Irina Temnikova
Due to the fact that emergency situations can lead to substantial losses, both financial and in terms of human lives, it is essential that texts used in a crisis situation be clearly understandable.
This thesis is concerned with the study of the complexity of the crisis management sub-language and with methods to produce new, clear texts and to rewrite pre-existing crisis management
documents which are too complex to be understood. By doing this, this interdisciplinary study makes several contributions to the crisis management field. First, it contributes to the knowledge of the complexity of the texts used in the domain, by analysing the presence of a set of written
language complexity issues derived from the psycholinguistic literature in a novel corpus of crisis management documents. Second, since the text complexity analysis shows that crisis management
documents indeed exhibit high numbers of text complexity issues, the thesis adapts to the English language controlled language writing guidelines which, when applied to the crisis management language, reduce its complexity and ambiguity, leading to clear text documents. Third, since low quality of communication can have fatal consequences in emergency situations, the proposed
controlled language guidelines and a set of texts which were re-written according to them are evaluated from multiple points of view. In order to achieve that, the thesis both applies existing evaluation approaches and develops new methods which are more appropriate for the task. These are used in two evaluation experiments – evaluation on extrinsic tasks and evaluation of users’
acceptability.
The evaluations on extrinsic tasks (evaluating the impact of the controlled language on text complexity, reading comprehension under stress, manual translation, and machine translation tasks) show a positive impact of the controlled language on simplified documents and thus ensure the quality of the resource. The evaluation of users’ acceptability contributes additional findings about
manual simplification and helps to determine directions for future implementation.
The thesis also gives insight into reading comprehension, machine translation, and cross-language adaptability, and provides original contributions to machine
translation, controlled languages, and
natural language generation evaluation techniques, which make it valuable for several scientific fields, including Linguistics, Psycholinguistics, and a number of different sub-fields of NLP.
Supervisors: Dr. Le An Ha , Prof. Dr. Galina Maneva , Dr. Anke Buttner, and Dr. Kevin Cohen
This thesis is concerned with the study of the complexity of the crisis management sub-language and with methods to produce new, clear texts and to rewrite pre-existing crisis management
documents which are too complex to be understood. By doing this, this interdisciplinary study makes several contributions to the crisis management field. First, it contributes to the knowledge of the complexity of the texts used in the domain, by analysing the presence of a set of written
language complexity issues derived from the psycholinguistic literature in a novel corpus of crisis management documents. Second, since the text complexity analysis shows that crisis management
documents indeed exhibit high numbers of text complexity issues, the thesis adapts to the English language controlled language writing guidelines which, when applied to the crisis management language, reduce its complexity and ambiguity, leading to clear text documents. Third, since low quality of communication can have fatal consequences in emergency situations, the proposed
controlled language guidelines and a set of texts which were re-written according to them are evaluated from multiple points of view. In order to achieve that, the thesis both applies existing evaluation approaches and develops new methods which are more appropriate for the task. These are used in two evaluation experiments – evaluation on extrinsic tasks and evaluation of users’
acceptability.
The evaluations on extrinsic tasks (evaluating the impact of the controlled language on text complexity, reading comprehension under stress, manual translation, and machine translation tasks) show a positive impact of the controlled language on simplified documents and thus ensure the quality of the resource. The evaluation of users’ acceptability contributes additional findings about
manual simplification and helps to determine directions for future implementation.
The thesis also gives insight into reading comprehension, machine translation, and cross-language adaptability, and provides original contributions to machine
translation, controlled languages, and
natural language generation evaluation techniques, which make it valuable for several scientific fields, including Linguistics, Psycholinguistics, and a number of different sub-fields of NLP.
Supervisors: Dr. Le An Ha , Prof. Dr. Galina Maneva , Dr. Anke Buttner, and Dr. Kevin Cohen
less
Related Authors
InterestsView All (15)
Uploads
Papers by Irina Temnikova
The terms have been collected from a seed set of terms manually annotated by a linguist and an emergency manager from tweets broadcast during 4 crisis events. A Conditional Random Fields (CRF) method was then applied to tweets from 35 crisis events, in order to expand the set of terms while overcoming the difficulty of getting more emergency managers’ annotations.
The terms are classified into 23 information-specific categories, by using a combination of expert annotations and crowdsourcing. This article presents the detailed terminology extraction methodology, as well as final results.
Emergency management specialists, crisis response practitioners, and scholars have long recognized that clear communication is essential during crises. To the best of our knowledge, the work we present here is the first to study the readability of crisis communications posted on Twitter—by governments, non-governmental organizations, and mainstream media. The data we analyze is comprised of hundreds of tweets posted during 15 different crises in English-speaking countries, which happened between 2012 and 2013. We describe factors which negatively affect comprehension, and consider how understanding can be improved.
Based on our analysis and observations, we conclude with several recommendations for how to write brief crisis messages on social media that are clear and easy to understand.
The terms have been collected from a seed set of terms manually annotated by a linguist and an emergency manager from tweets broadcast during 4 crisis events. A Conditional Random Fields (CRF) method was then applied to tweets from 35 crisis events, in order to expand the set of terms while overcoming the difficulty of getting more emergency managers’ annotations.
The terms are classified into 23 information-specific categories, by using a combination of expert annotations and crowdsourcing. This article presents the detailed terminology extraction methodology, as well as final results.
Emergency management specialists, crisis response practitioners, and scholars have long recognized that clear communication is essential during crises. To the best of our knowledge, the work we present here is the first to study the readability of crisis communications posted on Twitter—by governments, non-governmental organizations, and mainstream media. The data we analyze is comprised of hundreds of tweets posted during 15 different crises in English-speaking countries, which happened between 2012 and 2013. We describe factors which negatively affect comprehension, and consider how understanding can be improved.
Based on our analysis and observations, we conclude with several recommendations for how to write brief crisis messages on social media that are clear and easy to understand.