research aims to understand how computational methods and tools can be
used to aid terminologists in their work. In particular, I will be researching methods that can improve mono- and
bilingual term extraction, as well as data driven mono- and bilingual concept analysis.

Eva Forsbom

Derpartment of Linguistics and Philology, Uppsala University
web: http://stp.ling.uu.se/~evafo/

Academic background: B.A. in English and Swedish, Stockholm University; University Diploma in Language Consultancy, Stockholm University
Thesis Description:

The purpose of the thesis is to investigate if and how some automated textlinguistic methods can give more relevant hits in information retrieval, and give coherent summaries that are more query and user adapted than those usually given in information systems.

A lexical cohesion analysis is used as a basis for indexing, searching and a short summary in an information system. The analysis is based on a number of knowledge bases containing linguistic or world knowledge, and the result will mainly depend on what knowledge is available.

By combining the lexical cohesion analysis with a Rhetorical Structure Theory analysis, it should be possible to come to terms with some coherence problems in summaries only based on lexical cohesion analysis. At the same time, the less computationally costly lexical cohesion analysis could reduce the number of possible RST analyses, since it also gives an estimate on how closely sentences are related.

 

Ebba Gustavii

Department of Linguistics and Philology, Uppsala University
web: http://stp.ling.uu.se/~ebbag/

Academic background: M.A. in Language Engineering
Thesis Description:

Several aspects of compounds make them particularly difficult to handle in a system performing automatic translation, being it a machine translation system or a system for cross-language information retrieval. The relation between the parts of a compound is implicit, and thus its interpretation is never wholly compositional. This is, in particular, a problem when translating from a language making frequent use of compounds, to a language generally preferring syntactic constructions instead, since an overt syntactic marker is then to be generated. Related problems concern the choice of morhposyntactic features of the target constuction and its integration into the target sentence. In this thesis I will look into the methods for translating compounds suggested sofar, and look for ways to improve them. In particular I will focus on corpus-based methods (as opposed to interpretation-based methods).

Christian Hardmeier

Department of Linguistics and Philology, Uppsala University

Aron Henriksson

Department of Computer and System Sciences, Stockholm University

web: http://aron.blogs.dsv.su.se/

Academic background: M.Sc. in Engineering and Management of Information Systems

Thesis Description: Clinical Text Mining Group, which primarily conducts research in the field of language technology/natural language processing in the medical domain (electronic patient records).

 

Katarina Heimann Mülenbock

The Department of Swedish Language, University of Gothenburg

Academic background: M.A. in computational linguistics

Thesis Description: Texts can be assigned different degrees of readability, depending on the textual units studied. Swedish readability
research, grading texts into a readability index, has generally used characters and words as units of study. I want to
find other parameters that can indicate whether a text can be considered as appropriate for readers belonging to a
specific target group, namely persons with mild intellectual disabilities. A corpus, consisting of authentic easy-toread
fiction, news and information texts and children’s books will be used as material. A computer-based application
will be implemented, where the linguistic parameters will be set in order to react if certain rules are ignored and
suggest simplifications on lexical, syntactic and/or semantic level. Another objective of the thesis project is to
design and implement a tool, providing symbol support to the same texts for persons with more severe language
impairments, i.e. users of AAC (augmentative and alternative communication).

Cecilia Hemming

Department of Language, University of Skövde

Academic background: B.A. French and Linguistics, Skövde
Thesis Description:

Technical word formation in both French and Swedish is facilitated by using affixation, compounding and combinations of both. There are production patterns though that seem to be language specific. Contrastive studies can reveal what similarities, divergences and relations there are in the two languages. The result of this linguistic analysis can be used in the development of applications for Machine Translation or Information Extraction.

Maria Holmqvist

Department of Computer and Information Science, Linköping University
web: http://www.ida.liu.se/~marho/

Academic background: MS in Cognitive Science
Thesis Description:

The thesis topic is word alignment and extraction of translation information from parallel text. Word alignment is the task of finding corresponding words in a text and its translation. Word alignments are a prerequisite for many NLP tasks, such as the creation of bilingual dictionaries and term banks, machine translation applications and multilingual word sense disambiguation, but they can also be a resource for researchers in contrastive linguistics and translation studies. In my thesis I will examine methods that reuse longer word aligned text segments in alignment of new text. By reusing longer aligned text segments, new word alignments can be created automatically with very high accuracy although recall will not be as high as when using statistical methods for word alignment. Since manual word alignment is time-consuming and difficult, automatically produced word alignments of high quality could be used to speed up manual word alignment of parallel treebanks or other types of word aligned reference data. The thesis will present experiments where the method is applied to translated text from varying text domains, for example, EU proceedings, software manuals and fiction.

Taraka Rama Kasicheyanula

The Department of Swedish Language, University of Gothenburg

Monica Lassi

Swedish School of Library and Information Sciences University College of Borås
web: http://etjanst.hb.se/hb/forskning/forskare.asp?forskareId=mol

Academic background: M.Sc. Library and Information Science
Thesis Description:

The goal of the thesis is to find out which socio-technical factors that are important for Library and Information Science (LIS) researchers concerning adoption and use of a collaboratory for data collection instruments used within LIS research. I am interested in finding out how LIS researchers choose and evaluate data collection instruments today, and how that process can be supported by a collaboratory by enabling sharing and use data collection instruments through a new medium. Another component of this thesis is to find out what LIS researchers prefer considering the representation of the data collection instruments by metadata in general and a controlled vocabulary in particular. A controlled vocabulary is used to represent the content aspects of a document, what is called a subject representation, in order to provide access points to documents through what they are about. In interdisciplinary areas, such as LIS, there is no single vocabulary that incorporates all concepts of the discipline - terms are used in different ways, and one concept may be denoted by many different terms. The main research question is "Which sociotechnical factors affect the design and development and the adoption and use of a collaboratory for data collection instruments within Library and Information Science?" The main themes include collaboratories, data collection instruments, knowledge architecture (metadata and vocabularies) and Library and Information Science. A grounded theory approach will be used, data including interviews, literature reviews, discipline analysis and a user study evaluating a prototype of a collaboratory for data collection instruments. The dissertation's relation to Language Technology (LT) is at least twofold: firstly, the focus on vocabularies and metadata are connected to semantics, in the case of the dissertation the meaning of concepts of an interdisciplinary discipline. In LT, semantic relations are represented by e.g. ontologies and thesauri, which would in LIS be classified as types of controlled vocabularies. Second, the attempt to enhance scholarly communication in a system for collaboration is connected to Computer Mediated Communication, a field that e.g. studies how language is used in online environments.

Jonas Lindh

Department of Philosophi, Linguistics, and Theory of Science, University of Gothenburg
web: http://www.flov.gu.se/om/personal/jonas_lindh/

Academic background: Bachelor of Arts in Linguistics/Phonetics
Thesis Description: http://www.ling.gu.se/~jonas/Thesis.html

Current working title: A Theoretical Basis and Framework for Methods and Measures in Forensic Speaker Identification.

Kristina Lundholm Fors

Department of Philosophi, Linguistics, and Theory of Science, University of Gothenburg

web: http://www.flov.gu.se/om/personal/kristina_lundholm_fors/ 

 

Raveesh Meena 

Department of Speech, Music and Hearing, KTH

 

Jenny Myrendal

Department of Philosophy, Linguistics and
Theory of Science, University of Gothenburg

Academic background: Master’s degree in Computational Linguistics, University of Gothenburg

Thesis Description:Computer-Mediated Text-Based Dialogues on the Web: Communication and Collaborative Meaning Making
in Social Media and Other Web-Based Environments

Web-based technologies such as social media software, blogs, online discussion groups and communities all offer
and afford different technical conditions and opportunities for communication, but they all have in common that
they rely on text-based written interactions between participants as basis for communication. This thesis aims to
develop knowledge about this text-based communication, in which people through written contributions in digital
dialogues exchange opinions, cooperate and engage in collaborative meaning making. The project will study
different kinds of web-based dialogues and analyse what traits can be said to be typical characteristics of such
dialogues.
The study will employ linguistic methods, such as discourse analysis and methods for conversational analysis, to
code and analyse the functions of the different dialogue contributions and the relations between them, to reveal
discourse patterns and distinctive features, for example pertaining to:

  •  Addressivity, i.e. how a participant addresses a dialogue contribution to another or several other participants.

  • Intertextuality, i.e. how the different dialogue contributions relate to each other and bear traces of each other, for example by referring or alluding to previous contributions.

  • Intersubjectivity, i.e. to what extent the different dialogue contributions share common assumptions and perspectives.

  • Grounding, i.e. how participants in dialogue agree on what is common ground (mutually understood).

Primarily, the study will focus on asynchronous computer-mediated communication, and examine dialogues that
take place over an extended period of time, taken from various web-based contexts dealing with different topics of
conversation. The study will be divided into smaller sub-studies, where data from dialogues from diverse contexts
will be collected, for example from open web-based discussion forums and communities, active blogs where current
topics are being discussed, and perhaps also from closed web-based environments such as virtual learning
environments and formal learning platforms.
The collected dialogue samples will then be analysed and contrasted with each other in order to shed light on how
the functions and characteristics of these kinds of dialogue can be described and understood. The study will also try
to draw conclusions about how the characteristics of such web-based interaction influence how the dialogue
develops with regard to in-depth reasoning, collaborative meaning making and learning.
 

Eva Pettersson 

Department of Linguistics and Philology, Uppsala University

 

Mojgan Seraji

Department of Linguistics and Philology, Uppsala University

Academic background: M.A. in Computational Linguistics
Thesis Description:

My research project is a corpus-based contrastive grammatical analysis of Persian and Swedish. More specifically, it is a detailed syntactic analysis of noun phrases in Persian and Swedish. As a basis for the corpus-based analysis, I will also develop a parallel corpus of Persian and Swedish texts. The results of this analysis will be relevant for machine translation.

Alexander Simonsson

Department of Speech, Music and Hearing, KTH

Maria Skeppstedt

Department of Computer and System Sciences, Stockholm University

web: http://people.dsv.su.se/~mariask/

Academic background: M.Sc. and Ph. Licentiate in Computer Science

Thesis Description:Text mining in Swedish Health Records

Mustapha Skhiri

Department of Computer and Information Science, Linköping University

Academic background: M.Sc. Computer Science
Thesis Description:

Dialogue is an interactive communication of information mainly based on speech, but also visual information such as gesture, facial expression and head movement clearly makes the conversations much smoother and more natural. In my research I will first do an examination of the face and head movements in turntaking situations and also study how these movements are related to the speech signal in human-human and human-computer situation. Then I will make a computational model and implement these movements in a Òtalking headÓ and study the perceptual relevance of this in a dialog system that includes visual information.

Sofia Strömbergsson

Department of Speech, Music and Hearing, KTH

Academic background: M.Sc. in Language technology, Uppsala university, 2000
Thesis Description:

Phonological impairment is characterised by speech production deviations, but often also by deviant auditory perception of speech sounds. Phonological intervention often focuses on making the child aware of functional linguistic differences between speech sounds through listening and speech production exercises. As of today, there are few computer applications (e.g. SpeechViewer, Box of tricks) that are used in pronunciation and articulation exercises clinically, and the ones that exist are primarily designed to be used in the intervention of speech production deviations other than phonological impairments. Hence, these applications often focus on the motor aspect of articulation, and on the link between articulatory movements and auditory feedback, and rarely include exercises targeting the conceptual level of phonology. This research aims at finding answers to what elements in traditional phonological intervention that are possible -and that make sense -to implement computationally, and to find how speech technology can be used in developing new methods for phonologic intervention. One possibility in this direction would be "corrected re-synthesis"of the speech sounds that are difficult for the child -how would a child react to hearing his/her own voice produce phonological distinctions that the child himself/herself don’t produce? For example, how would a child who velarises all dental stops react to hearing a synthesised version of his/her own voice pronouncing the word "dad" correctly, and not /gæg/ as he/she herself would? Visual feedback has proven useful in computer-assisted speech training (Öster, 2006), and the idea of introducing visual feedback into computer-assisted phonological intervention is appealing, since it is expected to provide additional amplification of phonological distinctions.

Sara Stymne

Department of Computer and Information Science, Linköping University
web: http://www.ida.liu.se/~sarst

Academic background: M.A. in Cognitive Science
Thesis Description:

Licentiate thesis title: Compound processing for phrase-based statistical machine translation.

The focus of the licentiate thesis is processing of German and Swedish compounds for statistical machine
translation.

The overall focus on my work is on pre- and postprocessing for statistical machine translation between English and
the Germanic languages, which have more complex morphology than English.

Håkan Sundblad

Department of Computer and Information Science, Linköping University
web: http://www.ida.liu.se/~hakjo/

Academic background: M.A. in Cognitive Science
Thesis Description:

In recent years, there has been a surge in interest surrounding question answering (QA) systems and technologies. The perhaps the most exciting features of QA systems are that they work against open domains and unstructured information. So far, QA systems have not been designed to support dialogic features, such as posing clarification questions. Regarding these issues, results from research on dialogue systems can contribute to a large extent. Traditional dialogue systems, however, have been designed for small, rather restricted domains, and often make use of a structured information source, such as a database. This thesis investigates how we can combine the features of QA systems with the features of dialogue systems. More specifically, the thesis investigates how the question analysis must be modified in order to accommodate the new features that emerge. The thesis proposes a question taxonomy, an answer taxonomy, and a model for automatic content analysis.

Oscar Täckström

Department of Linguistics and Philology, Uppsala University
web: http://www.sics.se/people/oscar

Academic background: MSc
Thesis Description:

Supervised learning of semantic roles has been shown to be effective for verbal predicates and acceptable fornominal predicates. However, in order to get good results, large amounts of annotated data is required. Semi-supervised learning methods could potentially drastically reduce the need for annotated data, allowing semantic rolelabeling systems to be developed for the many languages for which corpora annotated with semantic roles do notexist. I aim to develop semi-supervised methods for learning of semantic roles. I further aim to minimize the need for pre-processing (e.g. by syntactic parsing and part-of-speech tagging) to a minimum.

Marcus Uneson

Department of Linguistics, Lund University
web: http://www.ling.lu.se/persons/MarcusUneson.html

Academic background: MA (Phonetics)
Thesis Description:

My areas of interest fall within computational phonology, dealing with questions such as: Given a pair of words
represented as strings in some phonetic (or at least alphabetic) transcription, how different are they? Given two
languages represented as a set of such pairs, how different are they? How can one transform one member of a pair
into the other? Borrowing methods from computational biology, I'd like to explore ways of inducing such rules
automatically.
Automatically learned phonological rules may be of interest for instance for pronunciation modelling and historical
linguistics. Similarly, phonetically or phonologically motivated string comparison measures have applications in
very diverse areas, including dialectometry, speech technology, historical linguistics, and identification of
confusable drug names.
 

Shafqat Mumtaz Virk

Department of Applied Information Technology, University of Gothenburg

Per Weijnitz

Department of Linguistics and Philology, Uppsala University
web: http://stp.ling.uu.se/~perweij/

Academic background: M.A. of Philosophy in Language Engineering
Thesis Description:

The purpose of this work is to explore how far a rule-based system can be improved, with the main focus on
coverage and robustness. Each candidate system implements a different approach to machine translation, with a
different balance between rule-based and statistically based. The main focus will be on a rule-based system inspired
by linguistic tradition with hybrid offsprings which borrow techniques from other machine translation paradigms,
and the baseline is a strictly rule-based system. The translation quality will be measured and evaluated using both
manual and automatic methods. Manual methods focus on concepts such as adequacy and acceptance, whereas
automatic methods define various kinds of similarity between translated text and one or more reference translations.

 

Jessica Villing

Department of Philosophi, Linguistics, and Theory of Science, University of Gothenburg
web: http://www.flov.gu.se/om/personal/jessica_villing/

Academic background: M.A. Computational Linguistics 2005
Thesis Description:

The aim of the thesis is to investigate methods for improving human-machine interaction in spoken in-vehicle
dialogue systems. The in-vehicle environment poses heavy demands on the usability of the dialogue system since
the user has to pay full attention to something other than the system, namely the traffic situation. The main focus
will be on the cognitive workload of the driver. I will investigate if it is possible to decrease the cognitive load by
changing dialogue management strategies, and study the interruption and resumption behaviour of dialogue partners
and use this as a basis for implementation. 

Roberg Östling

Department of Linguistics, Stockholm University

Academic background: M.Sc. in Computer Science and Engineering, Royal Institute of Technology

Thesis Description:Extraction of constructions from Swedish text

From a foundation of construction grammar, the goal of the project is to investigate methods for automatic
extraction of constructions from Swedish text, for comparative analysis. While the core of the project is theoretical,
possible applications include language checking tools, text categorization and genre detection.
The first part of the project is to adapt a suitable formalization of the sometimes vague construction grammar school
of thought.
The second part of the project is to develop practical methods to extract actual construction data from a text corpus,
and to analyze the use of constructions in a specific text.
The third and final part of the project is to apply these methods in some area, such as automatic language checking.

 

To the top