Computational Linguistics @ QMUL

31 posts

Computational Linguistics @ QMUL

Computational Linguistics @ QMUL

@QMCompLing

Computational Linguistics Lab @ QMUL; language, meaning, social media, web searches, dialogue systems, creativity, mental health,...

London, UK Katılım Nisan 2020
26 Takip Edilen302 Takipçiler
Computational Linguistics @ QMUL retweetledi
yin, wenjie
yin, wenjie@eijnewniy·
📑new paper on the generalisability of abusive language detection models and the role of slurs and profanity in training! 💭non-abusive swearing and implicit abuse are often confusing for models. _how_ do such biases come into play? 💡open access on: sciencedirect.com/science/articl…
English
1
8
20
0
Computational Linguistics @ QMUL retweetledi
Matthew Purver
Matthew Purver@mpurver·
Nice blog post @BBCRD by @QMCompLing 's very own Iacopo Ghinassi - read about his work in topic segmenation #NLProc
QMUL Electronic Engineering and Computer Science@QMEECS

EECS PhD student, Iacopo Ghinassi, has been working with @BBCRD to explore ways to segment and annotate media content without direct human effort. The project is part of the DAME doctoral training programme sponsored by the BBC. Read here: bbc.co.uk/rd/blog/2022-0… @QMULSciEng

English
0
2
8
0
Computational Linguistics @ QMUL
Computational Linguistics @ QMUL@QMCompLing·
(5/5) We then propose two approaches to improve robustness: MAS augments input data with synonyms to increase lexical variation during training, and ADV uses adversarial training. Both are pretty effective at improving the performance. See the paper for more detail:
Computational Linguistics @ QMUL tweet media
English
0
0
0
0
Computational Linguistics @ QMUL
Computational Linguistics @ QMUL@QMCompLing·
(4/5) Now we can test how much substituting in synonymous material affects state-of-the-art models. Accuracy drops dramatically!
Computational Linguistics @ QMUL tweet media
English
1
0
0
0
Computational Linguistics @ QMUL
Computational Linguistics @ QMUL@QMCompLing·
(1/5) New ACL paper out now: Towards Robustness of Text-to-SQL Models against Synonym Substitution, Gan, Chen, Chang, Purver, Woodward, Xie, Huang: improving Text-to-SQL robustness via a new dataset, input data augmentation and adversarial training arxiv.org/abs/2106.01065 #NLProc
English
2
1
2
0
Computational Linguistics @ QMUL retweetledi
Xia Zeng
Xia Zeng@XiaZeng2·
Excited to shared our binary step-by-step approach that achieves substantial improvements over the baseline!🍻 No extra data or huge models required!😆
Arkaitz Zubiaga@arkaitz

🔵 We (w/ @XiaZeng2) participated in the SCIVER scientific claim verification task at NAACL / SDP. Here's the paper describing our approach! 👇 arxiv.org/abs/2104.11572

English
1
6
11
0
Computational Linguistics @ QMUL retweetledi
Pat Healey
Pat Healey@Pat_Healey·
Congratulations to Morteza Rohanian and Julian Hough, on their COLING 2020 Outstanding Paper: "Re-framing Incremental Deep Language Models for Dialogue Processing with Multi-task Learning"! @QMCogSci coling2020.org/2020/11/29/out…
English
0
4
4
0

Keşfet

@BBCRD @embeddiaproject @QMEECS @guardiantech @EPSRC @mpurver @QMCogSci @QMUL