Dataset Information

Semantic Textual Similarity in Japanese Clinical Domain Texts Using BERT.

ABSTRACT:

Background

Semantic textual similarity (STS) captures the degree of semantic similarity between texts. It plays an important role in many natural language processing applications such as text summarization, question answering, machine translation, information retrieval, dialog systems, plagiarism detection, and query ranking. STS has been widely studied in the general English domain. However, there exists few resources for STS tasks in the clinical domain and in languages other than English, such as Japanese.

Objective

The objective of this study is to capture semantic similarity between Japanese clinical texts (Japanese clinical STS) by creating a Japanese dataset that is publicly available.

Materials

We created two datasets for Japanese clinical STS: (1) Japanese case reports (CR dataset) and (2) Japanese electronic medical records (EMR dataset). The CR dataset was created from publicly available case reports extracted from the CiNii database. The EMR dataset was created from Japanese electronic medical records.

Methods

We used an approach based on bidirectional encoder representations from transformers (BERT) to capture the semantic similarity between the clinical domain texts. BERT is a popular approach for transfer learning and has been proven to be effective in achieving high accuracy for small datasets. We implemented two Japanese pretrained BERT models: a general Japanese BERT and a clinical Japanese BERT. The general Japanese BERT is pretrained on Japanese Wikipedia texts while the clinical Japanese BERT is pretrained on Japanese clinical texts.

Results

The BERT models performed well in capturing semantic similarity in our datasets. The general Japanese BERT outperformed the clinical Japanese BERT and achieved a high correlation with human score (0.904 in the CR dataset and 0.875 in the EMR dataset). It was unexpected that the general Japanese BERT outperformed the clinical Japanese BERT on clinical domain dataset. This could be due to the fact that the general Japanese BERT is pretrained on a wide range of texts compared with the clinical Japanese BERT.

SUBMITTER: Mutinda FW

PROVIDER: S-EPMC8294940 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundNatural Language Understanding enables automatic extraction of relevant information from clinical text data, which are acquired every day in hospitals. In 2018, the language model Bidirectional Encoder Representations from Transformers (BERT) was introduced, generating new state-of-the-art results on several downstream tasks. The National NLP Clinical Challenges (n2c2) is an initiative that strives to tackle such downstream tasks on domain-specific clinical data. In this paper, we present the results of our participation in the 2019 n2c2 and related work completed thereafter.ObjectiveThe objective of this study was to optimally leverage BERT for the task of assessing the semantic textual similarity of clinical text data.MethodsWe used BERT as an initial baseline and analyzed the results, which we used as a starting point to develop 3 different approaches where we (1) added additional, handcrafted sentence similarity features to the classifier token of BERT and combined the results with more features in multiple regression estimators, (2) incorporated a built-in ensembling method, M-Heads, into BERT by duplicating the regression head and applying an adapted training strategy to facilitate the focus of the heads on different input patterns of the medical sentences, and (3) developed a graph-based similarity approach for medications, which allows extrapolating similarities across known entities from the training set. The approaches were evaluated with the Pearson correlation coefficient between the predicted scores and ground truth of the official training and test dataset.ResultsWe improved the performance of BERT on the test dataset from a Pearson correlation coefficient of 0.859 to 0.883 using a combination of the M-Heads method and the graph-based similarity approach. We also show differences between the test and training dataset and how the two datasets influenced the results.ConclusionsWe found that using a graph-based similarity approach has the potential to extrapolate domain specific knowledge to unseen sentences. We observed that it is easily possible to obtain deceptive results from the test dataset, especially when the distribution of the data samples is different between training and test datasets.

Project description:BackgroundSemantic similarity measures estimate the similarity between concepts, and play an important role in many text processing tasks. Approaches to semantic similarity in the biomedical domain can be roughly divided into knowledge based and distributional based methods. Knowledge based approaches utilize knowledge sources such as dictionaries, taxonomies, and semantic networks, and include path finding measures and intrinsic information content (IC) measures. Distributional measures utilize, in addition to a knowledge source, the distribution of concepts within a corpus to compute similarity; these include corpus IC and context vector methods. Prior evaluations of these measures in the biomedical domain showed that distributional measures outperform knowledge based path finding methods; but more recent studies suggested that intrinsic IC based measures exceed the accuracy of distributional approaches. Limitations of previous evaluations of similarity measures in the biomedical domain include their focus on the SNOMED CT ontology, and their reliance on small benchmarks not powered to detect significant differences between measure accuracy. There have been few evaluations of the relative performance of these measures on other biomedical knowledge sources such as the UMLS, and on larger, recently developed semantic similarity benchmarks.ResultsWe evaluated knowledge based and corpus IC based semantic similarity measures derived from SNOMED CT, MeSH, and the UMLS on recently developed semantic similarity benchmarks. Semantic similarity measures based on the UMLS, which contains SNOMED CT and MeSH, significantly outperformed those based solely on SNOMED CT or MeSH across evaluations. Intrinsic IC based measures significantly outperformed path-based and distributional measures. We released all code required to reproduce our results and all tools developed as part of this study as open source, available under http://code.google.com/p/ytex. We provide a publicly-accessible web service to compute semantic similarity, available under http://informatics.med.yale.edu/ytex.web/.ConclusionsKnowledge based semantic similarity measures are more practical to compute than distributional measures, as they do not require an external corpus. Furthermore, knowledge based measures significantly and meaningfully outperformed distributional measures on large semantic similarity benchmarks, suggesting that they are a practical alternative to distributional measures. Future evaluations of semantic similarity measures should utilize benchmarks powered to detect significant differences in measure accuracy.

Project description:BackgroundNeural network based embedding models are receiving significant attention in the field of natural language processing due to their capability to effectively capture semantic information representing words, sentences or even larger text elements in low-dimensional vector space. While current state-of-the-art models for assessing the semantic similarity of textual statements from biomedical publications depend on the availability of laboriously curated ontologies, unsupervised neural embedding models only require large text corpora as input and do not need manual curation. In this study, we investigated the efficacy of current state-of-the-art neural sentence embedding models for semantic similarity estimation of sentences from biomedical literature. We trained different neural embedding models on 1.7 million articles from the PubMed Open Access dataset, and evaluated them based on a biomedical benchmark set containing 100 sentence pairs annotated by human experts and a smaller contradiction subset derived from the original benchmark set.ResultsExperimental results showed that, with a Pearson correlation of 0.819, our best unsupervised model based on the Paragraph Vector Distributed Memory algorithm outperforms previous state-of-the-art results achieved on the BIOSSES biomedical benchmark set. Moreover, our proposed supervised model that combines different string-based similarity metrics with a neural embedding model surpasses previous ontology-dependent supervised state-of-the-art approaches in terms of Pearson's r (r = 0.871) on the biomedical benchmark set. In contrast to the promising results for the original benchmark, we found our best models' performance on the smaller contradiction subset to be poor.ConclusionsIn this study, we have highlighted the value of neural network-based models for semantic similarity estimation in the biomedical domain by showing that they can keep up with and even surpass previous state-of-the-art approaches for semantic similarity estimation that depend on the availability of laboriously curated ontologies, when evaluated on a biomedical benchmark set. Capturing contradictions and negations in biomedical sentences, however, emerged as an essential area for further work.

Project description:BackgroundExome sequencing is a promising method for diagnosing patients with a complex phenotype. However, variant interpretation relative to patient phenotype can be challenging in some scenarios, particularly clinical assessment of rare complex phenotypes. Each patient's sequence reveals many possibly damaging variants that must be individually assessed to establish clear association with patient phenotype. To assist interpretation, we implemented an algorithm that ranks a given set of genes relative to patient phenotype. The algorithm orders genes by the semantic similarity computed between phenotypic descriptors associated with each gene and those describing the patient. Phenotypic descriptor terms are taken from the Human Phenotype Ontology (HPO) and semantic similarity is derived from each term's information content.ResultsModel validation was performed via simulation and with clinical data. We simulated 33 Mendelian diseases with 100 patients per disease. We modeled clinical conditions by adding noise and imprecision, i.e. phenotypic terms unrelated to the disease and terms less specific than the actual disease terms. We ranked the causative gene against all 2488 HPO annotated genes. The median causative gene rank was 1 for the optimal and noise cases, 12 for the imprecision case, and 60 for the imprecision with noise case. Additionally, we examined a clinical cohort of subjects with hearing impairment. The disease gene median rank was 22. However, when also considering the patient's exome data and filtering non-exomic and common variants, the median rank improved to 3.ConclusionsSemantic similarity can rank a causative gene highly within a gene list relative to patient phenotype characteristics, provided that imprecision is mitigated. The clinical case results suggest that phenotype rank combined with variant analysis provides significant improvement over the individual approaches. We expect that this combined prioritization approach may increase accuracy and decrease effort for clinical genetic diagnosis.