Claire Nédellec

Knowledge Acquisition of Predicate Argument Structures from Technical Texts Using Machine Learning: The System ASIUM

In this paper, we describe the Machine Learning system, ASIUM, which learns Subcaterorization Frames of verbs and ontologies from the syntactic parsing of technical texts in natural language. The restrictions of selection in the subcategorization frames are filled by the ontologys concepts. Applications requiring such knowledge are crucial and numerous. The most direct applications are semantic control of texts and syntactic parsing disambiguation. This knowledge acquisition task cannot be fully automatically performed. Instead,we propose a cooperative ML method which provides the user with a global view of the acquisition task and also with acquisition tools like automatic concepts splitting, example generation, and an ontology view with attachments to the verbs. Validation steps using these features are intertwined with learning steps so that the user validates the concepts as they are learned. Experiments performed on two different corpora (cooking domain and patents) give very promising results.

Sentence Filtering for Information Extraction in Genomics, a Classification Problem

In some domains, Information Extraction (IE) from texts requires syntactic and semantic parsing. This analysis is computationally expensive and IE is potentially noisy if it applies to the whole set of documents when the relevant information is sparse. A preprocessing phase that selects the fragments which are potentially relevant increases the efficiency of the IE process. This phase has to be fast and based on a shallow description of the texts. We applied various classification methods -- IVI, a Naive Bayes learner and C4.5 -- to this fragment filtering task in the domain of functional genomics. This paper describes the results of this study. We show that the IVI and Naive Bayes methods with feature selection gives the best results as compared with their results without feature selection and with C4.5 results.

Event-based information extraction for the biomedical domain: the Caderige project

This paper gives an overview of the Caderige project. This project involves teams from different areas (biology, machine learning, natural language processing) in order to develop highlevel analysis tools for extracting structured information from biological bibliographical databases, especially Medline. The paper gives an overview of the approach and compares it to the state of the art.

Knowledge refinement using knowledge acquisition and machine learning methods

APT system integrates Machine Learning (ML) and Knowledge Acquisition (KA) methods in the same framework. Both kinds of methods closely cooperate to concur in the same purpose: the acquisition, validation and maintenance of problem-solving knowledge. The methods are based on the same assumption: knowledge acquisition and learning are done through experimentation, classification and comparison of concrete cases. This paper details APTs mechanisms and shows through examples and applications how APT underlying principles allow various methods to fruitfully collaborate.

ATOL: The Multi-species Livestock Trait Ontology

This paper presents the multi-species Animal Trait Ontology for Livestock (ATOL) and the methodology used for its design. ATOL has been designed as a reference source for indexing phenotype databases and scientific papers. It covers five major topics related to animal productions: growth and meat quality, animal nutrition, milk production, reproduction and welfare. It is composed of species-independent concepts subsuming species-specific ones so that cross-species and species-specific reasoning can be performed consistently. In order to ensure a large consensus, three complementary approaches have successively been applied to its design: reuse of existing ontologies, integration of production-specific livestock traits by a large team of domain experts and curators and terminology analysis of scientific papers. It resulted in a detailed taxonomy of 1,654 traits that is available at http://www.atol-ontology.com

Overview of the Bacteria Biotope Task at BioNLP Shared Task 2016.

This paper presents the Bacteria Biotope task of the BioNLP Shared Task 2016, which follows the previous 2013 and 2011 editions. The task focuses on the extraction of the locations (biotopes and geographical places) of bacteria from PubMe abstracts and the characterization of bacteria and their associated habitats with respect to reference knowledge sources (NCBI taxonomy, OntoBiotope ontology). The task is motivated by the importance of the knowledge on bacteria habitats for fundamental research and applications in microbiology. The paper describes the different proposed subtasks, the corpus characteristics, the challenge organization, and the evaluation metrics. We also provide an analysis of the results obtained by participants.

Building large lexicalized ontologies from text: a use case in automatic indexing of biotechnology patents

This paper presents a tool, TyDI, and methods experimented in the building of a termino-ontology, i.e. a lexicalized ontology aimed at fine-grained indexation for semantic search applications. TyDI provides facilities for knowledge engineers and domain experts to efficiently collaborate to validate, organize and conceptualize corpus extracted terms. A use case on biotechnology patent search demonstrates TyDIs potential.

Overview of the Regulatory Network of Plant Seed Development (SeeDev) Task at the BioNLP Shared Task 2016.

This paper presents the SeeDev Task of the BioNLP Shared Task 2016. The purpose of the SeeDev Task is the extraction from scientific articles of the descriptions of genetic and molecular mechanisms involved in seed development of the model plant, Arabidopsis thaliana. The SeeDev task consists in the extraction of many different event types that involve a wide range of entity types so that they accurately reflect the complexity of the biological mechanisms. The corpus is composed of paragraphs selected from the full-texts of relevant scientific articles. In this paper, we describe the organization of the SeeDev task, the corpus characteristics, and the metrics used for the evaluation of participant systems. We analyze and discuss the final results of the seven participant systems to the test. The best F-score is 0.432, which is similar to the scores achieved in similar tasks on molecular biology.

MACHINE LEARNING GOES TO THE BANK

This paper describes how the APT system has been applied to loan analysis to generalized and refine the knowledge previously used by an expert system, in order to increase the efficiency and the compactness of the decision rule base. The decision to lend money to industrial companies is a complex and risky activity for financial institutions. They need much expertise to deal with the large amount of information that has to be considered for this process, and the analysis must be carefully done in order to avoid misjudgments that would result in severe losses of unrecoverable credit. An expert system named SPAC had been developed to deal with this task without fulfilling the users expectations. This paper presents the drawbacks of SPACs approach and how APT, an integrated machine learning system, has been used to acquire and refine domain knowledge and general decision rules from basic descriptions of cases provided by SPAC. The learning methodology is detailed, and a complete example of a learning session with APT is given. The final results are then compared with those obtained with SPAC

Representation of complex terms in a vector space structured by an ontology for a normalization task

We propose in this paper a semisupervised method for labeling terms of texts with concepts of a domain ontology. The method generates continuous vector representations of complex terms in a semantic space structured by the ontology. The proposed method relies on a distributional semantics approach, which generates initial vectors for each of the extracted terms. Then these vectors are embedded in the vector space constructed from the structure of the ontology. This embedding is carried out by training a linear model. Finally, we apply a cosine similarity to determine the proximity between vectors of terms and vectors of concepts and thus to assign ontology labels to terms. We have evaluated the quality of these representations for a normalization task by using the concepts of an ontology as semantic labels. Normalization of terms is an important step to extract a part of the information contained in texts, but the vector space generated might find other applications. The performance of this method is comparable to that of the state of the art for this task of standardization, opening up encouraging prospects.