Information Extraction Tools Research Articles

A Multi-Label Machine Learning Approach to Support Pathologist's Histological Analysis

This paper proposes a new tool in the field of telemedicine, defined as a specific branch where IT supports medicine, in case distance impairs the proper care to be delivered to a patient. All the information contained into medical texts, if properly extracted, may be suitable for searching, classification, or statistical analysis. For this reason, in order to reduce errors and improve quality control, a proper information extraction tool may be useful. In this direction, this work presents a Machine Learning Multi-Label approach for the classification of the information extracted from the pathology reports into relevant categories. The aim is to integrate automatic classifiers to improve the current workflow of medical experts, by defining a Multi-Label approach, able to consider all the features of a model, together with their relationships.

Recovering Damaged Documents to Improve Information Retrieval Processes

Although computer forensics is frequently related to the investigation of computer crimes, it can also be used in civil procedures. An example of case of use is information retrieval from damaged documents, where words have undergone alterations, either accidentally or intentionally. In this paper, we present a new tool able to retrieve information from large volumes of documents whose contents have been damaged. We have designed a new approach to recover the original words, composed of two stages: a text cleaning filter, able to remove non relevant information, and a text correction unit, which gather a general purpose spell checker with a N-gram based spell checker built specifically for the domain of the documents. The benefits of using this combined approach are two-fold: on the one hand, the general spell checker allows us to leverage all the general purpose techniques that are usually used to perform the corrections; on the other hand, the use of an N-gram based model allows us to adapt them to the particular domain we are tackling exploiting text regularities detected in successfully processed domain documents. The result of the correction allows us to improve automatic information retrieval tasks of from the texts. We have tested it using a real data set by using an information extraction tool based on semantic technologies in collaboration with the Spanish company InSynergy Consulting.

SIA: a scalable interoperable annotation server for biomedical named entities

Recent years showed a strong increase in biomedical sciences and an inherent increase in publication volume. Extraction of specific information from these sources requires highly sophisticated text mining and information extraction tools. However, the integration of freely available tools into customized workflows is often cumbersome and difficult. We describe SIA (Scalable Interoperable Annotation Server), our contribution to the BeCalm-Technical interoperability and performance of annotation servers (BeCalm-TIPS) task, a scalable, extensible, and robust annotation service. The system currently covers six named entity types (i.e., chemicals, diseases, genes, miRNA, mutations, and organisms) and is freely available under Apache 2.0 license at https://github.com/Erechtheus/sia.

Extraction of Meaningful Information from the Web: a Brief Survey

There is an explosive growth of information on Internet that makes extraction of relevant data from various sources, a difficult task for its users. Therefore, to transform the Web pages into databases, Information Extraction (IE) systems are needed. Relevant information in Web documents can be extracted using information extraction and presented in a structured format.By applying information extraction techniques, information can be extracted from structured, semi-structured, and unstructured data. This paper presents some of the major information extraction tools. Here, advantages and limitations of the tools are discussed from a user’s perspective.

Identifying direct temporal relations between time and events from clinical notes

BackgroundMost of the current work on clinical temporal relation identification follows the convention developed in the general domain, aiming to identify a comprehensive set of temporal relations from a document including both explicit and implicit relations. While such a comprehensive set can represent temporal information in a document in a complete manner, some of the temporal relations in the comprehensive set may not be essential depending on the clinical application of interest. Moreover, as the types of evidence that should be used to identify explicit and implicit relations are different, current clinical temporal relation identification systems that target both explicit and implicit relations still show low performances for practical use.MethodsIn this paper, we propose to focus on a sub-task of conventional temporal relation identification task in order to provide insight into building practical temporal relation identification modules for clinical text. We focus on identification of direct temporal relations, a subset of temporal relations that is chosen to minimize the amount of inference required to identify the relations. A corpus on direct temporal relations between time expressions and event mentions is constructed, and an automatic system tailored for direct temporal relations is developed.ResultsIt is shown that the direct temporal relations constitute a major category of temporal relations that contain important information needed for clinical applications. The system optimized for direct temporal relations achieves better performance than the state-of-the-art system developed with comprehensive set of both explicit and implicit relations in mind.ConclusionsWe expect direct temporal relations to facilitate the development of practical temporal information extraction tools in clinical domain.

Analysis Of Single And Hybrid Data Mining Techniques For Prediction Of Heart Disease Using Real Time Dataset

Data mining and healthcare has strong relations as data mining is a process where we can analyze enormous set of data and after extraction meaning of data can be understood. As medical data is in bulk which leads to the need of data analytics tool for extraction of useful information or knowledge .Disease Prediction is one of the applications where many researchers are working so that they can answer whether to apply single or compound data mining technique. Heart related disease is the major cause of deaths in past decades. Use of single technique in treatment of heart disease is showing variable accuracy. Current leading research study is all about to check the effect of compound data mining technique will show enhanced results in the prognosis of heart related disease or not, but the major issue arises of identification a suitable treatment for heart disease patient. This survey paper notify gap in research findings on heart disease prognosis and prescription data and proposes a model which will fill the gap and shows enhanced accuracy.

Feature Extraction From Multitemporal SAR Images Using Selforganizing Map Clustering and Object-Based Image Analysis

We introduce a new architecture for feature extraction from multitemporal synthetic aperture radar (SAR) data. Its the purpose is to combine classic SAR processing and geographical object-based image analysis to provide a robust unsupervised tool for information extraction from time series images. The architecture takes advantage from the characteristics of the recently introduced RGB products of the Level-1 $\alpha$ and Level-1 $\beta$ families, and employs self-organizing map clustering and object-based image analysis. In particular, the input products are clustered using color homogeneity and automatically enriched with a semantic attribute referring to clusters’ color, providing a preclassification mask. Then, in the frame of an application-oriented object-based image analysis, opportune layers measuring scattering and geometric properties of candidate objects are evaluated, and an appropriate rule-set is implemented in a fuzzy system to extract the feature of interest. The obtained results have been compared with those given by existing techniques and turned out to provide high degree of accuracy and negligible false alarms. The discussion is supported by an example concerning small reservoir mapping in semiarid environment.

Respiration Symptoms Monitoring in Body Area Networks

This work presents a framework that monitors particular symptoms such as respiratory conditions (abnormal breathing pattern) experienced by hyperthyreosis, sleep apnea, and sudden infant death syndrome (SIDS) patients. The proposed framework detects and monitors respiratory condition using S-Band sensing technique that leverages the wireless devices such as antenna, card, omni-directional antenna operating in 2 GHz to 4 GHz frequency range, and wireless channel information extraction tool. The rhythmic patterns extracted using S-Band sensing present the periodic and non-periodic waveforms that correspond to normal and abnormal respiratory conditions, respectively. The fine-grained amplitude information obtained using aforementioned devices is used to examine the breathing pattern over a period of time and accurately identifies the particular condition.

Twitter Users’ Privacy Concerns: What do Their Accounts’ First Names Tell Us?

Abstract Purpose In this paper, we describe how gender recognition on Twitter can be used as an intelligent business tool to determine the privacy concerns among users, and ultimately offer a more personalized service for customers who are more likely to respond positively to targeted advertisements. Design/methodology/approach We worked with two different data sets to examine whether Twitter users’ gender, inferred from the first name of the account and the profile description, correlates with the privacy setting of the account. We also used a set of features including the inferred gender of Twitter users to develop classifiers that predict user privacy settings. Findings We found that the inferred gender of Twitter users correlates with the account’s privacy setting. Specifically, females tend to be more privacy concerned than males. Users whose gender cannot be inferred from their provided first names tend to be more privacy concerned. In addition, our classification performance suggests that inferred gender can be used as an indicator of the user’s privacy preference. Research limitations It is known that not all twitter accounts are real user accounts, and social bots tweet as well. A major limitation of our study is the lack of consideration of social bots in the data. In our study, this implies that at least some percentage of the undefined accounts, that is, accounts that had names non-existent in the name dictionary, are social bots. It will be interesting to explore the privacy setting of social bots in the Twitter space. Practical implications Companies are investing large amounts of money in business intelligence tools that allow them to know the preferences of their consumers. Due to the large number of consumers around the world, it is very difficult for companies to have direct communication with each customer to anticipate market changes. For this reason, the social network Twitter has gained relevance as one ideal tool for information extraction. On the other hand, users’ privacy preference needs to be considered when companies consider leveraging their publicly available data. This paper suggests that gender recognition of Twitter users, based on Twitter users’ provided first names and their profile descriptions, can be used to infer the users’ privacy preference. Originality/value This study explored a new way of inferring Twitter user’s gender, that is, to recognize the user’s gender based on the provided first name and the user’s profile description. The potential of this information for predicting the user’s privacy preference is explored.

Integrative image segmentation optimization and machine learning approach for high quality land-use and land-cover mapping using multisource remote sensing data

The growing use of optimization for geographic object-based image analysis and the possibility to derive a wide range of information about the image in textual form makes machine learning (data mining) a versatile tool for information extraction from multiple data sources. This paper presents application of data mining for land-cover classification by fusing SPOT-6, RADARSAT-2, and derived dataset. First, the images and other derived indices (normalized difference vegetation index, normalized difference water index, and soil adjusted vegetation index) were combined and subjected to segmentation process with optimal segmentation parameters obtained using combination of spatial and Taguchi statistical optimization. The image objects, which carry all the attributes of the input datasets, were extracted and related to the target land-cover classes through data mining algorithms (decision tree) for classification. To evaluate the performance, the result was compared with two nonparametric classifiers: support vector machine (SVM) and random forest (RF). Furthermore, the decision tree classification result was evaluated against six unoptimized trials segmented using arbitrary parameter combinations. The result shows that the optimized process produces better land-use land-cover classification with overall classification accuracy of 91.79%, 87.25%, and 88.69% for SVM and RF, respectively, while the results of the six unoptimized classifications yield overall accuracy between 84.44% and 88.08%. Higher accuracy of the optimized data mining classification approach compared to the unoptimized results indicates that the optimization process has significant impact on the classification quality.

Quantitative computed tomography applied to interstitial lung diseases

ObjectivesTo evaluate a new image marker that retrieves information from computed tomography (CT) density histograms, with respect to classification properties between different lung parenchyma groups. Furthermore, to conduct a comparison of the new image marker with conventional markers. Materials and methodsDensity histograms from 220 different subjects (normal = 71; emphysema = 73; fibrotic = 76) were used to compare the conventionally applied emphysema index (EI), 15th percentile value (PV), mean value (MV), variance (V), skewness (S), kurtosis (K), with a new histogram’s functional shape (HFS) method. Multinomial logistic regression (MLR) analyses was performed to calculate predictions of different lung parenchyma group membership using the individual methods, as well as combinations thereof, as covariates. Overall correct assigned subjects (OCA), sensitivity (sens), specificity (spec), and Nagelkerke’s pseudo R2 (NR2) effect size were estimated. NR2 was used to set up a ranking list of the different methods. ResultsMLR indicates the highest classification power (OCA of 92%; sens 0.95; spec 0.89; NR2 0.95) when all histogram analyses methods were applied together in the MLR. Highest classification power among individually applied methods was found using the HFS concept (OCA 86%; sens 0.93; spec 0.79; NR2 0.80). Conventional methods achieved lower classification potential on their own: EI (OCA 69%; sens 0.95; spec 0.26; NR2 0.52); PV (OCA 69%; sens 0.90; spec 0.37; NR2 0.57); MV (OCA 65%; sens 0.71; spec 0.58; NR2 0.61); V (OCA 66%; sens 0.72; spec 0.53; NR2 0.66); S (OCA 65%; sens 0.88; spec 0.26; NR2 0.55); and K (OCA 63%; sens 0.90; spec 0.16; NR2 0.48). ConclusionThe HFS method, which was so far applied to a CT bone density curve analysis, is also a remarkable information extraction tool for lung density histograms. Presumably, being a principle mathematical approach, the HFS method can extract valuable health related information also from histograms from complete different areas.

Lightweight predicate extraction for patient-level cancer information and ontology development

BackgroundKnowledge engineering for ontological knowledgebases is resource and time intensive. To alleviate these issues, especially for novices, automated tools from the natural language domain can assist in the development process of ontologies. We focus towards the development of ontologies for the public health domain and use patient-centric sources from MedlinePlus related to HPV-causing cancers.MethodsThis paper demonstrates the use of a lightweight open information extraction (OIE) tool to derive accurate knowledge triples that can lead to the seeding of an ontological knowledgebase. We developed a custom application, which interfaced with an information extraction software library, to help facilitate the tasks towards producing knowledge triples from textual sources.ResultsThe results of our efforts generated accurate extractions ranging from 80–89% precision. These triples can later be transformed to OWL/RDF representation for our planned ontological knowledgebase.ConclusionsOIE delivers an effective and accessible method towards the development ontologies.

Study of the variability of the Leap Motion’s measures for its use to characterize air strokes

This study presents an evaluation of different measurements provided by the Leap Motion sensor for characterization of air strokes and its stability. This test consists on the performance of a daily in air signature done by the same user during 4 months. Each one of the strokes, which represent the signatures of the users, is characterized for 21 different measures given by the sensor. The experiment is based on the evaluation of the temporal variability of the strokes according to these 21 measures. To compute the temporal variability of the measurements, different values as the length of the strokes, the variability of the mean and standard deviation of the measures, alignment costs referenced to the signature of the first day, etc. are obtained. In general, the results show some differences between the stability of the different measurements which can be produced either by the user or the sensor. In our conclusion, the results are promising and the Leap Motion could be used as a tool for biometric information extraction from the air strokes, as it has been already proposed. This novel study will help to understand and know the option of the use of Leap Motion from the measure stability point of view, for other application as signature identification.

Event-Based Trading: Building Superior Trading Strategies with State-of-the-Art Information Extraction Tools

Advanced sentiment analysis systems, which typically produce a positive or negative signal on firms for financial traders, consider financial events when calculating aggregate sentiment scores. However, these systems typically produce overly noisy signals for traders because they aggregate the information culled from the events rather than address each distinct event-type differently (all event-types receive the same weight in a text’s final sentiment score). We demonstrate that event-types have different informative values for traders, and specific event-types may be particularly relevant for specific investment periods. In this work we evaluate and compare the effects of event-type signals and sentiment analysis signals, automatically extracted from a given text, on stock prices in different investment periods. First, we establish that the polarity of different event-types is reflected in the stock price with different time lags. Second, we use benchmark portfolio and show that, when considering the polarity of different events for different time periods, one can generate excess return. Finally, we use the Quantopian Research Platform to test investment strategies based on different event-types and based on sentiment analysis signals, over various validation periods. For the experiments, we extracted event-types and sentiments from a large dataset of articles on a large number of firms using VIP, a state-of-the-art information extraction platform, and obtained the corresponding historical stock prices. We show that traders using trading strategies based on event-types and sentiment analysis obtain superior performance compared to acceptable benchmarks. We further show that when such investment strategies are based on daily aggregated event score, performance is not superior to strategies based on daily sentiment analysis scores. Performance is enhanced further when strategies use select event-type scores. Finally, we show that different event-types have different temporal effects on portfolio performance.

A Distributed Cross-layer Recommender System Incorporating Product Diffusion

In this era of online retailing, personalisation of web content has become very essential. Recommender system is a tool for extraction of relevant information to render personalisation in web information retrieval systems. With an inclination towards customer oriented service, there is a need to understand the adaptability of customers, to provide products/services of interest at the right time. In this paper, a model for distributed context aware cross layer recommender system incorporating the principle of product diffusion is proposed. The offline-online modelled recommender system learns offline about the adaptation time of users using the principle of product diffusion and then, uses online explore-then-exploit strategy to make effective recommendations to the user at the most probable time of consumption. Also, an algorithm based on product adaptability is proposed for recommending new items to the most probable users. The extensive experiments and results demonstrate the efficiency, scalability, reliability and enhanced retrieval effectiveness of the proposed recommender system model.

Data Mining and Analytics in the Process Industry: The Role of Machine Learning

Data mining and analytics have played an important role in knowledge discovery and decision making/supports in the process industry over the past several decades. As a computational engine to data mining and analytics, machine learning serves as basic tools for information extraction, data pattern recognition and predictions. From the perspective of machine learning, this paper provides a review on existing data mining and analytics applications in the process industry over the past several decades. The state-of-the-art of data mining and analytics are reviewed through eight unsupervised learning and ten supervised learning algorithms, as well as the application status of semi-supervised learning algorithms. Several perspectives are highlighted and discussed for future researches on data mining and analytics in the process industry.

Retrieval of similar topographic areas.

Image processing and understanding and further pattern recognition comprises a precious tool for the automatic extraction of information using digital topography. The aim of this work is the retrieval of areas with similar topography using digital elevation data. It can be applied to geomorphology, forestry, regional and urban planning, and many other applications for analyzing and managing natural resources. In specifics, the user selects the area of interest, navigating overhead a high resolution elevation image and determines two (3) parameters (step, number of local minima and display scale). Furthermore the regions with similar relief to the initial model are determined. Experimental results show high efficiency of the proposed scheme.

How to improve information extraction from German medical records

Abstract Vast amounts of medical information are still recorded as unstructured text. The knowledge contained in this textual data has a great potential to improve clinical routine care, to support clinical research, and to advance personalization of medicine. To access this knowledge, the underlying data has to be semantically integrated – an essential prerequisite to which is information extraction from clinical documents. A body of work, and a good selection of openly available tools for information extraction and semantic integration in the medical domain exist, yet almost exclusively for English language documents. For German texts the situation is rather different: research work is sparse, tools are proprietary or unpublished, and rarely any freely available textual resources exist. In this survey, we (1) describe the challenges of information extraction from German medical documents and the hurdles posed to research in this area, (2) especially address the problems of missing German language resources and privacy implications, and (3) identify the steps necessary to overcome these hurdles and fuel research in semantic integration of textual clinical data.

Reduction of spectral interferences and noise effects in laser ablation molecular isotopic spectrometry with partial least square regression – a computer simulation study

The fundamental analytical accuracies and precisions attainable by laser ablation molecular isotopic spectrometry (LAMIS), with emphasis on the impacts from spectral interferences and measurement noise, were investigated by means of computer simulation. The study focused on the analysis of a minor isotope at sub- to single-percentage abundance level. With a natural abundance about 1.1% for 13C, the C2 Swan band (d3Πg–a3Πu) with Δν=+1 was selected as a representative system. The characteristics (e.g., noise amplitude and distribution, signal strength, and signal-to-background ratio) of the simulated spectra were experimentally characterized. Partial least square (PLS) regression was used to extract isotopic information from the simulated molecular spectra. In the absence of any spectral interference and with the use of a calibration set consisting of eleven isotopic standards, the theoretical accuracies and precisions with signal accumulation from 100 laser shots are about 0.002% and 0.001%, respectively, in absolute percentage abundance of 13C. The theoretical analytical accuracies slightly degrade, but are adequate for many applications, to 0.004% and 0.008% respectively, for calibrations involving only three and two isotopic standards. It was found that PLS regression is not only immune to both source-flicker and photon-shot noise, but is also effective in differentiating the spectral patterns from the analyte against those from spectral interferences. The influences of spectral interference from single or multiple atomic emission lines were simulated, and new ways to minimize their impacts were formulated and demonstrated. It was found that the wavelength range selected for the computation of the normalization factor should not contain any spectral-interfering peak, and a properly chosen wavelength range increases the tolerance of spectral interference by at least one order of magnitude. With matrix-matched calibration standards, the precisions (expressed as RSDs of the determined 13C isotopic abundances) degrade from ~1‰ in the absence of spectral interference, to ~3‰, 10‰ and 20‰ with multiple spectral-interfering peaks that are 10×, 100× and 1000×, respectively, stronger than the molecular bandhead of the analyte. The study concluded that PLS regression is a powerful and indispensable tool for extraction of isotopic information from LAMIS spectra.

Improvement of the size estimation of 3D tracked droplets using digital in-line holography with joint estimation reconstruction

Digital holography is a valuable tool for three-dimensional information extraction. Among existing configurations, the originally proposed set-up (i.e. Gabor, or in-line holography), is reasonably immune to variations in the experimental environment making it a method of choice for studies of fluid dynamics. Nevertheless, standard hologram reconstruction techniques, based on numerical light back-propagation are prone to artifacts such as twin images or aliases that limit both the quality and quantity of information extracted from the acquired holograms. To get round this issue, the hologram reconstruction as a parametric inverse problem has been shown to accurately estimate 3D positions and the size of seeding particles directly from the hologram. To push the bounds of accuracy on size estimation still further, we propose to fully exploit the information redundancy of a hologram video sequence using joint estimation reconstruction. Applying this approach in a bench-top experiment, we show that it led to a relative precision of 0.13% (for a 60 μm diameter droplet) for droplet size estimation, and a tracking precision of .