[Paper Review] ELICA: An Automated Tool for Dynamic Extraction of Requirements Relevant Information

Zahra Shakeri Hossein Abad, Vincenzo Gervasi+2 more|arXiv (Cornell University)|Jul 21, 2018

Software Engineering Research21 references4 citations

TL;DR

ELICA is an automated, real-time tool that uses Weighted Finite State Transducers (WFSTs) and statistical language models to dynamically extract and label requirements-relevant information from domain documents and stakeholder interactions. It enhances analyst understanding by identifying key terms, analyzing speaker intent (confidence, tone, emotion), and delivering visualized, exportable snippets—demonstrated effective in real-time with a 95% significance level (p=0.02) in validating contextual lexical association as a relevance indicator.

ABSTRACT

Requirements elicitation requires extensive knowledge and deep understanding of the problem domain where the final system will be situated. However, in many software development projects, analysts are required to elicit the requirements from an unfamiliar domain, which often causes communication barriers between analysts and stakeholders. In this paper, we propose a requirements ELICitation Aid tool (ELICA) to help analysts better understand the target application domain by dynamic extraction and labeling of requirements-relevant knowledge. To extract the relevant terms, we leverage the flexibility and power of Weighted Finite State Transducers (WFSTs) in dynamic modeling of natural language processing tasks. In addition to the information conveyed through text, ELICA captures and processes non-linguistic information about the intention of speakers such as their confidence level, analytical tone, and emotions. The extracted information is made available to the analysts as a set of labeled snippets with highlighted relevant terms which can also be exported as an artifact of the Requirements Engineering (RE) process. The application and usefulness of ELICA are demonstrated through a case study. This study shows how pre-existing relevant information about the application domain and the information captured during an elicitation meeting, such as the conversation and stakeholders' intentions, can be captured and used to support analysts achieving their tasks.

Motivation & Objective

• Address the challenge of incomplete or inaccurate requirements elicitation due to analysts' lack of domain knowledge and memory limitations during stakeholder interviews.
• Reduce the misinformation effect in interviews by providing real-time, context-aware support during requirements gathering.
• Enhance analyst effectiveness by dynamically surfacing relevant domain knowledge and capturing non-linguistic cues such as confidence, tone, and emotion from stakeholder interactions.
• Provide a tool-supported, interactive environment that integrates textual and non-linguistic data into a structured, exportable artifact for traceability and collaboration.
• Demonstrate the feasibility and effectiveness of real-time, automated extraction of relevant information in a realistic software requirements elicitation scenario.

Proposed method

• Leverages Weighted Finite State Transducers (WFSTs) to model variable-length textual snippets and enable flexible, dynamic natural language processing for requirements extraction.
• Applies statistical language models (LMs) to identify and rank relevant terms from domain documents and interaction transcripts based on lexical association with the current conversation.
• Integrates multi-modal input including spoken dialogue (via third-party transcription) and written text (e.g., chat, emails) to support real-time processing during elicitation sessions.
• Analyzes non-linguistic cues—such as confidence level, analytical tone, and emotional state—using automated intention recognition models to enrich interpretation of stakeholder input.
• Visualizes extracted information by highlighting key terms and displaying intention metadata, enabling analysts to quickly assess relevance and tone.
• Exports all extracted information, including labeled snippets and intention data, in multiple formats (PDF, JSON, CSV) for use as formal artifacts in the Requirements Engineering process.

Experimental results

Research questions

• RQ1 To what extent does contextual lexical association between extracted snippets and relevant terms improve the accuracy of requirement information extraction?
• RQ2 Can real-time dynamic extraction of domain-relevant information enhance analyst understanding during requirements elicitation meetings with unfamiliar stakeholders?
• RQ3 How effectively can non-linguistic cues such as speaker confidence, tone, and emotion be captured and utilized to support interpretation of stakeholder input?
• RQ4 Does the integration of domain repository knowledge with live interaction data improve the completeness and relevance of elicited requirements?
• RQ5 Can the tool maintain real-time responsiveness during lengthy, complex elicitation sessions with large domain documents?

Key findings

• The null hypothesis that contextual lexical association has no impact on snippet relevance was rejected at a p-value of 0.02, indicating that lexical overlap is a statistically significant indicator of relevance.
• ELICA successfully processes and extracts relevant information in real time during a simulated real-world interview, keeping pace with the conversation flow.
• The tool effectively identifies and highlights key terms in extracted snippets, supported by visual cues and labeled intention data (e.g., confidence, tone).
• The system enables export of all extracted information in multiple formats (PDF, JSON, CSV), supporting traceability and collaboration as formal RE artifacts.
• The case study confirms that ELICA can integrate domain documents (e.g., RE Essentials book) with live interaction data to provide contextually relevant, analyst-ready information.
• Feedback collection mechanisms are in place to iteratively improve both the extraction and intention recognition components based on analyst input.