R&D
Cutting-edge NLP solutions
Our team consists of six PhDs and has produced around 10 scientific papers in the field of Natural Language Processing (NLP). We are contributing to the development of state-of-the-art NLP technology.

reciTAL and its research lab actively contribute to new global advances in the field of NLP and regularly publish new papers in the most prestigious conferences such as EMNLP, ACL or ICML.
Our main areas of research are:
Document Layout Understanding: Using models combining textual and visual features to automatically detect the internal layout of documents and facilitate document analysis and use.
Active learning et small data: Reducing the number of examples needed to improve the performance of models.
Question Answering et Question Generation: The full range of approaches involving the transition from keyword questions to natural language questions.
R&D Partnerships
1 Scientific Advisory Board: Stuart Russell, Antoine Bordes.
1 Academic Chair: Joint reciTAL-ESILV (Ecole Supérieure d’Ingénieurs Léonard de Vinci) Artificial Intelligence Chair.
3 CIFRE theses: aWith LIP6 (Paris VI Sorbonne Université and CNRS) and ESILV in the fields of Text Generation, Document Layout Understanding and Question Answering.
1 supercomputer: GENCI partnership to access the Jean Zay supercomputer installed and operated within the Institut du Développement et des Ressources en Informatique Scientifique at the CNRS (IDRIS) in Orsay, France.
SCIENTIFIC ADVISORY BOARD

« reciTAL is one of a small number of companies that is developing the next generation of natural langage technology »

« reciTAL is currently positioned in the exact market sectors where AI can help businesses now and make an impact on their operations. »
Scientific publications
-
- ARXIV
- GENERATIVE AI
-
- ARXIV
- GENERATIVE AI
Generative Cooperative Networks for Natural Language Generation
-
- GAN
- NeurIPS
- SUMMARISATION
To Beam Or Not To Beam: That is a Question of Cooperation for Language GANs
-
- EMNLP
- EVALUATION
-
- EMNLP
- EVALUATION
Data-QuestEval: A Referenceless Metric for Data to Text Semantic Evaluation
-
- EMNLP
- EVALUATION
- SUMMARISATION
-
- EMNLP
-
- EMNLP
- MULTILINGUAL
- QUESTION ANSWERING
Synthetic Data Augmentation for Zero-Shot Cross-Lingual Question Answering
-
- GAN
- NeurIPS
ColdGANs: Taming Language GANs with Cautious Sampling Strategies
-
- EMNLP
- MULTILINGUAL
- SUMMARISATION
-
- FRANCOPHONIA
- LREC
- QUESTION ANSWERING
Project PIAF: Building a Native French Question-Answering Dataset
-
- NER
Complex Named Entities Extraction on the Web: Application to Social Events
-
- ICML
- NATURAL LANGUAGE GENERATION
Discriminative Adversarial Search for Abstractive Summarization
-
- ARXIV
- TRANSFER LEARNING
BERT Can See Out of the Box: On the Cross-modal Transferability of Text Representations
-
- EMNLP
- EVALUATION
- NATURAL LANGUAGE GENERATION
Answers Unite! Unsupervised Metrics for Reinforced Summarization Models
-
- IEEE
- SENTIMENT ANALYSIS
- Transactions on Affective Computing
DepecheMood++: a Bilingual Emotion Lexicon Built Through Simple Yet Powerful Techniques
-
- ACL
- EVALUATION
- NATURAL LANGUAGE GENERATION
- QUESTION ANSWERING
Self-Attention Architectures for Answer-Agnostic Neural Question Generation