Julien Plu

A hybrid approach for entity recognition and linking

Numerous research efforts are tackling the entity recognition and entity linking tasks resulting in a large body of literature. One could roughly categorize the proposed approaches in two different strategies: linguistic-based and semantic-based methods. In this paper, we present our participation to the OKE challenge, where we experiment with a hybrid approach, which combines the strength of a linguistic-based method augmented by a high coverage in the annotation obtained by using a large knowledge base as entity dictionary. The main goal of this hybrid approach is to improve the extraction and recognition level to get the best recall in order to apply a pruning step. On the training set, the results are promising and the breakdown figures are comparable with the state of the art performance of top ranked systems. Our hybrid approach has been ranked first to the OKE Challenge on the test set.

Enhancing Entity Linking by Combining NER Models

Numerous entity linking systems are addressing the entity recognition problem by using off-the-shelf NER systems. It is, however, a difficult task to select which specific model to use for these systems, since it requires to judge the level of similarity between the datasets which have been used to train models and the dataset at hand to be processed in which we aim to properly recognize entities. In this paper, we present the newest version of ADEL, our adaptive entity recognition and linking framework, where we experiment with an hybrid approach mixing a model combination method to improve the recognition level and to increase the efficiency of the linking step by applying a filter over the types. We obtain promising results when performing a 4-fold cross validation experiment on the OKE 2016 challenge training dataset. We also demonstrate that we achieve better results that in our previous participation on the OKE 2015 test set. We finally report the results of ADEL on the OKE 2016 test set and we present an error analysis highlighting the main difficulties of this challenge.

The 3cixty Knowledge Base for Expo Milano 2015: Enabling Visitors to Explore the City

In this paper, we present the 3cixty Knowledge Base, which collects and harmonizes descriptions of events, places, transportation facilities and user-generated data such as reviews of the city and Expo site of Milan. This knowledge base is used by a set of web and mobile applications to guide Expo Milano 2015 visitors in the city and in the exhibit, allowing them to find places, satellite events and transportation facilities around Milan. As of July 24th, 2015 the knowledge base contains 18665 unique events, 225821 unique places, 94789 reviews, and 9343 transportation facilities, collected from several static, near- and real time local and global data providers, including Expo Milano 2015 official services and numerous social media platforms. The ontologies used as a backbone for structuring the knowledge base follow a rigorous development method where the design principle has generally been to re-use existing ontologies when they exist. We think that the lessons learned from this development will be useful for similar endeavors in other cities or large events around the world with a similar ecosystem of data provisioning services.

3cixty: Building Comprehensive Knowledge Bases For City Exploration

Abstract Planning a visit to Expo Milano 2015 or simply touring in Milan are activities that require a certain amount of a priori knowledge of the city. In this paper, we present the process of building such comprehensive knowledge bases that contain descriptions of events and activities, places and sights, transportation facilities as well as social activities, collected from numerous static, near- and real-time local and global data providers, including hyper local sources such as the Expo Milano 2015 official services and several social media platforms. Entities in the 3cixty KB are deduplicated, interlinked and enriched using semantic technologies. The 3cixty KB is empowering the ExplorMI 360 multi-device application, which has been officially endorsed by the E015 Technical Management Board and has gained the patronage of Expo Milano in 2015, thus has offered a unique testing scenario for the 20 million visitors along the 6 months of the exhibit. In 2016–2017, new knowledge bases have been created for the cities of London, Madeira and Singapore, as well as for the entire French Cote d’Azur area. The 3cixty KB is accessible at https://kb.3cixty.com/sparql while ExplorMI 360 at https://www.3cixty.com and in the Google Play Store and Apple App Store.

Knowledge Extraction in Web Media: At The Frontier of NLP, Machine Learning and Semantics

We identify two main factors that can cause numerous difficulties when developing a generic entity linking system: i) the amount of data currently available on the Web that do not stop to increase and where a large part comes in the form of natural language texts; ii) the velocity at which data is published that may impose to process streams of text in near real-time. Social media platforms such as Twitter, Facebook or LinkedIn become a reliable source of news and play a key role for being aware of events around the world. Encyclopedia and newspaper articles contain general knowledge of our world and they can be used to explain concepts and known entities. Videos can be associated with subtitles and images may have captions. Depending on where a text comes from, it can have different properties such as a specific language, style of writing or topic. In this research, we present a preliminary framework based on a novel hybrid architecture for an entity linking system, that combines methods from the Natural Language Processing (NLP), information retrieval and semantic fields. In particular, we propose a modular approach in order to be as independent as possible of the text to be processed. Our evaluation suggests that this framework can outperform the state-of-the-art systems or show encouraging results on three datasets: OKE2015, #Micropost 2014 and #Micropost 2015. We identify the current limitations and we provide promising future research directions.

ADEL@OKE 2017: A Generic Method for Indexing Knowledge Bases for Entity Linking

In this paper we report the participation of ADEL to the OKE 2017 challenge. In particular, an adaptive entity recognition and linking framework that combines various extraction methods for improving the recognition level and implements an efficient knowledge base indexing process to increase the performance of the linking step. We detail how we deal with fine-grained entity types, either generic (e.g. Activity, Competition, Animal for Task 2) or domain specific (e.g. MusicArtist, SignalGroup, MusicalWork for Task 3). We also show how ADEL can flexibly link entities from different knowledge bases (DBpedia and MusicBrainz). We obtain promising results on the OKE 2017 challenge test dataset for the first three tasks.