Altran

About: Altran is a company organization based out in Neuilly-sur-Seine, France. It is known for research contribution in the topics: Software development & Formal verification. The organization has 488 authors who have published 512 publications receiving 6395 citations.

Joint Intent Detection and Slot Filling via CNN-LSTM-CRF

Abstract: Intent detection and slot filling are two main tasks in the domain of Spoken Language Understanding (SLU) The methods employed may treat the intent detection and slot filling as two independent tasks or use a joint model Using a joint model takes into account the cross impact between the two tasks In this article, we introduce CoBiC a new model combining CNN (Convolutional Neural Network), Bidirectional LSTM (Long Short-Term Memory) and CRF (Conditional Random Field) to extract the intents and the related slots The same architecture of CoBiC can either be used as an independent model or joint model for intent detection and slot filling Our method improves the state-of-the-art results on ATIS (Airline Travel Information Systems) benchmark We also apply our model on a private dataset consisting of clients requests to a vocal assistant The results demonstrate that CoBiC has strong generalization capability

Schätzen in Normalverteilungsmodellen

Abstract: Kann eine Zufallsmessung prinzipiell einen beliebigen Wert auf einer metrischen Zahlenskala annehmen, spricht man von einer stetigen Zufallsvariablen. Die Korpertemperatur oder der systolische Blutdruck von Probanden einer klinischen Studie sind typischerweise stetige Zufallsvariablen. Zeit- und Langenmessungen sind ebenfalls typische Beispiele. Dass innerhalb eines Intervalls jeder beliebige Wert angenommen werden kann, ist eine theoretische Annahme und unserer Vorstellung von „Kontinuitat” geschuldet. In der Praxis werden die Daten naturgemas aufgrund der begrenzten Messgenauigkeit und unvermeidbaren Rundungen immer in diskreter Form vorliegen. Trotzdem ist es sinnvoll, bei der Modellierung die Merkmale als stetige Zufallsvariable aufzufassen.

Object Oriented Design Knowledge: Ontology and Measurement of Impact

Abstract: It has been a long time since appeared of the Object Oriented (OO) paradigm. From that moment, the designers have accumulated much knowledge in design and construction of OO systems. Patterns are the most refined OO Design Knowledge. However, there are many others kinds of knowledge than not yet classified and formalized. We distinguish and classify the following categories: principles, heuristics, patterns and refactorings. In this paper, we propose an Ontology for Object Oriented Design Knowledge and a measure of impact for patterns, two key elements to create a method based in knowledge.

“ADAPEI-TRANSPORT”: A GPS Based Mobile App for Learning Paths and Improving Autonomy for Young Adults Having Intellectual Disabilities to Take Public Transport

Abstract: Children and young adults with intellectual disabilities (from 10–20 years old) are trained by educators for their journeys taking public transport (buses). This training is done teaching them the different steps and actions to do in a sequential way using photos, pictograms, times, texts (e.g. walk to the bus stop Roppe and take the line 21 direction Republique, do not forget to press on the stop button to alert to take off, take off at the stop station Republique, etc.). Currently, specialists from ADAPEI association in Belfort (France) use booklets in which all the steps to follow are printed. With the increasing of the use of mobile apps in Assistive Technology, the aim of this partnership between Altran and ADAPEI from Belfort is to develop new supports to teach children (from the least to the most autonomous) the paths using GPS mobile and tablet apps before and while doing the journey. Moreover, for the most autonomous children familiarized with the use of apps, two ways of navigation using GPS have been developed. Either using maps showing the path or without using maps showing strategically the actions to do in an ordered way when the user is close to a point of interest. In this article, we will show the first results of the use of the app as a teaching system based on observations with disabled children in different situations, and the difficulties found to detect change of directions in noisy GPS data in cities.