Language & Semiotic Studies · Vol. 11(4) 2025, pp. 636–660 · De Gruyter
Allocution, Sentencing, and Viewers' Comments in YouTube-Mediated Trials of Convicted Young Murderers
An Appraisal-Sentiment Analysis of How Judges, Perpetrators, and the Public Evaluate Morality in High-Profile Courtroom Discourse
Ain Shams University, Cairo, Egypt · Indiana University Bloomington, USA
ALLOCUTION
COURTROOM DISCOURSE
APPRAISAL THEORY
SENTIMENT ANALYSIS
YOUTUBE
SENTENCING
FORENSIC LINGUISTICS
01Introduction & Motivation
▼
What Is This Study About?
Courtroom discourse is a highly conventionalized genre where power, identity, ideology, and affect converge. This study examines a specific intersection: high-profile US trials of young murderers, viewed through two complementary lenses — Martin & White's Appraisal Theory and computational sentiment analysis.
Critically, the study goes beyond the courtroom walls by analyzing 6,000 YouTube comments on the trial videos, treating the public audience as a third discursive actor alongside judges and perpetrators.
Key innovation: The study synthesizes Martin & White's (2005) Appraisal Theory with computational sentiment analysis — combining interpretive depth with scalability — and extends to YouTube viewer discourse, a layer rarely examined in forensic linguistics.
Three Discursive Actors
⚖️
Perpetrators
Allocution phase: plea for leniency
👨⚖️
Judges
Sentencing phase: moral condemnation
💬
YouTube Viewers
6,000 public comments on 12 videos
Example Utterances — Hover for Analysis
● Perpetrator
"I'm very sorry for the pain I caused. I regret all the decisions I made. If I were to get a second chance, I would do everything in my power to try to help others."
This statement clusters AFFECT (regret, sorrow) with positive Social Esteem — Tenacity. The perpetrator frames future reform potential as evidence of moral capacity. This is a classic allocution strategy: admit fault while projecting rehabilitative identity.
Hover for analysis ↑
● Judge
"There can be no mercy for you. You methodically planned, researched, conducted reconnaissance, and executed your hateful crimes. You will never see the light of day as a free man ever again."
Strong negative Propriety (Social Sanction) paired with Force Intensification. The premeditation narrative ("methodically planned") is a recurring pattern in judicial sentencing remarks — converting legal determination into moral condemnation.
Hover for analysis ↑
● YouTube Viewer
"They don't want your rehearsed apology… they want their kids back & loved ones you took."
Negative Veracity (doubting sincerity of allocution) combined with Unhappiness (grief for victims). This comment type is among the most frequent — viewers who reject the performative nature of the allocution while simultaneously expressing mourning.
Hover for analysis ↑
● YouTube Viewer
"evil to the core, killing a human for a low grade which rightfully deserved."
Negative Propriety expressed through moralized retribution language. "Evil to the core" is an example of invoked rather than inscribed JUDGMENT — the word "evil" carries a concentrated moral damnation without explicit reasoning.
Hover for analysis ↑
02Theoretical Framework: Appraisal Theory
▼
Martin & White's (2005) Appraisal Theory (AT) extends Halliday's Systemic Functional Linguistics to address evaluative language. This study focuses on two of AT's three systems: ATTITUDE and GRADUATION.
Interactive Appraisal Explorer — Click a Category
AFFECT
Emotional responses: happiness, security, satisfaction
JUDGMENT
Moral evaluation of human behavior & character
APPRECIATION
Evaluation of objects, texts & phenomena
Click a category above to explore its subcategories and examples from the corpus.
GRADUATION System
FORCE — Scaling Intensity
Intensification: scaling qualities/processes — "pure evil," "extremely sorry," "very deeply regret"
Quantification: measuring extent — "stabbed her 114 times," "the lives stolen"
Corpus pattern: Force dominates at 79% of all Graduation instances across all speaker types.
FOCUS — Sharpening/Softening
Sharpening: "I accept full responsibility" — defines guilt sharply with no ambiguity
Softening: "I didn't mean for this to happen" — mitigates moral agency
Corpus pattern: Focus accounts for 21% — more visible in judges' formal sentencing (sharpening) and perpetrators' mitigation attempts (softening).
Linking AT to Sentiment Analysis
The study innovatively bridges AT's AFFECT subsystem with computational sentiment labels:
Insecurity (AT)
→
Fear (Sentiment)
"I feel terrified…"
Unhappiness (AT)
→
Sadness (Sentiment)
"I am heartbroken…"
Displeasure (AT)
→
Anger (Sentiment)
"ROT IN JAIL…"
03Data: Six Trials, Six Perpetrators
▼
6
High-profile US trials (past 6 years)
12
YouTube videos (6 allocution + 6 sentencing)
6,050
Words in courtroom corpus
6,000
YouTube comments analyzed
382K
Total words in YouTube corpus
14–20
Age range of perpetrators
The Six Cases
All trials selected on criteria of public accessibility and availability of allocution + sentencing phases on YouTube.
| Perpetrator | Crime | Age | Sentence | Allocution type |
|---|---|---|---|---|
| P1 (Nikolas Cruz) | Mass shooting — 17 killed, Parkland FL | 19 | Life (consecutive) | Verbal |
| P2 (Ethan Crumbley) | Oxford High School shooting — 4 killed | 15 | Life sentence | Verbal |
| P3 (Aiden Fucci) | Murdered 13-year-old Tristyn Bailey | 14 | Life sentence | Written (read aloud) |
| P4 (Jeremy Goodale) | Co-murdered Spanish teacher Nohema Graber | 16 | Life + parole at 25 yrs | Verbal |
| P5 (Willard Miller) | Co-murdered Spanish teacher Nohema Graber | 17 | Life + parole at 25 yrs | Verbal |
| P6 (Payton Gendron) | Mass shooting — 10 Black victims, Buffalo NY | 18 | Life sentence | Verbal |
Selection rationale: All trials are publicly accessible via YouTube and occurred within the past 6 years, ensuring social recency. All perpetrators were minors or young adults at the time of the crime, making age-related mitigation arguments central to the allocution discourse.
04Methodology
▼
Data Collection
- 12 video transcripts manually transcribed
- 500 most recent comments per video downloaded via Google API Client Python library
- Total: 6,000 YouTube comments across 12 videos
Annotation Tool
A custom ChatGPT tool "Attitudinal Analysis" was developed, trained on AT papers and Grammatics.com (P.R.R. White's AT website), to assist with AT annotation. All annotations were then manually verified and corrected. The tool handled labeling; interpretation was human-led.
Sentiment Analysis
Applied Hartmann's (2022) Emotion English DistilRoBERTa-base model from HuggingFace — a fine-grained emotion classifier producing labels: sadness, anger, fear, joy, surprise, disgust, neutral.
Analysis Pipeline
Collect Data
12 videos transcribed + 6,000 comments via Google API
→
AT Annotation (ChatGPT Tool)
ATTITUDE + GRADUATION labels assigned sentence by sentence
→
Manual Review & Correction
Human expert validation of all annotations
→
Sentiment Analysis (HuggingFace)
Fine-grained emotion classification of all data
→
Cross-Analysis
Compare AT + sentiment across all 4 discourse contexts
Mixed-method design: Qualitative (close reading + AT annotation) + quantitative (frequency distributions + computational sentiment) — each layer validates and extends the other.
05Attitudinal Analysis: In-Court & YouTube Discourse
▼
Overarching finding: JUDGMENT dominates all four contexts — allocutions, sentencings, allocution comments, sentencing comments — confirming that moral evaluation is the primary discursive function in high-profile murder case discourse.
Figure 1: Frequency of Attitudinal Evaluations — In-Court Discourse
Perpetrators' Allocutions
Dominant AFFECT patterns
Allocutions are emotionally saturated: deep regret, sorrow, and emotional distress are foregrounded. This contrasts sharply with formal sentencing language.
JUDGMENT strategies in allocutions
Negative Propriety (self-directed)
"I did a terrible thing" / "I have done terrible things" — admitting moral failing, aligning with court's condemnation
Positive Tenacity (future-directed)
"If I were to get a second chance, I would help others" — projecting reform capacity to mitigate punishment
APPRECIATION: minimal, victim-directed
Sparse positive valuation of victims' lives — acknowledging impact without disputing guilt
Word Cloud: Top Content Words in Allocutions
Size reflects relative frequency in allocution transcripts
Judges' Sentencing Remarks
JUDGMENT distribution in sentencings
Judges perform a "unitary voice of condemnation" (Heffer 2008): evaluations are formal, declarative, and moralized — not hedged or emotionally displayed.
Exemplary sentencing strategies
● Judge (Negative Propriety)
"Given the manner in which you methodically planned, researched, conducted reconnaissance, and executed your hateful crimes…"
Premeditation narrative converts legal fact into moral failing. "Methodically" and "hateful" are Force-intensified evaluative choices that construct a picture of deliberate evil — not impulsive youth.
Hover for analysis ↑
● Judge (Positive Appreciation — victim)
"For the murder of Roberta Drury, a vibrant 32-year-old young woman, a daughter, a dedicated sister, and friend."
Judges consistently invoke positive APPRECIATION of victims to amplify the moral gravity of the offense. Enumerating social roles (daughter, sister, friend) transforms a legal case into a human tragedy.
Hover for analysis ↑
Word Cloud: Top Content Words in Sentencings
Size reflects relative frequency in sentencing transcripts
YouTube Comments on Allocution Videos
AFFECT distribution in allocution comments
JUDGMENT distribution in allocution comments
Viewers distrust the allocution. Negative Veracity is the second-most common JUDGMENT — viewers widely interpret the leniency plea as strategic performance, not genuine remorse: "the smirk on his face as he fakes an apology."
APPRECIATION: Predominant Negative Valuation
YouTube Comments on Sentencing Videos
AFFECT distribution in sentencing comments
JUDGMENT distribution in sentencing comments
Judges also criticized. Negative Capacity extends to the judges themselves — "this judge is too incompetent for a trial this serious" — and parents: "who buys their kids guns!" The public uses YouTube as a space for systemic critique, not just individual condemnation.
Figure 3: Attitudinal Evaluations in YouTube Comments
06Graduation Analysis — Force & Focus Across All Speakers
▼
Overall distribution: Force categories (Intensification + Quantification) account for 79.15% of all Graduation instances. Focus (Sharpening + Softening) = 20.85%.
Figure 4: Graduation Distribution Across All Speakers
Interactive Graduation Comparison
The stacked bars below visualize how each speaker type deploys Graduation strategies differently:
Intensification
Quantification
Sharpening
Softening
Perpetrators (Allocution)
42%
28%
18%
12%
Judges (Sentencing)
22%
30%
44%
4%
YT Alloc. Commenters
38%
40%
20%
2%
YT Sentencing Commenters
35%
38%
22%
5%
Perpetrators: Intensification-Dominant
Express remorse with emotional amplification ("very deeply regret," "so sorry") to appear genuine. Softening appears when mitigating responsibility: "I didn't mean for this to happen."
Judges: Sharpening-Dominant
Formal sentencing demands clarity — sharpening defines guilt without ambiguity. "The court imposes a mandatory life sentence without the possibility of parole" leaves no room for softening.
07Sentiment Analysis — Four Contexts Compared
▼
The HuggingFace DistilRoBERTa emotion classifier was applied across all four discourse contexts. Figure 5 reveals clear domain-specific emotional signatures.
Figure 5: Emotion Distribution Across All Four Contexts
Cross-context contrast: Perpetrators → sadness. Judges → neutrality. Public on allocutions → anger. Public on sentencings → sadness. The public shifts emotional register depending on whether they are watching a remorse performance (anger/disbelief) or a formal verdict (grief/mourning).
08Key Findings, Contrast & Future Directions
▼
Cross-Actor Comparison
| Dimension | 🔴 Perpetrators | 🔵 Judges | 🟡 YouTube Viewers |
|---|---|---|---|
| Dominant AT category | AFFECT (regret/sorrow) + neg. Propriety (self-directed) | JUDGMENT — neg. Propriety (Other-directed moral condemnation) | JUDGMENT — neg. Propriety (78–83% of comments) |
| Dominant sentiment | Sadness (~62%) | Neutrality (~58%) | Anger (allocution comments); Sadness (sentencing comments) |
| Graduation strategy | Intensification (amplify remorse); Softening (mitigate agency) | Sharpening (define guilt clearly); Quantification (enumerate harm) | Quantification (scale severity); Intensification (moral outrage) |
| Toward the allocution | Strategic self-presentation: reform narrative | Reception varies — may acknowledge or dismiss | Predominantly skeptical — Negative Veracity (21%) |
| Toward the victim(s) | Positive APPRECIATION — acknowledge loss | Positive APPRECIATION — enumerate victims' value to community | Unhappiness — grief, solidarity, mourning |
| Toward legal system | Implicit appeal — seek lenient judgment | Align with societal moral norms | Mixed: some celebrate, many critique (neg. Capacity toward judges) |
Key Findings
Finding 1
JUDGMENT Dominates EverywhereAcross all four discourse contexts, JUDGMENT — especially negative Propriety — is the most deployed AT category. Moral evaluation is the organizing function of courtroom-adjacent discourse.
Finding 2
AFFECT Belongs to the AllocutionAffectual evaluations (regret, sorrow, insecurity) are disproportionately concentrated in perpetrators' allocutions. This is the space where emotional display is both permitted and strategically necessary.
Finding 3
Public Distrusts the Leniency PleaNegative Veracity is the second-most common JUDGMENT in allocution comments. Viewers read allocutions as performance designed for mitigation, not genuine contrition.
Finding 4
Judges: Formal Condemnation MachineJudicial sentencing is a "unitary voice of condemnation." Sharpening + Quantification + negative Propriety construct an unambiguous moral narrative around premeditation and societal harm.
Methodological Finding
AT + Sentiment Validates MutuallyAFFECT's unhappiness maps to sentiment sadness; AFFECT's displeasure maps to anger. The two systems cross-validate, with sentiment analysis extending coverage to large-scale YouTube data where manual AT annotation is not feasible.
Methodological Finding
YouTube as Extended Jury RoomThe public uses YouTube comments to perform collective moral adjudication — passing judgment on perpetrators, judges, parents, and the legal system. Comment discourse is as morally evaluative as in-court discourse.
Limitation
Small In-Court Corpus6,050 words from 6 trials limits generalizability of in-court findings. Frequency distributions should be interpreted with caution; patterns rather than absolute proportions are the key takeaway.
Limitation
US-Centric, English-OnlyAll cases are from the US adversarial legal system. Findings cannot be directly extrapolated to inquisitorial systems or non-English courtroom discourse.
Future Work
Multimodal AnalysisAuthors explicitly call for nonverbal cue analysis — body language, prosody, gaze — during allocutions. "I regret everything" reads differently when paired with a smirk or tears.
Future Work
Cross-System ComparisonComparing US adversarial allocutions with inquisitorial system equivalent discourse (e.g., UK, France) would test how legal-procedural affordances shape evaluative language.
Future Work
Platform-Comparative StudyComparing YouTube comments with X/Twitter, Reddit, and TikTok reactions would reveal how platform affordances and community norms shape the moral evaluation of courtroom discourse.
09References & Publication Details
▼