Post-Doc Position at LIMSI-CNRS

Scientific context

The Post-Doc project takes place within the French VoxFactory project (project accredited with the Cap Digital Label) headed by Vocalcom in collaboration with the following other companies: EDF, SAFIG, VECSYS, TEMIS, SINEQUA and the public research institutions: LEM-CNRS and LIMSI-CNRS. Cap Digital is a French Business Cluster for Digital Content. A French competitiveness cluster is to be an industrial pillar for driving competitive economic development and stimulating job creation. It is designed to concentrate within a pre-defined territory or geography a broad range of companies, public and private research institutions and multi-level educational organizations focused on a specific technology-driven industry. The Computer Sciences Laboratory for Mechanics and Engineering Sciences (LIMSI) is a CNRS laboratory (UPR 3251) associated with Paris-6 and Paris-11 French Universities. The laboratory accommodates approximately 120 permanent personnel (researchers, professors and assistant professors, engineers, technicians) and about sixty PhD candidates. It undertakes multidisciplinary research in Sciences and Technologies for Information and Communication. The research fields cover a wide disciplinary spectrum from acoustics to cognition, encompassing spoken language and text processing, vision and virtual.

The VoxFactory project aims at analyzing and modeling of the Customer / Agent interaction in data recorded in call center. Call centers are an integral part of customer service, which in turn is what guarantees repeat customers. For that reason, corporations that put them to extensive use are being provided with tools to help improve their functionality. One of those tools is software that analyzes the vocal interaction of customer service calls. Performing a function known as 'interaction analytics,' this software can listen for keywords that indicate stress such as 'cancellation' or 'late delivery' or just plain 'you're terrible'. Beyond that, it can analyze the emotional content of a conversation and assign algorithmic values to what is monitored. The software will attach values to such components as vocal pitch, intensity, rapidity of speech, intonation and, of course, volume. From these values, emotion detection system can be built. Following the project Infom@gic (CallSurf, Cap Digital project), we propose in VoxFactory to analyze and to model the quality and efficiency of customer/agent interaction, using the technologies of speech processing, emotions detection and text mining. The originality consists in a mixed approach based on a textual analysis of the call, from its automatic transcription, and on the speech signal features for the detection of emotions in particular.

Work program.

The project basically consists of two parts. The first part of the project will exploit the available or to-be-invented methods and techniques from machine learning and pattern recognition in order to design an emotion detection system from acoustic cues (prosodic, spectral cues). Several acoustic libraries already used at LIMSI can serve for extracting cues. As a second task, the PhD is expected to get involved with all of the partners in order to use linguistic cues (topic, concept, words) coming from speech processing and text mining in order to build a mixed approach based on textual analysis and acoustic signal for emotion detection (satisfaction, stress, annoyance…). Efforts will be devoted to help final users to specify their needs in terms of system functionalities. The two parts of the project will interact and evolve through several iterations where prototypes will serve as a proof-of-concept for the project.

To apply, include the following information:

- C.V, - the names and contact email/phone of one or two academic referees who would willing to write letters of recommendation, - and a cover letter describing your research interests.

Salary : around 2000 euros net salary

Date of beginning: February 2009 (24 months)

Contact: the interested persons should directly contact: Laurence Devillers devil@limsi.fr