A Multi-Stage NLP Framework for Knowledge Discovery from Crop Disease Research Literature

Pneumonia is caused by microbes that establish an infectious process in the lungs. The gold standard for pneumonia diagnosis is radiologist-documented pneumonia-related features in radiology notes that are captured in electronic health records in an unstructured format. The study objective was to develop a methodological approach for assessing validity of a pneumonia diagnosis based on identifying presence or absence of key radiographic features in radiology reports with subsequent rendering of diagnostic decisions into a structured format. A pneumonia-specific natural language processing (NLP) pipeline was strategically developed applying Clinical Text Analysis and Knowledge Extraction System (cTAKES) to validate pneumonia diagnoses following development of a pneumonia feature-specific lexicon. Radiographic reports of study-eligible subjects identified by International Classification of Diseases (ICD) codes were parsed through the NLP pipeline. Classification rules were developed to assign each pneumonia episode into one of three categories: "positive," "negative," or "not classified: requires manual review" based on tagged concepts that support or refute diagnostic codes. A total of 91,998 pneumonia episodes diagnosed in 65,904 patients were retrieved retrospectively. Approximately 89% (81,707/91,998) of the total pneumonia episodes were documented by 225,893 chest X-ray reports. NLP classified and validated 33% (26,800/81,707) of pneumonia episodes classified as "Pneumonia-positive," 19% as (15401/81,707) as "Pneumonia-negative," and 48% (39,209/81,707) as "episode classification pending further manual review." NLP pipeline performance metrics included accuracy (76.3%), sensitivity (88%), and specificity (75%). The pneumonia-specific NLP pipeline exhibited good performance comparable to other pneumonia-specific NLP systems developed to date.

Dependency parsing, which is a fundamental task in Natural Language Processing (NLP), has attracted a lot of interest in recent years. In general, it is a module in an NLP pipeline together with word segmentation and Part-Of-Speech (POS) tagging in real Chinese NLP application. The NLP pipeline, which is a cascade system, will lead to error propagation for the parsing. This paper proposes a global discriminative re-ranking model using non-local features from word segmentation, POS tagging and dependency parsing to re-rank the parse trees produced by an N-best enhanced NLP pipeline. Experimental results indicate that the proposed model can improve the performance of dependency parsing as well as word segmentation and POS tagging in an NLP pipeline.

BackgroundDeveloping predictive models for precision psychiatry is challenging because of unavailability of the necessary data: extracting useful information from existing electronic health record (EHR) data is not straightforward, and available clinical trial datasets are often not representative for heterogeneous patient groups. The aim of this study was constructing a natural language processing (NLP) pipeline that extracts variables for building predictive models from EHRs. We specifically tailor the pipeline for extracting information on outcomes of psychiatry treatment trajectories, applicable throughout the entire spectrum of mental health disorders (“transdiagnostic”).MethodsA qualitative study into beliefs of clinical staff on measuring treatment outcomes was conducted to construct a candidate list of variables to extract from the EHR. To investigate if the proposed variables are suitable for measuring treatment effects, resulting themes were compared to transdiagnostic outcome measures currently used in psychiatry research and compared to the HDRS (as a gold standard) through systematic review, resulting in an ideal set of variables. To extract these from EHR data, a semi-rule based NLP pipeline was constructed and tailored to the candidate variables using Prodigy. Classification accuracy and F1-scores were calculated and pipeline output was compared to HDRS scores using clinical notes from patients admitted in 2019 and 2020.ResultsAnalysis of 34 questionnaires answered by clinical staff resulted in four themes defining treatment outcomes: symptom reduction, general well-being, social functioning and personalization. Systematic review revealed 242 different transdiagnostic outcome measures, with the 36-item Short-Form Survey for quality of life (SF36) being used most consistently, showing substantial overlap with the themes from the qualitative study. Comparing SF36 to HDRS scores in 26 studies revealed moderate to good correlations (0.62—0.79) and good positive predictive values (0.75—0.88). The NLP pipeline developed with notes from 22,170 patients reached an accuracy of 95 to 99 percent (F1 scores: 0.38 – 0.86) on detecting these themes, evaluated on data from 361 patients.ConclusionsThe NLP pipeline developed in this study extracts outcome measures from the EHR that cater specifically to the needs of clinical staff and align with outcome measures used to detect treatment effects in clinical trials.

Recently, deep clustering methods have achieved remarkable results compared to traditional clustering approaches. However, its performance remains constrained by the absence of annotations. A thought-provoking observation is that there is still a significant gap between deep clustering and semi-supervised classification methods. Even with only a few labeled samples, the accuracy of semi-supervised learning is much higher than that of clustering. Given that we can annotate a small number of samples in a certain unsupervised way, the clustering task can be naturally transformed into a semi-supervised setting, thereby achieving comparable performance. Based on this intuition, we propose ClusMatch, a unified positive and negative pseudo-label learning based semi-supervised learning framework, which is pluggable and can be applied to existing deep clustering methods. Specifically, we first leverage the pre-trained deep clustering network to compute predictions for all samples, and then design specialized selection strategies to pick out a few high-quality samples as labeled samples for supervised learning. For the unselected samples, the novel unified positive and negative pseudo-label learning is introduced to provide additional supervised signals for semi-supervised fine-tuning. We also propose an adaptive positive-negative threshold learning strategy to further enhance the confidence of generated pseudo-labels. Extensive experiments on six widely-used datasets and one large-scale dataset demonstrate the superiority of our proposed ClusMatch. For example, ClusMatch achieves a significant accuracy improvement of 5.4% over the state-of-the-art method ProPos on an average of these six datasets.

BackgroundProcessing data from electronic health records (EHRs) to build research-grade databases is a lengthy and expensive process. Modern arthroplasty practice commonly uses multiple sites of care, including clinics and ambulatory care centers. However, most private data systems prevent obtaining usable insights for clinical practice.ObjectiveThis study aims to create an automated natural language processing (NLP) pipeline for extracting clinical concepts from EHRs related to orthopedic outpatient visits, hospitalizations, and surgeries in a multicenter, single-surgeon practice. The pipeline was also used to assess therapies and complications after total hip arthroplasty (THA).MethodsEHRs of 1290 patients undergoing primary THA from January 1, 2012 to December 31, 2019 (operated and followed by the same surgeon) were processed using artificial intelligence (AI)–based models (NLP and machine learning). In addition, 3 independent medical reviewers generated a gold standard using 100 randomly selected EHRs. The algorithm processed the entire database from different EHR systems, generating an aggregated clinical data warehouse. An additional manual control arm was used for data quality control.ResultsThe algorithm was as accurate as human reviewers (0.95 vs 0.94; P=.01), achieving a database-wide average F1-score of 0.92 (SD 0.09; range 0.67‐0.99), validating its use as an automated data extraction tool. During the first year after direct anterior THA, 92.1% (1188/1290) of our population had a complication-free recovery. In 7.9% (102/1290) of cases where surgery or recovery was not uneventful, lateral femoral cutaneous nerve sensitivity (47/1290, 3.6%), intraoperative fractures (13/1290, 1%), and hematoma (9/1290, 0.7%) were the most common complications.ConclusionsAlgorithm evaluation of this dataset accurately represented key clinical information swiftly, compared with human reviewers. This technology may provide substantial value for future surgeon practice and patient counseling. Furthermore, the low early complication rate of direct anterior THA in this surgeon’s hands was supported by the dataset, which included data from all treated patients in a multicenter practice.

Natural language processing (NLP) is an important and challenging field of computer science. With the advent of computers and related technologies, NLP has grown out of the field of linguistics and is getting more attention in the computer field to support a wide variety of intelligent applications. NLP basically refers to automatic processing of natural languages by computers or machines in order to understand human language in written or spoken form. Ontologies are semantic models that provide general description of the types and structures of objects in a particular subject area or domain. Core components of ontology include concepts, relationships, instances, and axioms. Ontology is used to depict the hierarchical organization of concepts and their relationships. Knowledge graphs represent a set of interconnected descriptions of objects or concepts, and are sometimes referred to as semantic networks. In short, an ontology can be treated as a meta-model for data and knowledge graphs as a representation of the actual data or facts. An ontology (a formal description of relationships) plays a vital role in building a knowledge graph. Ontology and knowledge graphs have proven to be powerful tools to manage enterprise or big data. Today, a number of applications rely on ontology and knowledge graphs, such as knowledge management, information discovery, text extraction, semantic search, content-based recommendation, and intelligent chatbots. This chapter focuses on the importance of ontology and knowledge graphs for Developing NLP-based applications is difficult because computers require structured data, and human speech is highly complex, extremely diverse, unstructured in nature, and often ambiguous. In general, manual creation of knowledge graphs is expensive, therefore any automation of the process is highly desirable. Here, ontologies and knowledge graphs play a crucial role by providing a well-structured knowledge base, allowing semi-automatic creation of a lexicon to be used effectively in language translations, spellcheckers, keyword extraction and similar NLP applications.

This paper investigates the impact of knowledge extraction methods on the effectiveness of RAG (Retrieval-Augmented Generation) systems that utilize knowledge graphs. It highlights that the quality of the knowledge graph, which is formed using various knowledge extraction methods, is crucial for overcoming limitations of large language models (LLMs), such as “hallucinations”. The paper analyzes the architectures of LightRAG and GraphRAG, emphasizing that the selection of an optimal knowledge extraction strategy depends on specific tasks and the subject area. LLMs have advanced significantly, but they have limitations, including generating factually incorrect information (“hallucinations”) and possessing “outdated knowledge”. RAG systems were proposed to address these issues by combining LLMs with external knowledge bases. This approach reduces hallucinations, ensures factual accuracy, solves the problem of outdated knowledge, and increases transparency. Knowledge graphs are powerful tools for structuring information, consisting of entities (nodes) and relations (edges). They enhance RAG systems by enabling more precise and contextually grounded retrieval compared to keyword-based searches. The quality of a knowledge graph depends on the knowledge extraction methods used, which include named entity recognition (NER), relation extraction (RE), entity linking (EL), and event extraction. Different methods, such as rule-based, classical machine learning, and deep learning approaches, have varying trade-offs in terms of accuracy, completeness, and scalability. Entity linking and knowledge graph completion are also crucial for accuracy and richness. LightRAG and GraphRAG are two main graph-based RAG systems. LightRAG uses the knowledge graph as a quick reference, requiring high precision in knowledge extraction to avoid noise degradation. GraphRAG uses the knowledge graph as a domain model, where completeness of extracted knowledge is more critical, though systematic errors are still harmful. Both systems rely on LLMs for knowledge extraction, which makes them dependent on the LLM’s quality and the size of document fragments processed. The theoretical analysis confirms that the effectiveness of RAG systems is critically dependent on knowledge extraction methods. The quality, completeness, and accuracy of the knowledge graph directly influence the RAG system’s ability to provide relevant, accurate, and truthful answers. Different RAG system architectures like LightRAG and GraphRAG have distinct requirements for knowledge graph characteristics, prioritizing either accuracy or completeness in knowledge extraction.

Extracting information from unstructured and semi-structured geoscience literature is a crucial step in conducting geological research. The traditional machine learning extraction paradigm requires a substantial amount of high-quality manually annotated data for model training, which is time-consuming, labor-intensive, and not easily transferable to new fields. Recently, large language models (LLMs) (e.g., ChatGPT, GPT-4, and LLaMA), have shown great performance in various natural language processing (NLP) tasks, such as question answering, machine translation, and text generation. A substantial body of work has demonstrated that LLMs possess strong in-context learning (ICL) and even zero-shot learning capabilities to solve downstream tasks without specifically designed supervised fine-tuning. In this paper, we propose utilizing LLMs for geoscience literature information extraction. Specifically, we design a hierarchical PDF parsing pipeline and an automated knowledge extraction process, which can significantly reduce the need for manual data annotation, assisting geoscientists in literature data mining. For the hierarchical PDF parsing pipeline, firstly, a document layout detection model fine-tuned on geoscience literature is employed for layout detection, obtaining layout detection information for the document. Secondly, based on the document layout information, an optical character content parsing model is used for content parsing, obtaining the text structure and plain text corresponding to the content. Finally, the text structure and plain text are combined and reconstructed to ultimately obtain the parsed structured data. For the automated knowledge extraction process, firstly, the parsed long text is segmented into paragraphs to adapt to the input length limit of LLMs. Subsequently, a few-shot prompting method is employed for structured knowledge extraction, encompassing two tasks: attribute value extraction and triplet extraction. In attribute value extraction, prompts are generated automatically by the LLMs based on the subdomain and attribute names, facilitating the location and extraction of values related to subdomain attribute names in the text. For triplet extraction, the LLMs employ a procedural approach to entity extraction, entity type extraction, and relation extraction, following the knowledge graph structure pattern. Finally, the extracted structured knowledge is stored in the form of knowledge graphs, facilitating further analysis and integration of various types of knowledge from the literature. Our proposed approach turns out to be simple, flexible, and highly effective in geoscience literature information extraction. Demonstrations of information extraction in subdomains such as radiolarian fossils and fluvial facies have yielded satisfactory results. The extraction efficiency has significantly improved, and feedback from domain experts indicates a relatively high level of accuracy in the extraction process. The extracted results can be used to construct a foundational knowledge graph for geoscience literature, supporting the comprehensive construction and efficient application of a geoscience knowledge graph.

BackgroundSocial isolation is an important social determinant that impacts health outcomes and mortality among patients. The National Academy of Medicine recently recommended that social isolation be documented in electronic health records (EHR). However, social isolation usually is not recorded or obtained as coded data but rather collected from patient self-report or documented in clinical narratives. This study explores the feasibility and effectiveness of natural language processing (NLP) strategy for identifying patients who are socially isolated from clinical narratives.MethodWe used data from the Medical University of South Carolina (MUSC) Research Data Warehouse. Patients 18 years-of-age or older who were diagnosed with prostate cancer between January 1, 2014 and May 31, 2017 were eligible for this study. NLP pipelines identifying social isolation were developed via extraction of notes on progress, history and physical, consult, emergency department provider, telephone encounter, discharge summary, plan of care, and radiation oncology. Of 4195 eligible prostate cancer patients, we randomly sampled 3138 patients (75%) as a training dataset. The remaining 1057 patients (25%) were used as a test dataset to evaluate NLP algorithm performance. Standard performance measures for the NLP algorithm, including precision, recall, and F-measure, were assessed by expert manual review using the test dataset.ResultsA total of 55,516 clinical notes from 3138 patients were included to develop the lexicon and NLP pipelines for social isolation. Of those, 35 unique patients (1.2%) had social isolation mention(s) in 217 notes. Among 24 terms relevant to social isolation, the most prevalent were “lack of social support,” “lonely,” “social isolation,” “no friends,” and “loneliness”. Among 1057 patients in the test dataset, 17 patients (1.6%) were identified as having social isolation mention(s) in 40 clinical notes. Manual review identified four false positive mentions of social isolation and one false negatives in 154 notes from randomly selected 52 controls. The NLP pipeline demonstrated 90% precision, 97% recall, and 93% F-measure. The major reasons for a false positive included the ambiguities of the experiencer of social isolation, negation, and alternate meaning of words.ConclusionsOur NLP algorithms demonstrate a highly accurate approach to identify social isolation.

Neural symbolic reasoning with knowledge graphs: Knowledge extraction, relational reasoning, and inconsistency checking

Knowledge graphs have shown increasing value in semantic search, intelligent Q&A, data analysis, natural language processing, visual understanding, IoT devices, etc. It is undeniable that knowledge graphs have become the mandatory path for the artificial intelligence fields development. However, the information contained in existing knowledge graphs is incomplete, which attracts a large number of researchers to enhance the completeness of knowledge graphs utilizing knowledge graph completion methods. Most traditional embedding-based knowledge graph completion models use structural information within data-rich triples from the knowledge graph is limited by the long-tail distribution of the relations in the triples. To address this problem, we propose a Bert-based knowledge graph completion algorithm for few-shot knowledge graphs, with the main goal of implementing the knowledge graph completion task with only a few sample triples of training instances. We improve the baseline model GMatching for handling few-shot knowledge graphs by introducing the Bert pre-trained linguistic representation model to enhance the semantic representation of entities and relations in the triples. Through experiments, we demonstrate that our improved model B-GMatching achieves good results.

Most of the biomedical knowledge the research community has acquired during the past few decades has been deposited in scientific literature as unstructured text. Converting the unstructured text into the structured form will enable novel methodologies and applications for scientific discovery that can fully harness the power of the existing knowledge. To this end, two fundamental questions need to be addressed: named entity recognition (NER) and relation extraction (RE). NER deals with identifying the concepts or entities in texts, such as diseases, genes/proteins, chemical compounds, etc. while RE aims to extract the relations among these entities. Together, the extracted information forms a knowledge graph (KG) where the nodes are entities in the texts and the edges represent their relationships. KGs can link concepts within existing research to allow researchers to find connections that may have been difficult to discover without them. The LitCoin Natural Language Processing (NLP) Challenge was recently organized by NCATS of NIH and NASA to spur innovation by rewarding the most creative and high-impact uses of biomedical text to create KGs. Our team participated in the challenge and ranked first place. We have applied the methods we developed for the LitCoin NLP challenge to all PubMed abstracts and constructed the largest-scale biomedical KG to date. We show that powerful and versatile query functions can be implemented on top of the KG to enable highly specific and accurate knowledge retrieval and inference of causal and indirect relationships. Citation Format: Xin Sui, Yuan Zhang, Feng Pan, Donghu Sun, Menghan Chung, Jinfeng Zhang. Constructing the largest-scale knowledge graph using all PubMed abstracts and its application for highly specific and accurate knowledge retrieval. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5421.

A knowledge graph (KG) consists of numerous triples, in which each triple, i.e., (head entity, relation, tail entity), denotes a real-world assertion. Many large-scale KGs have been developed, e.g., general-purpose KGs Freebase and YAGO. Also, lots of domain-specific KGs are emerging, e.g., COVID-19 KGs, biomedical KGs, and agricultural KGs. By embedding KGs into low-dimensional vectors, i.e., representations of entities and relations, we could integrate KGs into machine learning models and enhance the performance of many prediction tasks, including search, recommendations, and question answering. During the construction, refinement, embedding, and application of KGs, a variety of KG learning algorithms have been developed to handle challenges in various real-world scenarios. Moreover, graph neural networks have also brought new opportunities to KG learning. This workshop aims to engage with active researchers from KG communities, recommendation communities, natural language processing communities, and other communities, and deliver state-of-the-art research insights into the core challenges in KG learning.

ObjectiveCluster analysis of spatio-temporal event-related potential (ERP) data is a promising tool for exploring the measurement time window of ERPs. However, even after preprocessing, the remaining noise can result in uncertain cluster maps followed by unreliable time windows while clustering via conventional clustering methods. MethodsWe designed an ensemble deep clustering pipeline to determine a reliable time window for the ERP of interest from temporal concatenated grand average ERP data. The proposed pipeline includes semi-supervised deep clustering methods initialized by consensus clustering and unsupervised deep clustering methods with end-to-end architectures. Ensemble clustering from those deep clusterings was used by the designed adaptive time window determination to estimate the time window. ResultsAfter applying simulated and real ERP data, our method successfully obtained the time window for identifying the P3 components (as the interest of both ERP studies) while additional noise (e.g., adding 20 dB to −5 dB white Gaussian noise) was added to the prepared data. ConclusionCompared to the state-of-the-art clustering methods, a superior clustering performance was yielded from both ERP data. Furthermore, more stable and precise time windows were elicited as the noise increased. SignificanceOur study provides a complementary understanding of identifying the cognitive process using deep clustering analysis to the existing studies. Our finding suggests that deep clustering can be used to identify the ERP of interest when the data is imperfect after preprocessing.

Trie-based rule processing for clinical NLP: A use-case study of n-trie, making the ConText algorithm more efficient and scalable.