Karel Ondrej

Martin Fajcik, Martin Docekal, Jan Dolezal et al.

We present BenCzechMark (BCM), the first comprehensive Czech language benchmark designed for large language models, offering diverse tasks, multiple task formats, and multiple evaluation metrics. Its duel scoring system is grounded in statistical significance theory and uses aggregation across tasks inspired by social preference theory. Our benchmark encompasses 50 challenging tasks, with corresponding test datasets, primarily in native Czech, with 14 newly collected ones. These tasks span 8 categories and cover diverse domains, including historical Czech news, essays from pupils or language learners, and spoken word. Furthermore, we collect and clean BUT-Large Czech Collection, the largest publicly available clean Czech language corpus, and use it for (i) contamination analysis and (ii) continuous pretraining of the first Czech-centric 7B language model with Czech-specific tokenization. We use our model as a baseline for comparison with publicly available multilingual models. Lastly, we release and maintain a leaderboard with existing 50 model submissions, where new model submissions can be made at https://huggingface.co/spaces/CZLC/BenCzechMark.

Juan Zuluaga-Gomez, Seyyed Saeed Sarfjoo, Amrutha Prasad et al.

Automatic speech recognition (ASR) allows transcribing the communications between air traffic controllers (ATCOs) and aircraft pilots. The transcriptions are used later to extract ATC named entities, e.g., aircraft callsigns. One common challenge is speech activity detection (SAD) and speaker diarization (SD). In the failure condition, two or more segments remain in the same recording, jeopardizing the overall performance. We propose a system that combines SAD and a BERT model to perform speaker change detection and speaker role detection (SRD) by chunking ASR transcripts, i.e., SD with a defined number of speakers together with SRD. The proposed model is evaluated on real-life public ATC databases. Our BERT SD model baseline reaches up to 10% and 20% token-based Jaccard error rate (JER) in public and private ATC databases. We also achieved relative improvements of 32% and 7.7% in JERs and SD error rate (DER), respectively, compared to VBx, a well-known SD system.

Martin Fajcik, Martin Docekal, Karel Ondrej et al.

This work presents a novel four-stage open-domain QA pipeline R2-D2 (Rank twice, reaD twice). The pipeline is composed of a retriever, passage reranker, extractive reader, generative reader and a mechanism that aggregates the final prediction from all system's components. We demonstrate its strength across three open-domain QA datasets: NaturalQuestions, TriviaQA and EfficientQA, surpassing state-of-the-art on the first two. Our analysis demonstrates that: (i) combining extractive and generative reader yields absolute improvements up to 5 exact match and it is at least twice as effective as the posterior averaging ensemble of the same models with different parameters, (ii) the extractive reader with fewer parameters can match the performance of the generative reader on extractive QA datasets.

Martin Fajcik, Martin Docekal, Karel Ondrej et al.

This work presents a novel pipeline that demonstrates what is achievable with a combined effort of state-of-the-art approaches. Specifically, it proposes the novel R2-D2 (Rank twice, reaD twice) pipeline composed of retriever, passage reranker, extractive reader, generative reader and a simple way to combine them. Furthermore, previous work often comes with a massive index of external documents that scales in the order of tens of GiB. This work presents a simple approach for pruning the contents of a massive index such that the open-domain QA system altogether with index, OS, and library components fits into 6GiB docker image while retaining only 8% of original index contents and losing only 3% EM accuracy.

Sewon Min, Jordan Boyd-Graber, Chris Alberti et al.

We review the EfficientQA competition from NeurIPS 2020. The competition focused on open-domain question answering (QA), where systems take natural language questions as input and return natural language answers. The aim of the competition was to build systems that can predict correct answers while also satisfying strict on-disk memory budgets. These memory budgets were designed to encourage contestants to explore the trade-off between storing retrieval corpora or the parameters of learned models. In this report, we describe the motivation and organization of the competition, review the best submissions, and analyze system predictions to inform a discussion of evaluation for open-domain QA.