Técnicas de machine learning para la detección de la negación en textos clínicos en español

En los últimos años han surgido nuevos campos de las tecnologías de la información que exploran el tratamiento de la gran cantidad de datos digitales existentes y cómo transformarlos en conocimiento explícito. Las técnicas de Procesamiento del Lenguaje Natural (NLP) son capaces de extraer información de los textos digitales presentados en forma narrativa. Además, las técnicas de machine learning clasifican instancias o ejemplos en función de sus atributos, en distintas categorías, aprendiendo de otros previamente clasificados. Los textos clínicos son una gran fuente de información no estructurada; en consecuencia, información no explotada en su totalidad. Algunos términos usados en textos clínicos se encuentran en una situación de afirmación, negación, hipótesis o histórica. La detección de esta situación es necesaria para la estructuración de información, pero a su vez tiene una gran complejidad. Extrayendo características lingüísticas de los elementos, o tokens, de los textos mediante NLP; transformando estos tokens en instancias y las características en atributos, podemos mediante técnicas de machine learning clasificarlos con el objetivo de detectar si se encuentran afirmados, negados, hipotéticos o históricos. La selección de los atributos que cada token debe tener para su clasificación, así como la selección del algoritmo de machine learning utilizado son elementos cruciales para la clasificación. Son, de hecho, los elementos que componen el modelo de clasificación. Consecuentemente, este trabajo aborda el proceso de extracción de características, selección de atributos y selección del algoritmo de machine learning para la detección de la negación en textos clínicos en español. Se expone un modelo para la clasificación que, mediante el algoritmo J48 y 35 atributos obtenidos de características lingüísticas (morfológicas y sintácticas) y disparadores de negación, detecta si un token está negado en 465 frases provenientes de textos clínicos con un F-Score del 73%, una exhaustividad del 66% y una precisión del 81% con una validación cruzada de 10 iteraciones. ---ABSTRACT--- New information technologies have emerged in the recent years which explore the processing of the huge amount of existing digital data and its transformation into knowledge. Natural Language Processing (NLP) techniques are able to extract certain features from digital texts. Additionally, through machine learning techniques it is feasible to classify instances according to different categories, learning from others previously classified. Clinical texts contain great amount of unstructured data, therefore information not fully exploited. Some terms (tokens) in clinical texts appear in different situations such as affirmed, negated, hypothetic or historic. Detecting this situation is necessary for the structuring of this data, however not simple. It is possible to detect whether if a token is negated, affirmed, hypothetic or historic by extracting its linguistic features by NLP; transforming these tokens into instances, the features into attributes, and classifying these instances through machine learning techniques. Selecting the attributes each instance must have, and choosing the machine learning algorithm are crucial issues for the classification. In fact, these elements set the classification model. Consequently, this work approaches the features retrieval as well as the attributes and algorithm selection process used by machine learning techniques for the detection of negation in clinical texts in Spanish. We present a classification model which, through J48 algorithm and 35 attributes from linguistic features (morphologic and syntactic) and negation triggers, detects whether if a token is negated in 465 sentences from historical records, with a result of 73% FScore, 66% recall and 81% precision using a 10-fold cross-validation.

Diseño de una herramienta para la aplicación de técnicas de minería de textos

El objetivo del presente trabajo es el estudio, diseño e implementación de una herramienta software, con interfaz gráfica de usuario, que permita aplicar diversas técnicas de análisis de textos de forma simple. Las técnicas de análisis, que serán implementadas en la herramienta, extraerán información de textos escritos en un lenguaje humano, es decir un lenguaje no artificial, y se le presentará al usuario. La herramienta permite la obtención de tres tipos de información: categorías a las que pertenece un texto, dentro de un conjunto de categorías predeterminadas; grupos de textos que son similares entre sí; y la polaridad de opinión expresada en un texto hacia el tema u objeto del que trata, que puede ser neutra, positiva o negativa.---ABSTRACT---The aim of this work is to study, design and implement a software tool, with graphical user interface, which will enable a user to easily apply various text analysis techniques. The techniques implemented in the tool will extract information from texts written in natural language, i.e. a non artificial language, and will present it to the user. The tool will extract three different types of information about a given set of texts: their categories (from a predefined set of categories), groups of similar texts, the polarity of the attitude expressed in the texts towards their topic.

Cerebral organization for language in deaf and hearing subjects: Biological constraints and effects of experience

Cerebral organization during sentence processing in English and in American Sign Language (ASL) was characterized by employing functional magnetic resonance imaging (fMRI) at 4 T. Effects of deafness, age of language acquisition, and bilingualism were assessed by comparing results from (i) normally hearing, monolingual, native speakers of English, (ii) congenitally, genetically deaf, native signers of ASL who learned English late and through the visual modality, and (iii) normally hearing bilinguals who were native signers of ASL and speakers of English. All groups, hearing and deaf, processing their native language, English or ASL, displayed strong and repeated activation within classical language areas of the left hemisphere. Deaf subjects reading English did not display activation in these regions. These results suggest that the early acquisition of a natural language is important in the expression of the strong bias for these areas to mediate language, independently of the form of the language. In addition, native signers, hearing and deaf, displayed extensive activation of homologous areas within the right hemisphere, indicating that the specific processing requirements of the language also in part determine the organization of the language systems of the brain.

The roles of language processing in a spoken language interface.

This paper provides an overview of the colloquium's discussion session on natural language understanding, which followed presentations by M. Bates [Bates, M. (1995) Proc. Natl. Acad. Sci. USA 92, 9977-9982] and R. C. Moore [Moore, R. C. (1995) Proc. Natl. Acad. Sci. USA 92, 9983-9988]. The paper reviews the dual role of language processing in providing understanding of the spoken input and an additional source of constraint in the recognition process. To date, language processing has successfully provided understanding but has provided only limited (and computationally expensive) constraint. As a result, most current systems use a loosely coupled, unidirectional interface, such as N-best or a word network, with natural language constraints as a postprocess, to filter or resort the recognizer output. However, the level of discourse context provides significant constraint on what people can talk about and how things can be referred to; when the system becomes an active participant, it can influence this order. But sources of discourse constraint have not been extensively explored, in part because these effects can only be seen by studying systems in the context of their use in interactive problem solving. This paper argues that we need to study interactive systems to understand what kinds of applications are appropriate for the current state of technology and how the technology can move from the laboratory toward real applications.

An approach based on Logic Forms and WordNet relationships to Textual Entailment performance

This paper outlines the approach adopted by the PLSI research group at University of Alicante in the PASCAL-2006 second Recognising Textual Entailment challenge. Our system is composed of several components. On the one hand, the first component performs the derivation of the logic forms of the text/hypothesis pairs and, on the other hand, the second component provides us with a similarity score given by the semantic relations between the derived logic forms. In order to obtain this score we apply several measures of similitude and relatedness based on the structure and content of WordNet.

PHORA: a system to solve the anaphora in Spanish

In this paper we present a whole Natural Language Processing (NLP) system for Spanish. The core of this system is the parser, which uses the grammatical formalism Lexical-Functional Grammars (LFG). Another important component of this system is the anaphora resolution module. To solve the anaphora, this module contains a method based on linguistic information (lexical, morphological, syntactic and semantic), structural information (anaphoric accessibility space in which the anaphor obtains the antecedent) and statistical information. This method is based on constraints and preferences and solves pronouns and definite descriptions. Moreover, this system fits dialogue and non-dialogue discourse features. The anaphora resolution module uses several resources, such as a lexical database (Spanish WordNet) to provide semantic information and a POS tagger providing the part of speech for each word and its root to make this resolution process easier.

Going beyond traditional QA systems: challenges and keys in opinion question answering

The treatment of factual data has been widely studied in different areas of Natural Language Processing (NLP). However, processing subjective information still poses important challenges. This paper presents research aimed at assessing techniques that have been suggested as appropriate in the context of subjective - Opinion Question Answering (OQA). We evaluate the performance of an OQA with these new components and propose methods to optimally tackle the issues encountered. We assess the impact of including additional resources and processes with the purpose of improving the system performance on two distinct blog datasets. The improvements obtained for the different combination of tools are statistically significant. We thus conclude that the proposed approach is adequate for the OQA task, offering a good strategy to deal with opinionated questions.

Prueba de concepto de expansión de consultas basada en ontologías de dominio financiero

Este trabajo presenta el uso de una ontología en el dominio financiero para la expansión de consultas con el fin de mejorar los resultados de un sistema de recuperación de información (RI) financiera. Este sistema está compuesto por una ontología y un índice de Lucene que permite recuperación de conceptos identificados mediante procesamiento de lenguaje natural. Se ha llevado a cabo una evaluación con un conjunto limitado de consultas y los resultados indican que la ambigüedad sigue siendo un problema al expandir la consulta. En ocasiones, la elección de las entidades adecuadas a la hora de expandir las consultas (filtrando por sector, empresa, etc.) permite resolver esa ambigüedad.

COMPENDIUM: a text summarisation tool for generating summaries of multiple purposes, domains, and genres

In this paper, we present a Text Summarisation tool, compendium, capable of generating the most common types of summaries. Regarding the input, single- and multi-document summaries can be produced; as the output, the summaries can be extractive or abstractive-oriented; and finally, concerning their purpose, the summaries can be generic, query-focused, or sentiment-based. The proposed architecture for compendium is divided in various stages, making a distinction between core and additional stages. The former constitute the backbone of the tool and are common for the generation of any type of summary, whereas the latter are used for enhancing the capabilities of the tool. The main contributions of compendium with respect to the state-of-the-art summarisation systems are that (i) it specifically deals with the problem of redundancy, by means of textual entailment; (ii) it combines statistical and cognitive-based techniques for determining relevant content; and (iii) it proposes an abstractive-oriented approach for facing the challenge of abstractive summarisation. The evaluation performed in different domains and textual genres, comprising traditional texts, as well as texts extracted from the Web 2.0, shows that compendium is very competitive and appropriate to be used as a tool for generating summaries.

PoS-tagging the Web in Portuguese. National varieties, text typologies and spelling systems

The great amount of text produced every day in the Web turned it as one of the main sources for obtaining linguistic corpora, that are further analyzed with Natural Language Processing techniques. On a global scale, languages such as Portuguese - official in 9 countries - appear on the Web in several varieties, with lexical, morphological and syntactic (among others) differences. Besides, a unified spelling system for Portuguese has been recently approved, and its implementation process has already started in some countries. However, it will last several years, so different varieties and spelling systems coexist. Since PoS-taggers for Portuguese are specifically built for a particular variety, this work analyzes different training corpora and lexica combinations aimed at building a model with high-precision annotation in several varieties and spelling systems of this language. Moreover, this paper presents different dictionaries of the new orthography (Spelling Agreement) as well as a new freely available testing corpus, containing different varieties and textual typologies.

Proyecto FIRST (Flexible Interactive Reading Support Tool): desarrollo de una herramienta para ayudar a personas con autismo mediante la simplificación de textos

El Trastorno de Espectro Autista (TEA) es un trastorno que impide el correcto desarrollo de funciones cognitivas, habilidades sociales y comunicativas en las personas. Un porcentaje significativo de personas con autismo presentan además dificultades en la comprensión lectora. El proyecto europeo FIRST está orientado a desarrollar una herramienta multilingüe llamada Open Book que utiliza Tecnologías del Lenguaje Humano para identificar obstáculos que dificultan la comprensión lectora de un documento. La herramienta ayuda a cuidadores y personas con autismo transformando documentos escritos a un formato más sencillo mediante la eliminación de dichos obstáculos identificados en el texto. En este artículo se presenta el proyecto FIRST así como la herramienta desarrollada Open Book.

ATTOS: Análisis de Tendencias y Temáticas a través de Opiniones y Sentimientos

El proyecto ATTOS centra su actividad en el estudio y desarrollo de técnicas de análisis de opiniones, enfocado a proporcionar toda la información necesaria para que una empresa o una institución pueda tomar decisiones estratégicas en función a la imagen que la sociedad tiene sobre esa empresa, producto o servicio. El objetivo último del proyecto es la interpretación automática de estas opiniones, posibilitando así su posterior explotación. Para ello se estudian parámetros tales como la intensidad de la opinión, ubicación geográfica y perfil de usuario, entre otros factores, para facilitar la toma de decisiones. El objetivo general del proyecto se centra en el estudio, desarrollo y experimentación de técnicas, recursos y sistemas basados en Tecnologías del Lenguaje Humano (TLH), para conformar una plataforma de monitorización de la Web 2.0 que genere información sobre tendencias de opinión relacionadas con un tema.

A Theorical Point of View of Reality, Perception, and Language

It is possible to view the relations between mathematics and natural language from different aspects. This relation between mathematics and language is not based on just one aspect. In this article, the authors address the role of the Subject facing Reality through language. Perception is defined and a mathematical theory of the perceptual field is proposed. The distinction between purely expressive language and purely informative language is considered false, because the subject is expressed in the communication of a message, and conversely, in purely expressive language, as in an exclamation, there is some information. To study the relation between language and reality, the function of ostensibility is defined and propositions are divided into ostensives and estimatives.

Lightweight Internet Traffic Classification - A Subject Based Solution with Word Embeddings

Internet traffic classification is a relevant and mature research field, anyway of growing importance and with still open technical challenges, also due to the pervasive presence of Internet-connected devices into everyday life. We claim the need for innovative traffic classification solutions capable of being lightweight, of adopting a domain-based approach, of not only concentrating on application-level protocol categorization but also classifying Internet traffic by subject. To this purpose, this paper originally proposes a classification solution that leverages domain name information extracted from IPFIX summaries, DNS logs, and DHCP leases, with the possibility to be applied to any kind of traffic. Our proposed solution is based on an extension of Word2vec unsupervised learning techniques running on a specialized Apache Spark cluster. In particular, learning techniques are leveraged to generate word-embeddings from a mixed dataset composed by domain names and natural language corpuses in a lightweight way and with general applicability. The paper also reports lessons learnt from our implementation and deployment experience that demonstrates that our solution can process 5500 IPFIX summaries per second on an Apache Spark cluster with 1 slave instance in Amazon EC2 at a cost of $ 3860 year. Reported experimental results about Precision, Recall, F-Measure, Accuracy, and Cohen's Kappa show the feasibility and effectiveness of the proposal. The experiments prove that words contained in domain names do have a relation with the kind of traffic directed towards them, therefore using specifically trained word embeddings we are able to classify them in customizable categories. We also show that training word embeddings on larger natural language corpuses leads improvements in terms of precision up to 180%.

Information extraction from clinical and radiology notes for liver cancer staging

Thesis (Ph.D.)--University of Washington, 2016-06