Sentiment Monitoring: a Tokeny case study
Search brand or topic sentiment online, analyzing positive/negative distribution and key topics. Follow the recorded task, execution screenshots, actual files and verification notes.
By the Tokeny team · Windows · Tokeny 1.7.8 · gpt-6-sol
Prepare the inputs
Create a separate workspace in Tokeny, download and unpack the sample inputs, then select a model that supports tools. The sample uses synthetic data or the public sources named in the article.
- feedback.csv
Send the task to Tokeny
The original task was executed in Chinese. Copying these instructions reproduces the same task; the results below include an English report.
对提供的合成公开评论样本做舆情分析，逐条标注情感、主题和依据，汇总分布。此例仅验证离线样本分析，不声称覆盖全网监测。 交付文件：sentiment.csv、sentiment.html。 验收要点： - 保留 12 条输入 - 分类可以逐条复核 - 明确样本范围 这是使用合成演示输入或指定公开资料的真实操作案例。请先读取工作空间输入，再实际生成文件并自检。所有写入、脚本与临时服务限于当前工作空间；临时服务只监听 127.0.0.1，执行后关闭；不触达外部收件人、不发布、不付款、不更改系统设置，不调用子智能体。需要公开资料时可读取给定 URL。只启用完成任务必需的工具。另写 report.md（中文）和 report.en.md（英文），简洁、完整，包含输入、处理步骤、真实结果、检查方式和局限；失败或未验证部分如实记录，不能宣称成功。最终回复用中文列出产物、关键结果与待核实项。 排版检查：Word 主标题仅出现一次，标题为黑色且与后续正文或表格保持同页；Excel 表头和长文字不能被窄列截断，合理设置列宽、换行、行高；HTML 在 390px 视口不能整页横向溢出。报告没有固定字数要求，不要为字数反复改写。
The recorded run used an isolated demonstration workspace. Review the commands and file paths before approving operations in your own workspace.
What actually ran
The task was submitted through the Tokeny input box. An online model read the inputs and executed tools. This run covers the local sample task and its actual files. Results and checks are recorded below.
Results and verification
Input
The workspace-provided feedback.csv was read. It contains 12 valid comments with the fields id, rating, and comment. This analysis covers only these 12 synthetic public comments. No external pages were accessed, and no whole-web monitoring or population-level inference was performed.
Processing Steps
- Preserve each input ID, rating, and original comment.
- Assign one sentiment label per comment: positive, negative, neutral, or mixed, using wording and rating together.
- Assign one primary topic and record a reviewable textual rationale for every row.
- Aggregate the sentiment distribution across the 12 valid comments and create CSV and HTML outputs.
Labeling rule: feature requests such as “希望……” (“hope to add/support…”) are neutral when they do not directly express satisfaction or dissatisfaction. Comments containing both praise and a problem are mixed. Explicit problems such as excessive waiting, unclear messages, or inability to find something are negative; the text problem can make a rating-3 comment negative.
Actual Results
| Sentiment | Count | Share |
|---|---|---|
| Positive | 3 | 25.0% |
| Negative | 6 | 50.0% |
| Neutral | 2 | 16.7% |
| Mixed | 1 | 8.3% |
| Total | 12 | 100.0% |
Negative comments are the largest group, with 6 comments (50.0%). They mainly concern large-file import performance, model-configuration comprehension/error messages, finding historical tasks, and save-state visibility after a network disconnect. Positive comments concern table organization, local file preview, and batch conversion. Row-level labels, topics, original text, and rationales are in sentiment.csv; the reviewable HTML presentation is in sentiment.html.
Checks
- Row-by-row comparison of
sentiment.csvagainstfeedback.csvfor all 12 IDs and original comments. - Verify that sentiment counts are 3 + 6 + 2 + 1 = 12 and shares total 100.0%.
- Verify that the HTML includes 12 detail rows, a scope statement, the distribution summary, and rationales.
- Verify responsive behavior: the table uses an inner scroll container on narrow screens and the page has no fixed-width layout; opening it at a 390px browser viewport is still required for a direct visual check.
Limitations
This is an offline analysis of a synthetic demonstration sample. It does not claim whole-web coverage and cannot represent the broader user population. The sample is small, with unknown source and sampling mechanism. Sentiment and topic labels use human-readable rules and may have boundary disagreements. Each comment receives one primary topic, so secondary needs may be omitted. No external fact checking, model-consistency evaluation, or time-trend analysis was performed.
Download and inspect the outputs
The archive contains the files created during this run. Use these checks when adapting the example to your own data.
- Reconcile the reported results against the downloadable sample inputs.
- Open the generated files and check the contents relevant to your task.
- Review the limitations and any unverified steps in the report.