Publication2021
ConOffense: Multi-modal multitask Contrastive learning for offensive content identification
The record
- Type
- Publication
- Date
- 1 January 2021
- Level
- National
- Journal
- Proceedings 2021 IEEE International Conference on Big Data Big Data 2021
- Pages
- 4524-4529
- Published
- 1 January 2021
- DOI
- 10.1109/bigdata52589.2021.9671427
- Scopus ID
- 85125316860
- Authors
- 2
- Open access
- No
- Citations
- 7
- International co-authorship
- No
- Co-author countries
- IN
Classification
Every label below carries the reasoning it was assigned on, the confidence band it was assigned in, and what produced it. The rubric is published.
UN Sustainable Development Goals
16.1 Significantly reduce all forms of violence and related death rates everywhere
Multimodal deep learning system detects hateful and offensive social media content with 82.6% accuracy, addressing online hate speech and violence prevention.
- High confidence
- The subject matter is a strong and direct inference from the record type, the funding or organising body, or the programme context, even where the title alone is thin. Example: an ICMR-funded project against SDG 3, on the strength of the agency.
- Judged
- Judged against the published classification rubric, under the rubric and prompt versions recorded with the run that produced it.
- Produced by
- claude-haiku-4-5-20251001, under kortex-publication-v1
9.5 Enhance scientific research, upgrade technological capabilities, and increase researchers and R&D spending
Contrastive learning methodology combining text and visual encoders represents research innovation in multimodal machine learning.
- Moderate confidence
- The alignment is plausible and defensible but rests on an assumption a reasonable reviewer could decline. Published, and surfaced in the review queue for a human to confirm or override.
- Produced by
- claude-haiku-4-5-20251001, under kortex-publication-v1
Viksit Bharat @2047
vb-1.5 Caring society, social protection and inclusion
Technology enabling social protection and inclusion through prevention of online harassment.
- Produced by
- claude-haiku-4-5-20251001, under kortex-publication-v1
Each label above carries the instrument and the version that produced it. How each classification is reached.
Evidence
Kortex hosts no documents. Evidence is an outbound link to the authority that holds it.
No outbound evidence link is recorded for this activity. The record below states which source it came from and when it was last read.
Provenance
Which source system this record came from, how many source rows were resolved into it, and when each was last read.
- Source rows resolved into this record
- 1
Each record carries the date Kortex last read it from its source system. That is a freshness date and not a verification: Kortex records what its sources say and when it read them, and it does not check those statements against any external authority. No claim on this site is described as verified, because none is.
Citing this record
This URL is stable. A record’s permalink is frozen when it is published and is not regenerated if its title is later corrected.
KIIT Deemed to be University. “ConOffense: Multi-modal multitask Contrastive learning for offensive content identification”. Kortex.