Cover
Unit 02 • 6 Lecture Hours

Research Topic, Literature Survey & Problem Statement

Formulating research questions, discovering the research gap, and structuring testable hypotheses.

TU Weightage: 10 - 15 MarksCore: Research Gap • SMART Objectives • H₀ vs H₁
Core Memory Anchors: Topic, Literature & Hypotheses
10-Mark Question (Literature Review & Hypotheses)

💡 Elaboration Strategy: परीक्षा हलमा यी मुख्य Anchor Points स्मरण राख्नुभयो भने प्रत्येक बुँदालाई प्राज्ञिक रूपमा विस्तार गरेर सजिलै २ देखि ३ पृष्ठको पूर्ण १०-मार्क्सको उत्तर तयार गर्न सकिन्छ।

1

Topic Selection: हावामा होइन, Data र Feasibility हेरेर रोज्ने

धेरै student ले 'AI in Healthcare' जस्तो विशाल topic छानेर पछि फस्छन्। राम्रो research topic मा ३ वटा कुरा मिल्नुपर्छ: novelty (केही नयाँपन हुनुपर्‍यो), feasibility (समय र बजेटमा सक्किने हुनुपर्‍यो), र data availability (training data पाइनुपर्‍यो)।

Exam मा broad topic भर्सेस focused topic को उदाहरण दिनुस्: 'AI in Healthcare' (Poor) vs 'Pneumonia Detection on Chest X-Rays using CNN' (Excellent)।
2

Backward vs Forward Search: जरादेखि हाँगासम्म खोज्ने

Backward search भनेको हातमा भएको paper को References हेरेर पुराना foundational papers खोज्नु हो। Forward search भनेको Google Scholar मा 'Cited by' थिचेर त्यही paper लाई reference मानेर अरू कस-कसले नयाँ research गरे भनेर पछ्याउनु हो।

नेपालको स्थानीय context को लागि Nepal Journals Online (NepJOL) प्रयोग गरिन्छ भनेर एक लाइन थप्नुस्।
3

Research Gap: अघिल्ला लेखकले छाडेको खाली ठाउँ

अरूले गरिसकेको काम दोहोर्‍याएर research हुँदैन। अरू research paper को अन्तिम खण्ड 'Future Work' वा 'Limitations' पढेपछि बल्ल पत्ता लाग्छ कि उनीहरूले के गर्न सकेनन्—त्यही खाली ठाउँ (gap) नै हाम्रो research को मुख्य जग हो।

Exam मा ३ वटा स्तम्भ भएको Literature Matrix बनाउनुस्: [Author/Year | Method Used | Unresolved Limitation (Gap)]।
4

SMART Objectives र Hypotheses जोडी (H₀ vs H₁)

हाम्रो objective हावादारी हुनुहुँदैन, SMART (Specific, Measurable, Achievable, Relevant, Time-bound) हुनुपर्छ। अनि Hypothesis मा दुईवटा हुन्छन्: Null Hypothesis (H₀: केही फरक पर्दैन, status quo) र Alternative Hypothesis (H₁: साच्चिकै फरक पर्छ, significant effect)।

One-tailed (एक दिशा देखाउने, जस्तै μ₁ > μ₂) र Two-tailed (फरक मात्र देखाउने, μ₁ ≠ μ₂) को गणितीय notation लेख्नुस्।
Section 01

Selecting a Research Topic in Computer Applications

Selecting an academic research topic in Computer Applications follows a funnel approach: progressing from a broad Theme to a specific Topic, and narrowing down to an actionable Research Issue:

THE TOPIC SELECTION FUNNEL:
[Broad Research Theme] ──> Natural Language Processing (NLP)
↓ narrows to
[Focused Research Topic] ──> Optical Character Recognition (OCR) for Devanagari
↓ narrows to
[Actionable Research Issue] ──> Attention-based CNN for Low-Resolution Handwritten Nepali Numerals
👨‍🏫शिक्षकको बोर्ड नोट (Teacher's Board Note)
सुरुमै धेरै ठूलो विषय (जस्तै: "Artificial Intelligence in Education") छान्नु विद्यार्थीको सबैभन्दा ठूलो गल्ती हो। विषयलाई साँघुरो (Narrow) बनाएर स्पष्ट समस्यामा केन्द्रित गर्नुपर्छ, जसको डाटा ६ महिनाभित्र संकलन गर्न सकियोस्।
Section 02

Literature Survey & Citation Search Strategies

A thorough literature survey prevents “reinventing the computational wheel” and establishes the theoretical scaffolding of your study.

1. Backward Reference Searching

Examining the Bibliography / References section of an authoritative seminal paper to trace foundational algorithms and classical theories backward in time.

Vector: Present Paper ──> Past Foundational Literature

2. Forward Reference Searching

Using academic indexing engines (Google Scholar, Scopus, IEEE Xplore) to inspect the “Cited by” links of a paper to trace modern modifications forward in time.

Vector: Seminal Paper ──> Future Descendant Citations
🎯 TU Exam Tip:Lazar et al. को HCI पुस्तक अनुसार Literature Search गर्दा दुवै Backward र Forward विधि प्रयोग गर्दा मात्र कुनै पनि महत्त्वपूर्ण अनुसन्धान छुट्दैन।
Section 03

Identifying the Research Gap

A Research Gap represents the intellectual space where prior literature falls short. Without a clearly identified gap, a research proposal lacks scientific justification.

Type of GapNature of Limitation in Prior WorksComputer Applications Example
Methodological GapPast studies utilized legacy or suboptimal algorithms.Existing fraud detection uses Decision Trees; deep Transformer models have not been evaluated.
Contextual / Regional GapAlgorithms validated in Western contexts but untested regionally.Speech recognition tools benchmarked for US English, lacking evaluations on Nepali dialects.
Theoretical GapConflicting empirical evidence among published papers.Paper A claims Dark Mode increases coding speed; Paper B claims no statistical effect.
👨‍🏫शिक्षकको बोर्ड नोट (Teacher's Board Note)
रिसर्च ग्याप भेट्टाउने सबैभन्दा सजिलो सूत्र: नयाँ-नयाँ जर्नल पेपरहरूको अन्त्यमा रहेको “Limitations” र “Future Scope” पढ्नुहोस्। त्यहाँ लेखकहरूले नै "हामीले यो गर्न सकेनौं, भविष्यमा अरूले गरोस्" भनेर ग्याप औंल्याइदिएका हुन्छन्।
Section 04

Formulating the Problem Statement

A Problem Statement is a concise description of an issue that needs to be addressed. A complete TU problem statement contains four indispensable structural components:

1. The Ideal Context:What the system or computational process should achieve under optimal conditions.
2. The Reality / Defect:The actual operational bottleneck, error rate, or inefficiency occurring today.
3. The Consequences:What happens if this computing problem is left unresolved (financial loss, latency, bias).
4. The Proposed Remedy:How the proposed study addresses and resolves this explicit problem.
Section 05

SMART Research Objectives

Research objectives define what the investigator aims to accomplish. In scientific computing, objectives must adhere strictly to the SMART taxonomy:

SSpecificClear & unambiguous
MMeasurableMetric-quantifiable
AAchievableFeasible resources
RRelevantTied to Problem
TTime-boundTarget milestones
Section 06

Hypothesis Formulation: Null (H₀) vs Alternative (H₁)

A hypothesis is a tentative proposition formulated for empirical statistical testing. It forms the backbone of inferential statistics in Unit 4:

Null Hypothesis (H₀)

H₀: μ₁ = μ₂

States that there is no significant difference, no relationship, or no treatment effect between the observed variables. Represents the conservative status quo.

Example: “The new algorithm does not reduce execution latency compared to standard baseline.”

Alternative Hypothesis (H₁)

H₁: μ₁ ≠ μ₂ (or > / <)

States that a statistically significant difference, association, or effect truly exists. What the researcher aims to support with empirical data.

Example: “The new algorithm significantly reduces execution latency.”
👨‍🏫शिक्षकको बोर्ड नोट (Teacher's Board Note)
कोर्टको भाषा सम्झिनुहोस्: Null Hypothesis (H₀) भनेको "प्रमाण नभेटिएसम्म कसैलाई निर्दोष मान्नु" जस्तै हो। यदि परीक्षण गर्दा पर्याप्त प्रमाण (p < 0.05) भेटियो भने बल्ल हामी H₀ लाई Reject गरेर नयाँ अल्गोरिदमले काम गर्‍यो (H₁) भनी प्रमाणित गर्छौं।
High-Probability Exam Blueprint

TU Exam Model Q&A Bank — Unit 2

5 Past & Model Questions

💡 यी प्रश्नहरू विगतका TU परीक्षा तथा सिलेबसका मुख्य उद्देष्यहरूबाट संकलन गरिएका हुन्। उत्तरपुस्तिकामा Full Marks पाउन चाहिने मोडेल ढाँचा यहाँ प्रस्तुत गरिएको छ।

शिक्षकको छोटो सारांश (One-Minute Exam Recall):
टपिक छान्दा ४ कुरा याद राख्ने: १) Feasible (सकिने खालको हुनुपर्‍यो), २) Relevant (कम्प्युटर क्षेत्रको हुनुपर्‍यो), ३) Novel (केही नयाँ कुरा हुनुपर्‍यो), र ४) Data Available (डाटा पाइनुपर्‍यो)।

Standard Model Answer Structure:

1. Feasibility: The scope must match the available computational infrastructure, dataset access, financial budget, and graduation timeline.

2. Domain Relevance: Must address current challenges in computer applications, software engineering, human-computer interaction, or intelligent systems.

3. Originality & Novelty: Must avoid mere repetition of established solutions and introduce novel algorithmic modifications, contextual adaptations, or empirical evaluations.

4. Data Accessibility & Ethical Viability: Training datasets, human participant access, and system logs must be ethically obtainable without privacy infringement.

✍️ Essential Answer Keywords / Must-Draw Elements:
✓ Feasibility & Scope✓ Domain Relevance✓ Novelty & Non-duplication✓ Data Accessibility
Previous: Unit 1Next: Unit 3 (Research Design & Instrumentation)