Tencent Cloud OCR General High Accuracy Version
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md to install @tencent-adm/tencentcloud-ocr.
About this skill
Problem Context
In engineering workflows, processing unstructured document data is a frequent and challenging task. Developers often need to extract text from scanned documents, photos, screenshots, or PDFs for archiving, analysis, or system integration. Traditional OCR solutions may suffer from issues like insufficient accuracy, poor handling of complex layouts or low-quality images, and an inability to effectively recognize specific structures (such as resumes).
Core Capabilities and Implementation
This skill encapsulates Tencent Cloud's GeneralAccurateOCR API, providing the following core capabilities:
* High-Accuracy General Text Recognition: Extracts text content from images with precision, returning structured text even for images of moderate quality.
* Direct PDF File Processing: Supports recognizing content from a specified page (PdfPageNumber) of a PDF file directly via the IsPdf parameter, without needing to convert the PDF to images first.
* Granular Position and Confidence Data: Optionally enables the IsWords parameter to return bounding box coordinates and confidence scores for each recognized character, which is valuable for scenarios requiring precise location analysis or quality assessment.
* Structured Information Extraction Extension: Beyond pure text extraction, the skill offers the capability to perform structured extraction on multilingual resumes, transforming unstructured resume images or PDFs into formatted data.
The skill's workflow is straightforward: it initiates requests via the scripts/main.py script, primarily by constructing JSON parameters containing ImageBase64 or ImageUrl to call the Tencent Cloud API. The script automatically handles the detection and injection of the UserAgent parameter for channel tracking purposes, without affecting the main OCR process.
Applicable Boundaries and Considerations
Before use, it's important to note the following:
* Single-Page Processing: For PDFs, each request can only recognize a single specified page. Recognizing multiple pages requires sequential calls.
* Input Limitations: The Base64 size of the image must not exceed 10MB. Ensure a stable network environment for accessing the Tencent Cloud API endpoints.
* Service Dependency: The skill's operation is entirely dependent on the Tencent Cloud OCR service, requiring valid API credentials (TENCENTCLOUD_SECRET_ID and TENCENTCLOUD_SECRET_KEY) and a purchased service plan.
* Non-Real-Time Scenarios: As a server-side recognition service, network latency and API rate limits may affect response speed, making it unsuitable for real-time scenarios with millisecond latency requirements.
Use Cases
- After uploading scanned images of paper contracts or invoices, invoke this skill to precisely extract key textual information like clauses, amounts, and dates for subsequent data entry or auditing.
- When needing to quickly extract body text from a batch of academic papers or reports in PDF format for text analysis or information retrieval, avoiding manual copy-pasting.
- Upon receiving numerous resume images or PDFs from job applicants, use this skill to automatically recognize and structurally extract fields such as name, education history, and years of work experience to import into an ATS for initial screening.
- Perform high-accuracy OCR on blurry document photos from historical archives to restore editable electronic text, used for digitizing materials and archiving content.
Best For
- An archivist responsible for digitizing and filing large volumes of paper documents, who needs to efficiently and accurately obtain text content from the documents.
- A recruitment specialist who processes dozens of resumes weekly and wants to quickly extract structured information from resume images or PDFs for screening.
- An assistant in legal or financial analysis who needs to extract key clauses and data from unstructured document images like contracts or tender documents.
- A researcher or data analyst who needs to batch-process PDF documents to extract plain text for corpus building or content mining.
Related Skills
Transforms LLMs into knowledge base maintainers to incrementally build and maintain a persistent, interlinked Markdown wiki.
A mindmap generation skill specialized in distilling complex information into structured knowledge frameworks for easy understanding and reuse.
BookBone distills books into operational thinking frameworks, extracting core mental models, methodologies, and concepts into structured knowledge tools.
A Python function for local academic paper retrieval engine, supporting multi-platform parallel search, export, and citation analysis, with built-in classic paper index.