< Back
Posted: 17 Apr 2022 01:00

“Dialogue Summarization” April 2022 — summary from Astrophysics Data System and Arxiv

Brevi Assistant
Brevi Assistant

Business performance assistant

“Dialogue Summarization” April 2022 — summary from Astrophysics Data System and Arxiv main image

The content below is machine-generated by Brevi Technologies’ NLG model, and the source content was collected from open-source databases/integrate APIs.


Astrophysics Data System - summary generated by Brevi Assistant


We present ClidSum, a benchmark dataset for constructing cross-lingual summarization systems on dialogue documents. It contains 67k+ dialogue documents from two parts and 112k+ annotated summaries in different target languages. The numerous goals made use of in the more pre-training phase aid the pre-trained model record the structural features as well as important content in dialogues and the transformation from source to the target language.

Annotating task-oriented dialogues is well-known for the difficult and costly information collection process. In this paper, we assume that dialogue summaries are essentially unstructured dialogue states; For this reason, we recommend reformulate dialogue state monitoring as a dialogue summarization problem.

We empirically show that our approach to DS2 outshines previous works on few-shot DST in MultiWoZ 2. 0 and 2. 1, in both cross-domain and multi-domain settings. One of the most advanced abstractive dialogue summarizers lacks generalization capacity on new domains and the existing investigates for domain adaptation in summarization usually rely on large-scale pre-trainings. To explore the light-weight fine-tuning techniques for domain adaptation of dialogue summarization, in this paper, we recommend an effective and generalizable Domain-Oriented Prefix-tuning model, which makes use of a domain word booted up prefix module to alleviate domain complication and takes on distinct motivates to lead the model to concentrate on crucial components of dialogues and enhance model generalization.

We perform zero-shot experiments and construct domain adaptation standards on two multi-domain dialogue summarization datasets, TODSum and QMSum.


Source texts:



Arxiv - summary generated by Brevi Assistant


We present ClidSum, a benchmark dataset for building cross-lingual summarization systems on dialogue documents. It consists of 67k+ dialogue documents from 2 parts and 112k+ annotated summaries in different target languages. The numerous purposes made use of in the further pre-training phase aid the pre-trained model capture the structural qualities in addition to essential content in dialogues and the transformation from source to the target language.

Annotating task-oriented discussions is well-known for the pricey and challenging information collection procedure. In this paper, we hypothesize that dialogue summaries are basically disorganized dialogue states; thus, we suggest reformulate dialogue state monitoring as a dialogue summarization issue.

We empirically reveal that our technique DS2 surpasses previous works on few-shot DST in MultiWoZ 2. 0 and 2. 1, in both cross-domain and multi-domain setups. One of the most advanced abstractive dialogue summarizers lacks generalization capacity on new domains and the existing looks into domain adjustment in summarization normally rely on large-scale pre-trainings. To check out the light-weight fine-tuning approaches for domain adaptation of dialogue summarization, in this paper, we propose a generalizable and reliable Domain-Oriented Prefix-tuning model, which uses a domain word initialized prefix component to reduce domain entanglement and embraces distinct prompts to direct the model to focus on crucial components of dialogues and boost model generalization.

We perform zero-shot experiments and construct domain adjustment standards on 2 multi-domain dialogue summarization datasets, TODSum and QMSum. Text summarization assists readers catch significant information from documents, information, meetings, and meetings. Our structure can refine input text of approximate length by readjusting the variety of stages while keeping the LM input dimension dealt with.

Our experiments show that Summ ^N surpasses previous cutting edge approaches by enhancing ROUGE scores on 3 long meeting summarization datasets AMI, ICSI, and QMSum, 2 lengthy TV collection datasets from SummScreen, and a long document summarization dataset, GovReport.


This can serve as an example of how to use Brevi Assistant and integrated APIs to analyze text content.


Source texts:


logo

The Brevi assistant is a novel way to summarize, assemble, and consolidate multiple text documents/contents.

Partners:

© All rights reserved 2022 made by Brevi Technologies