Toqeer Ehsan

—

Unclaimed · AI-structured profilename-matched identity

NLP researcher spanning Urdu syntactic parsing with multi-task learning and LLM-based policy monitoring automation

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-01) 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 scorable: this snapshot covers only the two arXiv preprints in Evolve's catalog, with no aggregate publication counts. This axis would measure total research output volume and how productivity trends over time.

Impact—

No citation counts, h-index or venue-level influence data are available in this snapshot, so impact cannot be scored. This axis would measure how widely the work is cited and its uptake across the field.

Focus0/100

Scores at the low end (0) because recorded output splits evenly, 50/50, between computational linguistics (NLP) and natural language processing / policy informatics. The two strands — Urdu syntactic parsing with multi-task learning and LLM-based policy survey automation — share sequence-modeling and information-extraction methods, so the split reads as a deliberate two-track program rather than unfocused drift, though the evidence base is only two papers.

Collaboration46/100

Mid-range score (46) with 9 distinct co-authors recorded. Collaboration splits into two visible clusters: Miriam Butt, Sarmad Hussain, Hassan Alhuzali and Ali Al-Laith on Urdu parsing; Carolyn Cole, Matthias Deschryvere and Arash Hajikhani on policy informatics. First authorship on the parsing paper versus middle authorship on the policy paper suggests a leading role in one cluster and a contributing role in the other.

Activity—

Insufficient data to score: with only two 2026 preprints on record, there is no multi-year publication cadence. This axis would measure publishing tempo and recency of research activity over successive years.

Reproducibility20/100

Low score (20), with a mean of 20 across the 2 papers Evolve assessed for reproducibility readiness; both are rated low. Neither preprint currently ships the artifacts a third party would need, and Evolve's verdicts flag evaluation weaknesses — marginal gains over the single-task baseline, and circular self-evaluation with lenient metrics.

Research summary

Toqeer Ehsan is a natural language processing researcher whose cataloged work to date spans two strands: linguistic resources for Urdu and applied LLM evaluation for policy monitoring. As first author, Ehsan co-developed a multi-task BLSTM approach using contextualized word representations for syntactic parsing of a morphologically rich language, pairing phrase-structure-to-dependency-structure conversion with joint constituency and dependency parsing; Evolve's review credits an F1 of 91.39 and LAS of 85.69 on Urdu, calling it solid resource work while noting that gains over the authors' own single-task baseline are marginal. As a middle author, Ehsan contributed to an evaluation of GPT-4o for pre-filling OECD STIP policy surveys from policy documents, reporting 84–95% label overlap; Evolve cautions that circular self-evaluation and lenient metrics weaken the headline claim. Bibliometric signals mirror this two-track profile: field distribution splits evenly (50/50) between computational linguistics and NLP/policy informatics, yielding a focus score of 0; 9 distinct co-authors give a mid-range collaboration score of 46; and a mean reproducibility readiness of 20 across both papers reflects limited artifact availability. Co-authors include Miriam Butt, Sarmad Hussain, Hassan Alhuzali and Ali Al-Laith on parsing, and Carolyn Cole, Matthias Deschryvere and Arash Hajikhani on policy monitoring. Both papers are 2026 arXiv preprints; no institution, country or ORCID is on file.

Collaboration pattern

Ehsan's 9 distinct co-authors fall into two clusters. The Urdu parsing paper — where Ehsan is first author — connects to Miriam Butt, Sarmad Hussain, Hassan Alhuzali and Ali Al-Laith. The policy-monitoring paper — where Ehsan is a middle author — connects to Carolyn Cole, Matthias Deschryvere and Arash Hajikhani. The mid-range collaboration score (46) matches this pattern: concentrated, two-community co-authorship rather than a wide open network.

Research keywords

Urdu language processingSyntactic parsingConstituency parsingDependency parsingMulti-task learningContextualized word representationsSequence labelingLarge language modelsPolicy monitoringSurvey automationInformation extractionLong-context in-context learning

Highlights

Urdu parsing reaches F1 91.39 / LAS 85.69 with multi-task BLSTM
As first author, combined a phrase-structure-to-dependency-structure conversion with multi-task learning over contextualized word representations for constituency and dependency parsing of Urdu, a morphologically rich language. Evolve calls it solid resource work while noting the improvement over the authors' own single-task baseline is marginal; reproducibility readiness is rated low.
GPT-4o pre-fills OECD STIP policy surveys with 84–95% label overlap
As a middle author, evaluated large language models for policy monitoring by having GPT-4o turn policy documents into structured survey responses for STIP Compass. Evolve notes the 84–95% label overlap but cautions that circular self-evaluation and lenient metrics weaken the headline claim; reproducibility readiness is low.
A two-track program bridging language resources and policy informatics
Cataloged output splits evenly (50/50) between computational linguistics (NLP) and NLP/policy informatics, unified by sequence-modeling and information-extraction methods; the 9 distinct co-authors split into an Urdu-parsing cluster and a policy-monitoring cluster.

Timeline

2026First-author arXiv preprint on multi-task learning with contextualized word representations for syntactic parsing of Urdu, combining PS-to-DS conversion with joint constituency and dependency parsing (F1 91.39, LAS 85.69).
2026Co-authored arXiv preprint evaluating large language models for policy monitoring: GPT-4o pre-fills OECD STIP Compass survey responses with 84–95% label overlap.
2026Evolve reviews both preprints, rating each low on reproducibility readiness and flagging evaluation caveats in both verdicts.

Papers in the Evolve catalog

Frequent co-authors

Miriam Butt×1 Sarmad Hussain×1 Hassan Alhuzali×1 Ali Al-Laith×1 Carolyn Cole×1 Matthias Deschryvere×1 Arash Hajikhani×1 Toqeer Ehsan×1 Arash Hajikhani×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 scores derive from a minimal OpenAlex-based snapshot covering only two reviewed arXiv preprints; the output, impact and activity axes lack sufficient data to score.
No ORCID, institution or country is on file; this profile is name-resolved from public bibliographic data and identity should be verified before citing.
All quoted numbers (F1 91.39, LAS 85.69, 84–95% label overlap) come from Evolve's review verdicts on arXiv preprints, not from independently verified publications.
The provided frequent-coauthor list contains self and duplicate entries (Toqeer Ehsan himself; Arash Hajikhani twice), treated after deduplication here.