Danushka Bollegala

Amazon (United States) · United States

Unclaimed · AI-structured profileOpenAlex

Amazon-based NLP researcher focused on lexical semantics, multilingual LLMs, and cross-lingual representation alignment

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not yet scored: the underlying output data — complete publication counts and venue-normalized volume — is insufficient in this snapshot. This axis would measure total research production and how it scales across the researcher's career.

Impact—

Not yet scored: citation-based evidence in this snapshot is incomplete, so no impact estimate is offered. This axis would measure uptake of the work through citation counts, h-index trajectory, and downstream influence on follow-up research.

Focus0/100

Focus scores 0, the lowest concentration reading in this profile. The raw record supports that breadth: the field distribution splits exactly 0.5/0.5 between natural language processing and NLP (multilingual LLMs), and the topic list spans five areas — topic modeling (16), NLP techniques (10), sentiment analysis and opinion mining (6), multimodal machine learning applications (6), and ethics and social impacts of AI (4) — with no single theme dominating beyond topic modeling.

Collaboration41/100

Collaboration scores 41, a moderate reading. The raw record lists 7 distinct coauthors in the recent window, with recurring partners including Yi Zhou, Kiamehr Rezaee, Mohammad Taher Pilehvar, Jose Camacho-Collados, Taisei Yamamoto, Ryoma Kumon, and Hitomi Yanaka. Both Evolve-reviewed 2026 papers place this researcher in a middle author position without corresponding-author status, consistent with contributing within small academic teams rather than leading large consortia.

Activity90/100

Activity scores 90, the strongest axis in this profile. The raw counts by year show 5 works in 2024, 1 in 2025, and 15 in 2026, for 21 works over the last three years. The 2026 figure indicates a sharp recent acceleration in output, with the bulk of indexed work appearing in the current year.

Reproducibility78/100

Reproducibility scores 78, averaged over 2 Evolve-reviewed papers. The cross-lingual alignment paper (CAROT) rates high for reproducibility readiness, with solid, well-analyzed steering gains and honest reporting of modest training gains; the WiC paper rates medium, supported by ablations and human error analysis but lacking significance tests. Transparent analysis combined with occasional missing statistical testing defines this axis.

Research summary

Danushka Bollegala is a natural language processing researcher affiliated with Amazon (United States). The OpenAlex-indexed profile splits evenly (0.5/0.5) between natural language processing and NLP for multilingual LLMs, and the record shows 21 works over the last three years — 5 in 2024, 1 in 2025, and 15 in 2026 — an accelerating, highly current output pattern reflected in an activity score of 90. The topic profile is broad: topic modeling leads (16), followed by NLP techniques (10), sentiment analysis and opinion mining (6), multimodal machine learning applications (6), and ethics and social impacts of AI (4). Evolve has reviewed two 2026 arXiv papers: 'WiC is Not WSD', which argues that Word-in-Context difficulty stems from unspecified sense granularity rather than ambiguity resolution itself, supported by ablations and human error analysis; and CAROT, which aligns LLM token-level representations in a language-agnostic space by combining LEACE concept erasure with optimal transport, rated high for reproducibility readiness. Average reproducibility across the two reviewed papers is 78. Citation aggregates in this snapshot (h-index 1, i10-index 0) conflict with the publication record and are treated as incomplete.

Collaboration pattern

Collaboration is compact: the record lists 7 distinct coauthors in the recent three-year window, with Yi Zhou, Kiamehr Rezaee, Mohammad Taher Pilehvar, Jose Camacho-Collados, Taisei Yamamoto, Ryoma Kumon, and Hitomi Yanaka recurring as frequent partners. Both 2026 Evolve-reviewed papers place this researcher in a middle author position without corresponding-author status, suggesting contributions within small academic teams rather than sole or corresponding leadership of large groups.

Research keywords

Word-in-Context (WiC)Word Sense DisambiguationLexical SemanticsSense GranularityCross-Lingual AlignmentOptimal TransportConcept Erasure (LEACE)Representation EngineeringLarge Language ModelsSentiment Analysis

Highlights

Reframing Word-in-Context difficulty as a sense-granularity problem
In 'WiC is Not WSD' (2026, arXiv), the study argues that WiC task difficulty stems from unspecified sense granularity rather than lexical ambiguity resolution itself. Evolve finds the claim supported by ablations and human error analysis, while noting the absence of significance tests. Reproducibility readiness: medium.
Token-level optimal transport for cross-lingual LLM alignment
In a 2026 arXiv paper, CAROT aligns LLM token-level representations in a language-agnostic space by combining LEACE concept erasure with optimal transport. Evolve finds the steering gains solid and well-analyzed, with training gains modest but honestly reported. Reproducibility readiness: high.

Timeline

20245 works appear in the OpenAlex counts-by-year record, the earliest year shown in the current three-year window.
2025Only 1 work is indexed for 2025, the low point of the three-year window.
2026Output accelerates sharply: 15 indexed works in a single year, and two arXiv papers enter the Evolve catalog — 'WiC is Not WSD' on lexical ambiguity in LLMs, and CAROT on cross-lingual token-level alignment.

Papers in the Evolve catalog

Frequent co-authors

Yi Zhou×1 Kiamehr Rezaee×1 Mohammad Taher Pilehvar×1 Jose Camacho-Collados×1 Taisei Yamamoto×1 Ryoma Kumon×1 Hitomi Yanaka×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric figures come from an OpenAlex identity snapshot (identity: openalex); the indexed citation aggregates (h-index 1, i10-index 0) conflict with the listed output and likely reflect incomplete author disambiguation, so citation-based readings should be treated as unreliable.
The output and impact axes are unscored (dimMissing) because the underlying data is insufficient in this snapshot.
Only the two 2026 arXiv papers in the Evolve catalog were individually reviewed for this profile; the remaining indexed works are not assessed.
No ORCID is linked to this record, so author resolution relies on OpenAlex bibliographic name matching; identity confidence is name-based.