ABSTRACT
The state-of-the-art systems for most natural language engineering tasks employ machine learning methods. Despite the improved performances of these systems, there is a lack of established methods for assessing the quality of their predictions. This work introduces a method for explaining the predictions of any sequence-based natural language processing (NLP) task implemented with any model, neural or non-neural. Our method named EXSEQREG introduces the concept of region that links the prediction and features that are potentially important for the model. A region is a list of positions in the input sentence associated with a single prediction. Many NLP tasks are compatible with the proposed explanation method as regions can be formed according to the nature of the task. The method models the prediction probability differences that are induced by careful removal of features used by the model. The output of the method is a list of importance values. Each value signifies the impact of the corresponding feature on the prediction. The proposed method is demonstrated with a neural network based named entity recognition (NER) tagger using Turkish and Finnish datasets. A qualitative analysis of the explanations is presented. The results are validated with a procedure based on the mutual information score of each feature. We show that this method produces reasonable explanations and may be used for i) assessing the degree of the contribution of features regarding a specific prediction of the model, ii) exploring the features that played a significant role for a trained model when analyzed across the corpus.
Subject(s)
Natural Language Processing , Machine Learning , SoftwareABSTRACT
BACKGROUND: Coordination of an emergency response team is an important determinant of prompt treatment for combat injuries in hospitals. The authors hypothesized that instant messaging applications for smartphones could be appropriate tools for notifying emergency response team members. The objective of this study was to investigate the efficiency of a commercial instant messaging application (WhatsApp, Mountain View, CA) as a communication tool for the emergency team in a level-I trauma center. METHODS: We retrospectively evaluated the messages in the instant messaging application group that was formed to coordinate responses to patients who suffered from combat injuries and who were transported to our hospital via helicopter during an 8-week period. We evaluated the response times, response time periods during or outside of work hours, and the differences in the response times of doctors, nurses, and technicians among the members of the emergency team to the team leader's initial message about the patients. RESULTS: A total of 510 emergency call messages pertaining to 17 combat injury emergency cases were logged. The median time of emergency response was 4.1 minutes, 6 minutes, and 5.3 minutes for doctors, nurses, and the other team members, respectively. The differences in these response times between the groups were statistically significant (p=0.03), with subgroup analyses revealing significant differences between doctors and nurses (p=0.038). However, no statistically significant differences were observed between the doctors and the technicians (p=0.19) or the nurses and the technicians (p=1.0). From the team leader's perspective, using this application reduced the workload and the time loss, and also encouraged the team. CONCLUSION: Instant messaging applications for smartphones can be efficient, easy-to-operate, and time-saving communication tools in the transfer of medical information and the coordination of emergency response team members in hospitals.