Module D: Terminology, Definitions & Explainability Resources.¶

Example User Question for Pihu (Reminder):

"Compare the overall liking of SLHB5 and SLHB8 from the Signature Latte Cold V2 study. Did SLHB8 meet the benchmark, and what were the main aroma notes for it among millennials?"

How the (Revised) Module D helps answer our example question:

When Pihu synthesizes the final answer, it will use Module D to:

Understand User Language (General Vocabulary): Ensure it correctly interpreted terms like "compare," "meet the benchmark," "main aroma notes."
Explain Technical Terms (Statistical Term Definitions): If it mentions a "p-value" for the comparison, it can define it.
Use Standard Phrasing for Interpretations (Standard Interpretation Phrases): To say "SLHB8 Exceeds Expectations" in a consistent way.
Format the Output Correctly (Reporting Standards & Formatting Guidelines - NEW):
- State that liking scores are on a 9-point scale and what the anchors mean.
- List the main aroma notes with their percentages in the standard format (e.g., "Coffee-Roast (52%), Dairy (30%)").
- Use the standard sentence for statistical significance if the comparison between SLHB5 and SLHB8 was significant.
- Report mean scores to the standard number of decimal places.
Provide Transparency (Provenance & Explainability Templates): For the "Show Details" panel.
Maintain Persona (Pihu Persona & Tone): Deliver the answer in a helpful, analytical, and professional tone.
Define Acronyms (Glossary Terms): If it uses terms like "SLHB."

Module D: Explainability, Communication & Reporting Standards¶

Goal: To equip Pihu with the necessary vocabulary, definitions, standard phrasing, formatting rules, and persona guidelines to communicate insights clearly, consistently, accurately, and in a manner that aligns with TagTaste's standards and builds user trust.

1) Text-Based Structure with Actual Examples (Relevant to the User Question)¶

A. General Vocabulary & NLU Support Entry Form (Example: For the term "compare")

Term ID: VOCAB_COMPARE_001
Canonical Term: compare
Synonyms & User Phrases: vs, versus, difference between, how does X stack up against Y, X vs Y
Intent Association (Helps Query Understanding): COMPARISON_ANALYSIS
Notes: Indicates a direct comparison between two or more items on one or more attributes.
Created By: nlu_specialist_01
Created At: 2023-03-01T10:00:00Z

B. Statistical Term Definition Entry Form (Example: For "p-value" - relevant if the comparison SLHB5 vs SLHB8 was tested for significance)

Term ID: STAT_PVALUE_001
Term Name: p-value
Status: Active
Plain English Definition: The probability that an observed difference between groups (e.g., in liking scores) occurred purely by random chance. A small p-value (typically less than 0.05) suggests the difference is 'statistically significant,' meaning it's likely a real difference and not just due to chance.
TagTaste Example Pattern (use placeholders): The difference in {attributeName} between {productA} and {productB} was {significanceDescriptor} (p = {pValue}).
Visual Cue: Pihu flags p < 0.05 in green (significant), p >= 0.05 in grey (not significant).
Related Concepts: Statistical Significance, Hypothesis Testing, t-test, ANOVA
Notes for Pihu: When reporting a p-value, always state the alpha level used for significance (e.g., 0.05). Explain that 'statistically significant' does not automatically mean the difference is large or practically important, just that it's unlikely to be by chance.
Created By: stats_expert_01
Created At: 2023-04-10T10:00:00Z

C. Standard Interpretation Phrase Entry Form (Example: For reporting a benchmark result like "Exceeds Expectations")

Phrase ID: SIP_BENCH_EXCEEDS_001
Response Type / Context: BENCHMARK_COMPARISON_POSITIVE_EXCEEDS
Condition for Use (Conceptual): Product score is >= benchmark primary target AND falls within the 'Exceeds Expectations' band defined in Module B.
Template: For {attributeName}, {productName} scored {scoreValue} (out of {scaleMax}), which **Exceeds Expectations** based on the TagTaste benchmark target of {benchmarkTargetValue}. This indicates strong positive performance.
Variables to Fill: attributeName, productName, scoreValue, scaleMax, benchmarkTargetValue
Notes: Ensure scoreValue is formatted according to Reporting Standards.
Created By: report_writer_01
Created At: 2023-06-15T14:00:00Z

D. Reporting Standard & Formatting Guideline Entry Form (NEW)

Guideline: Reporting Mean Scores

Guideline ID: RSFG_MEAN_SCORE_001
Applies To: Mean scores for sensory attributes (e.g., Liking, Intensity)
Standard: Report mean scores to two decimal places (e.g., 7.05).
Example: Overall Liking for SLHB8 was 7.05.
Notes for Pihu: Always round to two decimal places for display.
Created By: style_guide_committee
Created At: 2023-07-01T11:00:00Z

Guideline: Reporting Scale Context

Guideline ID: RSFG_SCALE_CONTEXT_HEDONIC9_001
Applies To: Any score reported from a 9-point hedonic scale.
Standard: When first mentioning a score from a 9-point hedonic scale in a response, state: "(on a 9-point scale where 1=Dislike Extremely and 9=Like Extremely)" or similar concise explanation.
Example: SLHB8's overall liking score was 7.05 (on a 9-point scale where 1=Dislike Extremely and 9=Like Extremely).
Notes for Pihu: Provide this context at least once per major attribute type in an answer if space permits and it's the first mention.
Created By: style_guide_committee
Created At: 2023-07-01T11:05:00Z

Guideline: Reporting Top CATA Notes

Guideline ID: RSFG_CATA_TOPNOTES_001
Applies To: Reporting top mentioned items from a CATA question (e.g., aroma notes).
Standard: List the top 3-5 notes (or as appropriate) with their percentage of mentions, ordered by frequency. Format: "Note Name (XX%)".
Example: The main aroma notes for SLHB8 among millennials were: Coffee-Roast (52%), Dairy (30%), and Sugary (25%).
Notes for Pihu: Default to Top 3 unless the query specifies more or fewer. Ensure percentages are clearly linked to the specific filter group (e.g., "among millennials").
Created By: style_guide_committee
Created At: 2023-07-01T11:10:00Z

Guideline: Reporting Statistical Significance

Guideline ID: RSFG_STAT_SIG_PHRASING_001
Applies To: Comparisons where a statistical test (e.g., t-test) was performed.
Standard:
- If significant (p < 0.05): The difference was statistically significant (p = {p_value_formatted}). or Product A was rated significantly higher/lower... (p = {p_value_formatted}).
- If not significant (p >= 0.05): The difference was not statistically significant (p = {p_value_formatted}).
Example: The difference in overall liking between SLHB5 (7.10) and SLHB8 (7.05) was not statistically significant (p = 0.80).
Notes for Pihu: Format p-value to standard decimal places (e.g., two or three). Use the term "statistically significant."
Created By: style_guide_committee
Created At: 2023-07-01T11:15:00Z

E. Provenance & Explainability Template Entry Form (Example: For explaining a benchmark application)

Template ID: PROV_BENCHMARK_APPLIED_001
Template Type: BUSINESS_RULE_APPLIED
Text Template: The '{benchmarkName}' (ID: {benchmarkId}) was applied to evaluate {productName}'s score for {attributeName}. This benchmark defines '{benchmarkBandLabel}' as scores {benchmarkConditionDescription} (e.g., between {minScore} and {maxScore}). {productName}'s score of {scoreValue} falls into this band.
Variables: benchmarkName, benchmarkId, productName, attributeName, benchmarkBandLabel, benchmarkConditionDescription, minScore, maxScore, scoreValue
Notes: Used in the "Show Details" panel.
Created By: pihu_dev_team
Created At: 2023-08-01T16:00:00Z

F. Pihu Persona & Tone Guideline (Brief example)

Guideline ID: PPT_TONE_OBJECTIVE_001
Aspect: Tone
Guideline: Maintain an objective, analytical, and helpful tone. Avoid speculative language or personal opinions.
Example (Do): Analysis indicates SLHB8 (7.05) exceeded the benchmark (6.80).
Example (Don't): I think SLHB8 did really well, it's a great score!
Created By: ux_writing_team
Created At: 2023-02-10T12:00:00Z

G. Glossary Term Entry Form (Example: For "CLT")

Term ID: GLOSSARY_CLT_001
Term: CLT
Definition: Central Location Test: A sensory evaluation methodology where respondents are brought to a controlled, centralized facility to evaluate products.
Context/Usage: Often mentioned when describing study methodology.
Created By: sensory_training_lead
Created At: 2023-01-20T17:00:00Z

2) Actual Data Example in JSON (for Module D - with Reporting Standards)¶

{
  "moduleName": "Explainability, Communication & Reporting Standards",
  "version": "1.1", // Incremented version
  "lastUpdated": "2024-01-25T11:00:00Z",
  "resources": {
    "generalVocabulary": [
      {
        "termId": "VOCAB_COMPARE_001", "canonicalTerm": "compare",
        "synonymsAndUserPhrases": ["vs", "versus", "difference between"],
        "intentAssociation": "COMPARISON_ANALYSIS"
      }
      // ... more vocab
    ],
    "statisticalTermDefinitions": [
      {
        "termId": "STAT_PVALUE_001", "termName": "p-value", "status": "Active",
        "plainEnglishDefinition": "Probability that an observed difference is due to chance. Small p-value (e.g., <0.05) suggests a real, statistically significant difference.",
        "tagTasteExamplePattern": "The difference was {significanceDescriptor} (p = {pValue}).",
        "visualCue": "Pihu flags p < 0.05 in green, p >= 0.05 in grey.",
        "notesForPihu": "State alpha level (e.g., 0.05). 'Significant' != 'important' or 'large effect'."
      }
      // ... other stat terms
    ],
    "standardInterpretationPhrases": {
      "BENCHMARK_COMPARISON_POSITIVE_EXCEEDS": [ // Grouped by response type/context
        {
          "phraseId": "SIP_BENCH_EXCEEDS_001",
          "template": "For {attributeName}, {productName} scored {scoreValue} (out of {scaleMax}), which **Exceeds Expectations** based on the TagTaste benchmark target of {benchmarkTargetValue}. This indicates strong positive performance.",
          "variables": ["attributeName", "productName", "scoreValue", "scaleMax", "benchmarkTargetValue"]
        }
      ],
      "CATA_TOP_LIST": [
        {
            "phraseId": "SIP_CATA_TOP_LIST_001",
            "template": "For {productName}, the main {attributeName} mentioned by {demographic_group_description} were: {notes_list_formatted}.",
            "variables": ["productName", "attributeName", "demographic_group_description", "notes_list_formatted"]
        }
      ]
      // ... more phrase groups and phrases
    },
    "reportingStandardsAndFormattingGuidelines": [ // NEW SECTION
      {
        "guidelineId": "RSFG_MEAN_SCORE_001", "appliesTo": "Mean scores for sensory attributes",
        "standard": "Report mean scores to two decimal places (e.g., 7.05).",
        "example": "Overall Liking for SLHB8 was 7.05."
      },
      {
        "guidelineId": "RSFG_SCALE_CONTEXT_HEDONIC9_001", "appliesTo": "Score from a 9-point hedonic scale",
        "standard": "When first mentioning, state: \"(on a 9-point scale where 1=Dislike Extremely and 9=Like Extremely)\".",
        "example": "SLHB8's overall liking score was 7.05 (on a 9-point scale...)."
      },
      {
        "guidelineId": "RSFG_CATA_TOPNOTES_001", "appliesTo": "Reporting top CATA notes",
        "standard": "List top 3-5 notes with % mentions. Format: \"Note Name (XX%)\".",
        "example": "Main aroma notes: Coffee-Roast (52%), Dairy (30%)."
      },
      {
        "guidelineId": "RSFG_STAT_SIG_PHRASING_001", "appliesTo": "Statistical comparisons",
        "standard": "If p < 0.05: \"The difference was statistically significant (p = {formatted_p_value}).\" Else: \"The difference was not statistically significant (p = {formatted_p_value}).\"",
        "example": "Difference... was not statistically significant (p = 0.80)."
      }
      // ... other reporting standards (N-size, decimal for p-value, etc.)
    ],
    "provenanceAndExplainabilityTemplates": {
      "BUSINESS_RULE_APPLIED": [
        {
          "templateId": "PROV_BENCHMARK_APPLIED_001",
          "text": "The '{benchmarkName}' (ID: {benchmarkId}) was applied to evaluate {productName}'s score for {attributeName}. This benchmark defines '{benchmarkBandLabel}' as scores {benchmarkConditionDescription}.",
          "variables": ["benchmarkName", "benchmarkId", "productName", "attributeName", "benchmarkBandLabel", "benchmarkConditionDescription"]
        }
      ]
      // ... other provenance templates
    },
    "pihuPersonaAndTone": {
      "personaName": "Pihu - Your Smart Sensory-Data Assistant",
      "voiceCharacteristics": ["Helpful", "Precise", "Data-driven", "Approachable"],
      "communicationStyle": { /* ... */ }, "dos": [/* ... */], "donts": [/* ... */]
    },
    "glossaryTerms": [
        { "termId": "GLOSSARY_CLT_001", "term": "CLT", "definition": "Central Location Test..." }
        // ... other glossary terms
    ]
  }
}

3) JSON Structure (for Module D - with Reporting Standards, Simplified Comments)¶

{
  "moduleName": "Explainability, Communication & Reporting Standards",
  "version": "1.1",
  "lastUpdated": "YYYY-MM-DDTHH:MM:SSZ",
  "resources": {
    "generalVocabulary": [ /* ... */ ],
    "statisticalTermDefinitions": [ /* ... */ ],
    "standardInterpretationPhrases": { /* Grouped by response type, then array of phrases */ },
    "reportingStandardsAndFormattingGuidelines": [ // NEW: Array of guideline objects
      {
        "guidelineId": "STRING_UNIQUE_ID",
        "appliesTo": "STRING_DESCRIPTION_OF_CONTEXT", // e.g., "Mean scores", "CATA lists"
        "standard": "TEXT_AREA_STRING_RULE",         // The rule itself, e.g., "Report to 2 decimal places."
        "example": "NULLABLE_STRING_ILLUSTRATION",
        "notesForPihu": "NULLABLE_TEXT_AREA_STRING"  // Specific instructions for Pihu
      }
    ],
    "provenanceAndExplainabilityTemplates": { /* Grouped by template type, then array of templates */ },
    "pihuPersonaAndTone": { /* ... */ },
    "glossaryTerms": [ /* ... */ ]
  }
}

Team Tasks for the new "Reporting Standards" section in Module D:

Sensory/Analytical/Reporting Team (Content Owners):
- Compile a list of all standard ways TagTaste reports specific types of data (mean scores, percentages, CATA lists, significance, scale explanations, etc.).
- Define the rules for numerical precision (decimal places).
- Standardize phrasing for common reporting scenarios.
- This becomes the "style guide" for Pihu's numerical and textual output.
Technical Team (Implementation):
- The Response Synthesis Agent will be the primary consumer of this new section.
- When it has a piece of data (e.g., a mean score of 7.05342), it will consult these guidelines to format it (e.g., to "7.05").
- When it needs to explain a scale, it will use the standard descriptor from here.
- When listing CATA notes, it will use the specified format.