Search Articles

View query in Help articles search
e.g. mhealth

Search Results (1 to 4 of 4 Results)

Download search results: CSV END BibTex RIS

Developing an ICD-10 Coding Assistant: Pilot Study Using RoBERTa and GPT-4 for Term Extraction and Description-Based Code Selection

Developing an ICD-10 Coding Assistant: Pilot Study Using RoBERTa and GPT-4 for Term Extraction and Description-Based Code Selection

Encoder-only models, such as the BERT (Bidirectional Encoder Representations from Transformers) architecture and its variant, Ro BERTa (Robustly Optimized BERT Pretraining Approach) [15], excel at capturing relevant information from the input data and constructing meaningful representations [16-18] and can be found in current SOTA models in ICD coding [5,7]. Conversely, decoder models, exemplified by GPT, are specialized in generating coherent and contextually relevant text [19].

Sander Puts, Catharina M L Zegers, Andre Dekker, Iñigo Bermejo

JMIR Form Res 2025;9:e60095

Download Citation: END BibTex RIS

Download PDF Download XML
Article has an altmetric score of 1

Chinese Clinical Named Entity Recognition With Segmentation Synonym Sentence Synthesis Mechanism: Algorithm Development and Validation

Chinese Clinical Named Entity Recognition With Segmentation Synonym Sentence Synthesis Mechanism: Algorithm Development and Validation

To demonstrate the effectiveness of the algorithm, we constructed four models: (1) SSSS + BERT + CRF, (2) SSSS + BERT + Bi LSTM + CRF, (3) SSSS + Ro BERTa + CRF, and (4) SSSS + Ro BERTa + Bi LSTM + CRF. These were compared with BERT + CRF (baseline 1) and BERT + Bi LSTM + CRF (baseline 2). To investigate the impact of SSSS on Ro BERTa, we also performed an ablation study on the Ro BERTa + CRF and Ro BERTa + Bi LSTM + CRF models. The results for CCKS-2017 and CCKS-2019 are presented in Tables 7 and 8.

Jian Tang, Zikun Huang, Hongzhen Xu, Hao Zhang, Hailing Huang, Minqiong Tang, Pengsheng Luo, Dong Qin

JMIR Med Inform 2024;12:e60334

Download Citation: END BibTex RIS

Download PDF Download XML
Article has an altmetric score of 1

Classification of Patients’ Judgments of Their Physicians in Web-Based Written Reviews Using Natural Language Processing: Algorithm Development and Validation

Classification of Patients’ Judgments of Their Physicians in Web-Based Written Reviews Using Natural Language Processing: Algorithm Development and Validation

As part of our final sample of 345,053 reviews, the 2000 hand-coded reviews from our training data set were recoded by Ro BERTa. The interrater reliability between our hand-coding and Ro BERTa was Cohen κ =0.96 for both interpersonal manner and technical competence. Comparing the Ro BERTa codes with the original hand codes for these reviews, we found only 107 (5.4%) reviews had coding discrepancies.

Farrah Madanay, Karissa Tu, Ada Campagna, J Kelly Davis, Steven S Doerstling, Felicia Chen, Peter A Ubel

J Med Internet Res 2024;26:e50236

Download Citation: END BibTex RIS

Download PDF Download XML
Article has an altmetric score of 1

Global User-Level Perception of COVID-19 Contact Tracing Applications: Data-Driven Approach Using Natural Language Processing

Global User-Level Perception of COVID-19 Contact Tracing Applications: Data-Driven Approach Using Natural Language Processing

For this study, we used different transformer models, including BERT [13], Ro BERTa [37], XLM-Ro BERTa [37], and Distil BERT [38]. To measure the performance of each classifier, we used weighted average precision (P), recall (R), and F1. We used weighted metrics as they have the capability to take into account the class imbalance distribution.

Kashif Ahmad, Firoj Alam, Junaid Qadir, Basheer Qolomany, Imran Khan, Talhat Khan, Muhammad Suleman, Naina Said, Syed Zohaib Hassan, Asma Gul, Mowafa Househ, Ala Al-Fuqaha

JMIR Form Res 2022;6(5):e36238

Download Citation: END BibTex RIS

Download PDF Download XML
Article has an altmetric score of 7