How to Do Things With Words

L'article, publie dans Science, sur une des premieres utilisations analytiques de Google Books, fondee sur les n-grammes (Google Ngrams) We constructed a corpus of digitized texts containing about 4% of all books ever printed. Analysis of this corpus enables us to investigate cultural trends quantitatively. We survey the vast terrain of "culturomics", focusing on linguistic and cultural phenomena that were reflected in the English language between 1800 and 2000. We show how this approach can ...

Quantitative Analysis of Culture Using Millions of Digitized Books

The past four years have seen the rise of conversational agents (CAs) in everyday life. Apple, Microsoft, Amazon, Google and Facebook have all embedded proprietary CAs within their software and, increasingly, conversation is becoming a key mode of human-computer interaction. Whilst we have long been familiar with the notion of computers that speak, the investigative concern within HCI has been upon multimodality rather than dialogue alone, and there is no sense of how such interfaces are used in everyday life. This paper reports the findings of interviews with 14 users of CAs in an effort to understand the current interactional factors affecting everyday use. We find user expectations dramatically out of step with the operation of the systems, particularly in terms of known machine intelligence, system capability and goals. Using Norman's 'gulfs of execution and evaluation' [30] we consider the implications of these findings for the design of future systems.

/pdf/like-having-a-really-bad-pa-the-gulf-between-user-3jl39tt7wd.pdf

"Like Having a Really Bad PA": The Gulf between User Expectation and Experience of Conversational Agents

How To Do Things With Words.

Julius is an open-source large-vocabulary speech recognition software used for both academic research and industrial applications. It executes real-time speech recognition of a 60k-word dictation task on low-spec PCs with small footprint, and even on embedded devices. Julius supports standard language models such as statistical N-gram model and rule-based grammars, as well as Hidden Markov Model (HMM) as an acoustic model. One can build a speech recognition system of his own purpose, or can integrate the speech recognition capability to a variety of applications using Julius. This article describes an overview of Julius, major features and specifications, and summarizes the developments conducted in the recent years.

/pdf/recent-development-of-open-source-speech-recognition-engine-5ckfml6tlw.pdf

Recent Development of Open-Source Speech Recognition Engine Julius

We study in this work the importance of depth in convolutional models for text classification, either when character or word inputs are considered. We show on 5 standard text classification and sentiment analysis tasks that deep models indeed give better performances than shallow networks when the text input is represented as a sequence of characters. However, a simple shallow-and-wide network outper-forms deep models such as DenseNet with word inputs. Our shallow word model further establishes new state-of-the-art performances on two datasets: Yelp Binary (95.9%) and Yelp Full (64.9%).

https://hal.archives-ouvertes.fr/hal-01690601/document

Do Convolutional Networks Need to Be Deep for Text Classification

On noise masking for automatic missing data speech recognition: A survey and discussion

Because of license restrictions, it often becomes impossible to strictly reproduce most research results on Twitter data already a few months after the creation of the corpus. This situation worsened gradually as time passes and tweets become inaccessible. This is a critical issue for reproducible and accountable research on social media. We partly solve this challenge by annotating a new Twitter-like corpus from an alternative large social medium with licenses that are compatible with reproducible experiments: Mastodon. We manually annotate both dialogues and sentiments on this corpus, and train a multi-task hierarchical recurrent network on joint sentiment and dialog act recognition. We experimentally demonstrate that transfer learning may be efficiently achieved between both tasks, and further analyze some specific correlations between sentiments and dialogues on social media. Both the annotated corpus and deep network are released with an open-source license.

https://hal.archives-ouvertes.fr/hal-01838323/document

Multi-task dialog act and sentiment recognition on Mastodon

La reconnaissance automatique de la parole donne aujourd'hui lieu a un ensemble important d'applications de nature et de difficulte tres variees, concernant quotidiennement des millions de personnes a travers le monde. On peut prevoir que la parole fera de plus en plus partie des interfaces multimedia entre un utilisateur et un systeme automatique, d'une part grâce a l'amelioration de la robustesse des systemes de reconnaissance automatique de la parole et, d'autre part, du fait de la sensibilisation croissante du grand public a cette technologie encore peu connue. Cet ouvrage illustre le caractere pluridisciplinaire de la reconnaissance automatique de la parole qui fait appel a des notions variees allant des mathematiques, du traitement du signal, de l'informatique et de l'intelligence artificielle jusqu'a la phonetique, a la linguistique, aux sciences cognitives et aux neurosciences. Il presente les fondements theoriques du domaine ainsi que differents modeles et algorithmes utilises (statistiques, neuromimetiques, symboliques) en les illustrant par des exemples reels. Les grands domaines d'application sont egalement etudies : dictee, commandes, telematique, traduction, etc. Ce livre s'adresse aux etudiants en master et en doctorat, aux eleves ingenieurs, aux chercheurs, aux professionnels des interfaces homme-machine, ainsi qu'a toute personne ayant un niveau scientifique suffisant.

Reconnaissance Automatique de la Parole Du signal à son interprétation

This paper presents the recent development of ANTS, the Automatic News Transcription System of LORIA This system was designed in the framework of ESTER, the French broadcast radio news transcription task evaluation After describing its different components and some segmentation and recognition results on the ESTER database, we present a number of experiments focusing on the real-time version of ANTS We evaluate the system with different number of Gaussians, sizes of vocabulary and decoder settings The non real time version of ANTS yields an overall error rate of 36 % that can be compared to 445 % for the real time version

Christophe Cerisara

Papers

Do Convolutional Networks Need to Be Deep for Text Classification

On noise masking for automatic missing data speech recognition: A survey and discussion

Multi-task dialog act and sentiment recognition on Mastodon

Reconnaissance Automatique de la Parole Du signal à son interprétation

The Automatic News Transcription System: ANTS some Real Time experiments