Retrieval Challenges in Low-Resource Public Service Information: A Case Study on Food Pantry Access

Touseef Hasan; Laila Cure; Souvika Sarkar

cs.IRFeb 25, 2026

Retrieval Challenges in Low-Resource Public Service Information: A Case Study on Food Pantry Access

Authors: Touseef Hasan, Laila Cure, Souvika Sarkar

Organizations: Wichita State University, Wichita, KS, USA

Abstract

Public service information systems are often fragmented, inconsistently formatted, and outdated. These characteristics create low-resource retrieval environments that hinder timely access to critical services. We investigate retrieval challenges in such settings through the domain of food pantry access, a socially urgent problem given persistent food insecurity. We develop an AI-powered conversational retrieval system that scrapes and indexes publicly available pantry data and employs a Retrieval-Augmented Generation (RAG) pipeline to support natural language queries via a web interface. We conduct a pilot evaluation study using community-sourced queries to examine system behavior in realistic scenarios. Our analysis reveals key limitations in retrieval robustness, handling underspecified queries, and grounding over inconsistent knowledge bases. This ongoing work exposes fundamental IR challenges in low-resource environments and motivates future research on robust conversational retrieval to improve access to critical public resources.

Figures & tables

Type of User Query	Example	# Queries	Accuracy		Failure Notes
			Llama	GPT
Only location	Where can I find food pantries in Sedgwick County?	20	0.60	0.65	Retrieving counties
Location + hours	Are there food pantries in Derby open on Wednesdays?	13	0.38	0.38	Retrieving counties
Location + ID requirement	Find food pantries in Maize that do not require any ID.	13	0.15	0.30	ID constraint errors
Ambiguous location	Where can I get free food near me?	1	0.00	0.00	Needs clarification
Recall-based exact search	Where is the pantry I went to in 67214 on 21st Street?	4	1.00	1.00	No errors

Table 1. Queries categorized by user requirements, along with a systematic breakdown of LLM performance by query type. An example is provided for each type of query. Accuracy per query type has been calculated for the Llama and GPT models respectively. Notes indicate the most frequent cause of the erroneous responses for each type of user query.

Figure 1. Our proposed Pantry Assistant. Food pantries are scraped from Kansas Food Source website and stored into a structured JSON database. Data is then embedded and indexed for similarity search to enable semantic retrieval. The retrieved relevant pantries are then passed to an LLM, which recommends pantries to users based on their requirements.

Information Retrieval

Information Retrieval studies how to find, rank, and organize relevant information in large collections of data.

See all 471 papers

TRACE: Trustworthy Retrieval-Augmented Conversational Engine
Aug 10, 2026Touseef Hasan, Laila Cure, Souvika SarkarRetrieval-Augmented Generation Graph Retrieval-Augmented Generation
Wichita State University, Wichita, KS, USA
Evaluation of LLMs in retrieving food and nutritional context for RAG systems
Mar 10, 2026Maks Požarnik Vavken, Matevž Ogrinc, Tome Eftimov +1RAG Evaluation
Department of Computer Systems, Institute "Jožef Stefan", Ljubljana, Slovenija
PILLAR: Private Inverted-Index Lexical Lookup for Augmented Retrieval
Sep 28, 2026Truong Son Nguyen, Daniel Blackley, Ni Trieu +1Retrieval-Augmented Generation Privacy-Preserving RAG
Arizona State University · Department of Computer Science Arizona State University Tempe, AZ 85281, USA · George Mason University +1

Aug 10, 2026J/K move · Enter open · S save

TRACE: Trustworthy Retrieval-Augmented Conversational Engine

Touseef Hasan, Laila Cure, Souvika Sarkar

Wichita State University, Wichita, KS, USA

Retrieval-Augmented Generation Graph Retrieval-Augmented Generation Recommender Systems

Open paper AskPDF arXiv

Public service chatbots are expected to deliver recommendations from an underlying public service directory, while also making sure that the recommendations respect explicit user constraints. In practice, public service directories are noisy and inconsistent, and general-purpose large language model (LLM) or AI-based chatbots frequently generate unreliable recommendations, citing unverified sources from the web. We investigate the impact of retrieval quality on constraint-aware recommendation in public service conversational systems built over noisy and heterogeneous service directories. We propose TRACE (Trustworthy Retrieval-Augmented Conversational Engine), a retrieval-based, constraint-aware framework that parses input user queries into structural and semantic constraints for downstream retrieval, with the help of a dual data representation schema. Using a curated statewide pantry directory and a synthetic query benchmark, we evaluate multiple knowledge-representation variants with and without knowledge graphs (KGs). We experiment with several open-source LLMs and a proprietary model, showing that strengthening retrieval substantially improves user constraint satisfaction while reducing hallucinated recommendations. Performance differences across LLMs narrowed in our experiments as retrieval quality improved, making results less sensitive to model size. These findings suggest that the quality of retrieval is key for robust public service conversational systems.

Abstract

Information Retrieval

Explore similar work

TRACE: Trustworthy Retrieval-Augmented Conversational Engine

Evaluation of LLMs in retrieving food and nutritional context for RAG systems

PILLAR: Private Inverted-Index Lexical Lookup for Augmented Retrieval