Searching in Dialect: A Pilot Study of Query Failures and Lexical Fixes in a Palestinian Arabic E-commerce Store

Shoppers in small Arabic-speaking markets search the way they speak: in dialect, with non-standard spelling, truncated words, and English or other languages mixed in. We study six weeks of real search logs (9 Oct – 21 Nov 2026) from Glamora, a Palestinian e-commerce store with about 2,400 products on the web and a mobile app. Of 239 logged searches, 55.6% returned no results. We show first that 40% of the logged events were partial keystrokes produced by search-as-you-type logging. After filtering them out, the zero-result rate at logging time was still 52.4%. We then replay 74 cleaned and labeled product queries through a ladder of lexical methods: raw substring matching; Arabic orthographic normalization; clitic stripping, light stemming and a store-curated synonym table; partial-match fallback; and one-guess, dialect-aware spelling correction. We compare these with the production search engine. On queries whose intent exists in the catalog, Success@10 rises from 43.9% to 89.5% (95% CI 80.7–96.5), and the zero-result rate falls from 52.6% to 8.8%. Partial matching contributes most (+21 points, McNemar p < 0.001), more than spelling correction (+7 points, p = 0.13). The same components turn honest "no results" into misleading ones. For queries asking for products the store does not sell, the share returning only irrelevant items rises from 5.9% to 47.1% (p = 0.016). Such queries account for 22% of evaluated search events, led by dresses typed in English and Spanish, so zero-result logs are also a direct signal of unmet demand. We release the queries, labels and evaluation code. Pilot study, preprint, not peer reviewed.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24409546
Primary Topic
Information Retrieval and Search Behavior
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Searching in Dialect: A Pilot Study of Query Failures and Lexical Fixes in a Palestinian Arabic E-commerce Store

Yaser Sweity
Zenodo (CERN European Organization for Nuclear Research)
Information Retrieval and Search Behavior
preprint

Searching in Dialect: A Pilot Study of Query Failures and Lexical Fixes in a Palestinian Arabic E-commerce Store

Yaser Sweity
preprint en

Abstract

Shoppers in small Arabic-speaking markets search the way they speak: in dialect, with non-standard spelling, truncated words, and English or other languages mixed in. We study six weeks of real search logs (9 Oct – 21 Nov 2026) from Glamora, a Palestinian e-commerce store with about 2,400 products on the web and a mobile app. Of 239 logged searches, 55.6% returned no results. We show first that 40% of the logged events were partial keystrokes produced by search-as-you-type logging. After filtering them out, the zero-result rate at logging time was still 52.4%. We then replay 74 cleaned and labeled product queries through a ladder of lexical methods: raw substring matching; Arabic orthographic normalization; clitic stripping, light stemming and a store-curated synonym table; partial-match fallback; and one-guess, dialect-aware spelling correction. We compare these with the production search engine. On queries whose intent exists in the catalog, Success@10 rises from 43.9% to 89.5% (95% CI 80.7–96.5), and the zero-result rate falls from 52.6% to 8.8%. Partial matching contributes most (+21 points, McNemar p < 0.001), more than spelling correction (+7 points, p = 0.13). The same components turn honest "no results" into misleading ones. For queries asking for products the store does not sell, the share returning only irrelevant items rises from 5.9% to 47.1% (p = 0.016). Such queries account for 22% of evaluated search events, led by dresses typed in English and Spanish, so zero-result logs are also a direct signal of unmet demand. We release the queries, labels and evaluation code. Pilot study, preprint, not peer reviewed.

Zenodo (CERN European Organization for Nuclear Research)
Information Retrieval and Search Behavior
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
Searching in Dialect: A Pilot Study of Query Failures and Lexical Fixes in a Palestinian Arabic E-commerce Store — Yaser Sweity · Zenodo (CERN European Organization for Nuclear Research) (2026) | TGRS Research Map | TGRS