Semantic Relation Extraction with Semantic Patterns Experiment on Radiology Reports

Semantic RelationExtraction with Semantic Patterns: Experiment on Radiology Report Mathieu Lafourcade, Lionel Ramadier LIRMM, University of Montpellier 860, rue de St Priest, 34095 Montpellier E-mail: Mathieu.lafourcade@lirmm.fr, lionel.ramadier@lirmm.fr Abstract This work presents a practical system for indexing terms and relations from French radiology reports, called IMAIOS. In this paper, we present how semantic relations (causes, consequences, symptoms, locations, parts…) between medical terms can be extracted. For this purpose, we handcrafted some linguistic patterns from on a subset of our radiology report corpora. As semantic patterns (de (of)) may be too general or ambiguous, semantic constraints have been added. For instance, in the sentence néoplasie du sein (neoplasm of breast) the system knowing neoplasm as a disease and breast as an anatomical location, identify the relation as being a location: neoplasm r-lieu breast. An evaluation of the effect of semantic constraints is proposed. Keywords: Relation Extraction, Semantic constraints, NLP 1. Introduction In the domain of radiology, the amount of pictures and textual reports is growing at an unprecedented rate. This quantity of medical information exceeds the ability of the radiologist to manage it effectively. One challenge is then to figure how to make computer help to effectively use these findings and to improve the management of patients. NLP methods allow for enhanced indexing and searching of medical reports with the linking of relevant terms through semantic relations and enabling on the fly inference of new knowledge. This work presents a practical system for indexing terms and relations from French radiology reports, called IMAIOS. The main task of this system is to extract semantic relations from unstructured text that can be integrated in a lexical-semantic network in order to improve the quality of the latter. The IMAIOS system differs from other techniques by the use of the french lexical-semantic JeuxDeMots (JDM) network as supporting resource. Although the JDM network is primarily a knowledge base of common sense, it contains also many specialized facts, including medicine/radiology, which have been added within the framework of IMAIOS system. With such a large lexico-semantic network, we can devise quite crude algorithms that are very efficient. In this paper, we present how semantic relations (causes, consequences, symptoms, locations, parts…) between medical terms can be extracted. For this purpose, we handcrafted some linguistic patterns from on a subset of our radiology report corpora. As semantic patterns (de (of)) may be too general or ambiguous, semantic constraints have been added. For instance, in the sentence néoplasie du sein (neoplasm of breast) the system knowing neoplasm as a disease and breast as an anatomical location, identify the relation as being a location: neoplasm r-lieu breast. An evaluation of the effect of semantic constraints is proposed. 2. Background Most work concerning the extraction of semantic relation focus on domain-independent relations (Snow et al., 2006; Chklovski et al., 2004). In the general domain, the extraction of semantic relations between entities uses either statistical approaches (Hindle et al., 1990; Nazar et al 2012) either techniques of machine learning as well as approaches based on …

Content truncated to 500 words